CN115244149B - Paint manufacturing method, color data prediction method and computer color matching system - Google Patents

Abstract

The invention provides a method for producing a paint, which is used for obtaining various colors including bright colors with difficult optical characteristics to predict, is not influenced by the proficiency of operators, and is based on computer color matching that can finish color matching through less trial production times. The method for producing the paint includes the steps S101 to S111, wherein at least 1 or more composition data Y and corresponding color data X are registered in a database, and the computer is used to calculate the color by using the data registered in the database.

Description

Paint manufacturing method, color data prediction method and computer color matching system

Technical Field

The invention relates to a manufacturing method of paint and a method for predicting color data. More specifically, the present invention relates to a method for producing a paint by computer toning and a method for predicting color data of a coating film.

The present invention also relates to a computer tinting system, a system for predicting color data of a paint film, and application software for controlling and operating these systems.

Background

In recent years, from the viewpoints of diversification of personal preferences, improvement of beauty, and the like, as various industrial products, particularly as colors of automobiles, there has been an increase in bright colors based on bright pigments such as metal powder or bright mica. When repairing such a bright color, a repair coating containing a bright pigment is usually applied to the repaired portion.

When repairing a bright color, not only the hue but also the brightness needs to be adjusted so that the repaired portion and the like are not obvious. However, when a repair coating material satisfying both the hue and the brightness is prepared, even a skilled worker may have to repeatedly undergo trial and error to trial and produce a plurality of repair coating materials. Further, it is necessary for a less experienced operator to repeatedly undergo experiments and failures by a skilled operator to try and manufacture various repair paints, which is a very difficult operation.

Since the worker repeatedly undergoes the test and failure, the working time becomes long, and problems such as cost and disposal are caused by the manufactured repair paint which cannot be used, in addition to problems such as a long period of time during repair and a reduction in efficiency.

Accordingly, various studies have been made to produce a repair coating material quickly and efficiently without being affected by factors such as the proficiency of the operator. As one of them, a computer Color Correction (CCM) system has been studied in which a computer is used to obtain color data of a target color obtained by color measurement by a colorimeter and color data of a color sample of a known matching composition, and the matching composition of the target color is obtained. However, for example, metallic paint containing a brightening pigment has an angle dependence of spectral reflectance, and it is difficult to cope with the conventional CCM.

In addition, in the conventional color matching of the repair paint, it is often necessary to further perform fine color matching based on the fact that the color matching of the paint is approximately obtained (sometimes simply referred to as approximately matching), and particularly in the preparation of the repair paint which is a bright color, there is a limit in the accuracy and reproducibility of the approximately matching by the conventional CCM, and there is still a problem in rapidly and efficiently preparing the repair paint because man-hours in which an operator repeatedly experiences a test and failure of color matching often occur.

Patent document 1 discloses a method in which, in a color matching system using a computer, the total spectral reflectance of an unknown color plate is determined by a spectrophotometer, the reflectance data is transmitted to the computer, and the computer performs mathematical processing on pre-recorded data indicating the K value (light absorption coefficient) and S value (light scattering coefficient) of a pigment, thereby performing theoretical color matching.

Patent document 2 discloses a computer color matching device comprising a colorimeter, a microscopic shine meter, and a computer in which color data and microscopic shine data corresponding to the respective paint combinations, color characteristic data and microscopic shine characteristic data of a plurality of primary color paints are registered and color matching calculation logic functions, and a color matching method using the computer color matching device.

Patent document 3 discloses a color matching method comprising a colorimeter, a microscopic shine-by-sample color chart, and a computer in which a plurality of paint combinations, color data corresponding to the paint combinations, microscopic shine-by-sample color chart, and color characteristic data and microscopic shine-by-sample color chart of a plurality of primary color paints are registered, and color matching calculation logic functions.

Patent document 4 discloses a method of mixing colors of metallic paint, which is useful as the final fine color of a repair paint, wherein, when changing the mixing of metallic paint, the mixing conversion index of each metallic primary color paint is used, which is based on the characteristic of the ratio of the front color to the back color of the paint and the brightness of the front color does not change, depending on the difference between the front color and the back color at the time of viewing.

Patent document 5 discloses a method of determining or predicting visual texture parameters of a paint by an artificial neural network based on color components used in a paint dispensing composition. However, the optical characteristics such as spectral reflectance are determined or predicted by a physical model of a known paint dispensing composition. When the coating material for repairing is prepared for bright color, which is difficult to predict the optical characteristics, more tests and failures are predicted to be required.

Patent document 6 describes a method for searching for a color, in which a color database is created for determining a color having a desired texture or a color belonging to a desired color class in a bright coating film or the like, and the database includes a step of learning a neural network by using data such as spectral reflectance or microscopic bright feeling of the color. However, no disclosure is made regarding specific operations and the like when applying a coating for bright color modulation repair, which makes it difficult to predict optical characteristics.

Patent literature

Patent document 1: japanese patent publication No. Sho 50-28190

Patent document 2: japanese patent application laid-open No. 2001-221690

Patent document 3: international publication No. 2002/004567

Patent document 4: japanese patent application laid-open No. 2004-224966

Patent document 5: japanese patent application laid-open No. 2019-500588

Patent document 6: international publication No. 2008/156147

Disclosure of Invention

The 1 st object of the present invention is to provide a method for producing a paint, which is used for obtaining a paint color of various colors including a bright color whose optical characteristics are difficult to predict, and which is not affected by the proficiency of an operator, but is based on a computer color tone that can finish color mixing with a small number of trial production times.

The present invention has as its object to provide a method for predicting color data of a coating film, which can predict color data of a coating film of various compositions containing a luminescent pigment or the like with high accuracy.

The present invention has as its object to provide a computer-based color matching system for preparing a color-coated paint of various colors including a bright color whose optical characteristics are difficult to predict, which can finish color matching with a small number of trial-and-production times without being affected by the proficiency of the operator.

The 4 th object of the present invention is to provide a system for predicting color data of a coating film, which can predict color data of a coating film of various compositions containing a luminescent pigment and the like with high accuracy.

The present inventors have made intensive studies to solve the above problems, and have found that the above problems can be solved by the following configuration, and have completed the present invention.

Specifically, as described below.

Item 1: a method for producing a paint, based on computer toning using a device having a database and a computer,

the database is registered with the composition data Y1 to Yn of 1 or more compositions C1 to Cn (n is an integer of 2 or more) and the color data X1 to Xn corresponding to the composition data,

color matching calculation logic using data registered in the database functions in the computer,

the method includes the following steps S101 to S111.

S101, using the data registered in the database, inputting learning data into the computer,

s102 is a process of performing machine learning using the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model matching the composition data Y from the color data X,

s103 is a step of acquiring color data (Xp) of a target color whose matching composition Yp is unknown,

s104 is a step of inputting the color data Xp to the computer,

S105 is a step of obtaining predicted composition data Ya1 predicted from the color data Xp as composition data containing 1 or more of the compositions C1 to Cn as components by using the learned artificial intelligence model and/or a prediction formula other than the artificial intelligence model,

s106 is a step of obtaining predicted color data Xa1 predicted from the predicted cooperative constituent data Ya1 by using the learned artificial intelligence model and/or a predictive expression other than the artificial intelligence model, and comparing the predicted color data Xa with the color data Xp to determine whether the color data Xa is acceptable or not,

s107, when the composition is not acceptable in the step S106, obtaining predicted composition data Yai which is predicted from color data Xp and is different from the predicted composition data up to now as composition data containing 1 or more of the compositions C1 to Cn by using a predictive expression other than the learned artificial intelligence model and/or artificial intelligence model, obtaining predicted color data Xai predicted from predicted composition data Yai by using a predictive expression other than the learned artificial intelligence model, and repeating a step of judging whether or not the composition is acceptable by comparing the composition data with the color data Xp until the composition is acceptable,

S108 is a step of acquiring the fit composition data Yap1 when the data is qualified in either the step S106 or the step S107,

s109 is a step of modulating the actual candidate paint CMap1 based on the acceptable composition data Yap1 to obtain a coated sheet of the actual candidate paint CMap1 and obtaining actual measurement color data Xap1,

s110 is a step of determining whether or not the paint is acceptable by comparing the color data Xp with the actual measurement color data Xap1 and/or comparing the target color with the color of the paint sheet of the actual candidate paint CMap1,

s111 is a step of repeating the steps S105 to S110 or the steps S107 to S110 until the failure in the step S110 is reached.

Item 2: a method for producing a paint, wherein a database is used for registering at least 1 or more composition matching data Y and corresponding color data X based on a computer color matching using a device having the database and a computer, and the computer is operated by a color matching calculation logic using the data registered in the database, comprising the following steps S201 to S211.

S201 is a step of inputting learning data to the computer by using the data registered in the database,

S202 is a step of performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the matching composition data Y,

s203 is obtaining target color data X of the target color _t In the process of (a) and (b),

s204 is to input the target color data X to the computer _t In the process of (a) and (b),

s205, obtaining the color data approximate to the target color data X by searching with a computer _t Is to search for color data X _n1 Corresponding to the retrieved color data X _n1 Is approximately matched to form data Y _n1 At the same time for the target color data X _t And the search color data X _n1 A step of comparing the results to determine whether the test result is acceptable or not,

s206, when the color data is not qualified in the step S205, obtaining the predicted target color data X by using the computer _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni A step of comparing the results to determine whether the test result is acceptable or not,

s207, when the color data is not acceptable in the step S206, obtaining the predicted target color data X by using the computer _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni Comparing, repeating the step of judging whether the product is qualified or not until the product is qualified,

s208 is to obtain the qualified matching composition data Y when the product is qualified in any one of the steps S205 to S207 _C1 In the process of (a) and (b),

s209 is to compose data Y according to the qualified coordination _C1 Preparation of actual candidate paint CM _Ci Obtaining the actual candidate paint CM _Ci Obtaining measured color data X by coating a plate _Ci In the process of (a) and (b),

s210 is a color data X _t And the measured color data X _Ci And/or as a comparison of the target color with the actual candidate paint CM _Ci A step of comparing the colors of the coated plates to determine whether the plates are acceptable or not,

s211 is a step of repeating the steps S206 to S210 when the step is not acceptable in the step S210.

Item 3: a method for predicting color data of a coating film, wherein a device having a database in which at least 1 or more kinds of composition data Y and corresponding color data X are registered and a computer in which color matching calculation logic using the data registered in the database is used, the method comprising the following steps S301 to S309.

S301 is a step of inputting learning data to the computer by using the data registered in the database,

s302 is a step of performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the matching composition data Y,

s303 is paint CM for obtaining color data of predicted paint film _t Is matched with the composition data Y _CM S304 is to input the matching composition data Y to the computer _CM In the process of (a) and (b),

s305, according to need, obtaining the matching composition data Y by searching with a computer _CM Is to search for color data X _n1 In the process of (a) and (b),

s306 is a step of, when the corresponding search color data X is not searched in the step S305 _n1 At the time, or when the step S305 is not performed, data Y is composed from the fit using the at least 1 learned artificial intelligence model or the at least 1 learned artificial intelligence model and predictive formulas other than the artificial intelligence model _CM Obtaining predicted color data X _m1 In the process of (a) and (b),

s307 is to obtain the CM coated with the paint according to the need _t Measured color data X of coated board _CM And the predicted color data X _m1 And a step of comparing.

Item 4: the method for producing a paint according to item 1, wherein the step S105 and/or the step S107 includes a step of obtaining predicted blend composition data Ya1 and/or Yai predicted from the color data Xp as blend composition data containing 1 or more of the compositions C1 to Cn as components by using multi-label classification.

Item 5: the method for producing a paint according to item 1 or 4, wherein the predicted blend composition data Ya1 obtained in the step S105 and/or the predicted blend composition data Yai obtained in the step S107 are blend composition data containing 15 or less of the compositions C1 to Cn as components, and the number of metal pigment-containing compositions is 5 or less and the number of pearlescent pigment-containing compositions is 5 or less.

Item 6: the method of producing a paint according to item 2, wherein, when the paint is not acceptable in the step S211, the predicted color data X is obtained _ni And the measured color data X _Ci The difference Δ of (a) is inputted to a computer as a correction coefficient α, and the steps S206 to S211 are repeated.

Item 7: the method for producing a paint according to any one of items 1, 2, and 4 to 6 or the method for predicting color data of a coating film according to item 3, wherein the composition data Y and the corresponding color data X of 1 or more kinds of compositions registered in the database include actual measurement data or include actual measurement data and data calculated from the actual measurement data.

Item 8: the method for producing a coating material according to any one of items 1, 2 and 4 to 7,

in the step S102 or the step S202, the step of generating the learned artificial intelligence model includes:

(i) A step of learning an artificial intelligence model by using, as learning data, 1 or more kinds of composition data Y and color data X relating to a composition containing no luminescent pigment;

and (ii) a step of learning the artificial intelligence model by using, as learning data, 1 or more kinds of blending composition data Y and color data X of the composition containing the lustrous pigment.

Item 9: the method for producing a coating material according to any one of items 1, 2 and 4 to 8,

in the step S102 or the step S202, the step of generating a learned artificial intelligence model includes,

the data for learning is selected from the group consisting of 1 or more data of the content of the light-reflective pigment in the composition, the content of the light-interference pigment, the content of the orientation controlling agent, the content of each color phase of the light-reflective pigment in the composition, the content of each color phase of the light-interference pigment, the content of each color phase of the colorant, and 2 or more total of these contents, and/or the shape data of the coloring material contained in the composition,

And learning the artificial intelligence model.

Item 10: the method for producing a paint according to any one of items 1, 2, and 4 to 9, wherein the color data Xp in the step S103 or the target color data X in the step S203 is obtained by a method for producing a paint _t Color data of a coating film containing a lustrous pigment.

Item 11: the method for producing a paint according to any one of items 1, 2, and 4 to 10, wherein the step of switching to a predictive expression other than the artificial intelligence model is included in the repeating step when the paint fails in the step of determining.

Item 12: the method for producing a paint according to any one of items 1, 2, and 4 to 11, wherein a computer is used for the determination in the step of determining.

Item 13: the method for producing a paint according to any one of items 1, 2, and 4 to 12 or the method for predicting color data of a coating film according to item 3, which is used for repair coating of a vehicle.

Item 14: a computer toning system is provided with:

a database in which match composition data Y1 to Yn of 1 or more compositions C1 to Cn (n is an integer of 2 or more) and color data X1 to Xn corresponding to the match composition data are registered;

And a computer that functions by color matching calculation logic using the data registered in the database,

the system includes the following means S401 to S411.

S401 is means for inputting learning data to the computer by using the data registered in the database,

s402 is means for performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model matching the composition data Y from the color data X,

s403 is a means for acquiring color data Xp of a target color whose matching composition Yp is unknown,

s404 is means for inputting the color data Xp to the computer,

s405 is a means for obtaining predicted composition data Ya1 predicted from color data Xp using the learned artificial intelligence model and/or a prediction formula other than the artificial intelligence model as composition data containing 1 or more of the compositions C1 to Cn as components,

s406 is means for obtaining predicted color data Xa1 predicted from the predicted cooperative constituent data Ya1 by using the learned artificial intelligence model and/or a predictive expression other than the artificial intelligence model, and comparing the predicted color data Xa with the color data Xp to determine whether the color data Xa is acceptable or not,

S407 is a means for obtaining predicted fit composition data Yai predicted from color data Xp and other than the predicted fit composition data up to now as fit composition data containing 1 or more of the compositions C1 to Cn as components by using a predictive expression other than the learned artificial intelligence model and/or artificial intelligence model when the means S406 fails, then obtaining predicted color data Xai predicted from predicted fit composition data Yai by using a predictive expression other than the learned artificial intelligence model and repeatedly judging whether or not the composition data is acceptable by comparing the predicted fit composition data with the color data Xp until the composition data is acceptable,

s408 is a means for acquiring the fit composition data Yap1 when the fit is satisfied in either the means S406 or S407,

s409 is means for modulating the actual candidate paint CMap1 based on the acceptable composition data Yap1 to obtain a coated sheet of the actual candidate paint CMap1 and obtaining actual measurement color data Xap1,

s410 is means for determining whether or not the paint is acceptable by comparing the color data Xp with the actual measurement color data Xap1 and/or comparing the target color with the color of the paint sheet of the actual candidate paint CMap1,

S411 is a means for repeating the means S405 to S410 or S407 to S410 until the pass is reached when the pass is not made in the means S410.

Item 15: a computer toning system is provided with: a database in which at least 1 or more composition matching composition data Y and corresponding color data X are registered; and a computer that functions by using color matching calculation logic of data registered in the database, the system including the following means S501 to S511.

S501 is means for inputting learning data to the computer by using the data registered in the database,

s502 is means for performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the matching composition data Y,

s503 is target color data X for acquiring a target color _t Is characterized in that,

s504 is to input the target color data X to the computer _t Is characterized in that,

s505 is to obtain the color data X approximate to the target color data by searching using a computer _t Is to search for color data X _n1 Corresponding to the retrieved color data X _n1 Is approximately matched to form data Y _n1 At the same time for the target color data X _t And the search color data X _n1 Means for comparing the results to determine whether the test result is acceptable or not,

s506 is to use the computer to obtain the prediction to provide the target color data X when the color data is not qualified in the step S505 _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color of predictionColor data X _ni At the same time for the color data X _t And the predicted color data X _ni Means for comparing the results to determine whether the test result is acceptable or not,

s507 is to use a computer to obtain the predicted target color data X when the color data X is not qualified in the step S506 _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni Means for repeating the means for determining whether the test is acceptable or not until the test is acceptable,

s508 is to obtain the qualified matching composition data Y when the data is qualified in any one of the means S505 to S507 _C1 Is characterized in that,

s509 is the composition data Y according to the qualified coordination _C1 Preparation of actual candidate paint CM _Ci Obtaining the actual candidate paint CM _Ci Obtaining measured color data X by coating a plate _Ci Is characterized in that,

s510 is the color data X _t And the measured color data X _Ci And/or as a comparison of the target color with the actual candidate paint CM _Ci A means for comparing the colors of the coated plates to determine whether the plates are acceptable or not,

s511 is a means for repeating the means S506 to S510 when the means S510 fails.

Item 16: a system for predicting color data of a coating film, comprising: a database in which at least 1 or more composition matching composition data Y and corresponding color data X are registered; and a computer that functions by using color matching calculation logic of data registered in the database, the system including the following means S601 to S609.

S601 is a means for inputting learning data to the computer by using the data registered in the database,

s602 is means for performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the cooperative constituent data Y,

S603 is a paint CM for obtaining color data of a predicted paint film _t Is matched with the composition data Y _CM S604 is to input the matched composition data Y to the computer _CM Is characterized in that,

s605, according to need, obtaining the matching composition data Y by searching with a computer _CM Is to search for color data X _n1 Is characterized in that,

s606 is when no corresponding search color data X is searched in the means S605 _n1 At the time, or when the means S605 is not performed, data Y is composed from the fit using the at least 1 learned artificial intelligence model or the at least 1 learned artificial intelligence model and predictive formulas other than the artificial intelligence model _CM Obtaining predicted color data X _m1 Is characterized in that,

s607, obtaining the CM coated with the paint according to the need _t Measured color data X of coated board _CM And the predicted color data X _m1 Means for comparing.

Item 17: the computer-based color matching system according to item 14 or 15 or the system for predicting color data of a coating film according to item 16, which is provided with an automatic blending means for automatically blending and realizing color matching based on the obtained matching composition data.

Item 18: application software for controlling and operating the system according to any one of items 14 to 17.

According to the present invention, there is provided a method for producing a paint for obtaining a color of various colors including a bright color whose optical characteristics are difficult to predict, which is not affected by the proficiency of an operator, but is based on a computer color tone in which the color tone can be finished with a small number of trial production times.

According to the present invention, there is provided a method for predicting color data of a coating film, which can accurately predict color data of a coating film of various compositions including a luminescent pigment and the like.

According to the present invention, there is provided a computer toning system for preparing a paint for obtaining various colors including a bright color whose optical characteristics are difficult to predict, which is not affected by the proficiency of an operator, but can finish toning with a small number of trial production times.

According to the present invention, there is provided a system for predicting color data of a coating film, which can accurately predict color data of a coating film of various compositions including a luminescent pigment and the like.

Accordingly, by reducing the working time and the number of times of color mixing, the burden on the operator can be reduced, and by reducing the number of paint to be produced, the reduction of waste and the energy saving can be achieved, and stable color mixing and color tone prediction can be performed without being affected by the skill of the operator, and the effect such as improvement of the workability can be obtained, which is industrially extremely useful.

Drawings

Fig. 1 is a schematic configuration diagram showing an embodiment of an apparatus used in the method according to the present invention.

Fig. 2 is a schematic configuration diagram showing another embodiment of an apparatus used in the method according to the present invention.

Fig. 3 is a schematic structural diagram of a neural network in the artificial intelligence model of the present invention.

Fig. 4 is a schematic configuration diagram showing an embodiment of variable angle color measurement using a multi-angle spectrophotometer according to the present invention.

Fig. 5 is a schematic configuration diagram showing another embodiment of variable angle color measurement using a multi-angle spectrophotometer according to the present invention.

Fig. 6 is a flowchart showing an embodiment of a method for producing a paint by computer toning according to embodiment 1 of the present invention.

Fig. 7 is a flowchart showing an embodiment of a method for producing a paint by computer toning according to embodiment 2 of the present invention.

Fig. 8 is a flowchart showing an embodiment of a method for predicting color data of a coating film according to embodiment 3 of the present invention.

Symbol description

1-a database; 2. 21-24-computer; 31. 32-an input device; 41. 42-an output device; 51. 52-a display device; 61. 62-colorimeter; 71. 72-camera instrument; 81. 82- (automatic) blender; 9-neural networks; 91-input layer; 911-input node; 92-hidden layer; 921-hidden node; 93-an output layer; 931-output node.

Detailed Description

The invention comprises the following steps:

(i) The method for producing a paint according to claim 1, comprising a step of generating a learned artificial intelligence model including at least 1 artificial intelligence model for estimating the cooperative composition data Y from the color data X;

(ii) A method for producing a paint according to claim 2, comprising a step of generating a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from matching composition data Y;

(iii) The method for predicting color data of a coating film according to the embodiment 3 includes a step of generating a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the cooperative composition data Y;

(iv) The computer toning system according to embodiment 4, which includes means for generating a learned artificial intelligence model including at least 1 of artificial intelligence models that cooperate with the composition data Y from the color data X;

(v) The computer toning system according to embodiment 5, comprising means for generating a learned artificial intelligence model including at least 1 artificial intelligence model for deducing color data X from the cooperative constituent data Y;

(vi) The system for predicting color data of a coating film according to the embodiment 6, comprising means for generating a learned artificial intelligence model including at least 1 artificial intelligence model for deducing color data X from fitting composition data Y;

and (vii) application software for controlling and operating the system according to the 4 th to 6 th embodiments.

The configuration and operation of the method, system and application software of the present invention will be described in more detail with reference to fig. 1 to 8 by way of the following exemplary embodiments. The means and steps in the present specification are not limited in any way as long as the operations, functions, or steps described in the means and steps can be completed. The present invention is not limited to the embodiments described below, as long as the gist of the present invention is not deviated.

Device and method for controlling the same

In the method for producing a computer-tinting paint and the method for predicting color data of a coating film according to the present invention, a device having a database in which at least color data X and matching composition data Y of a composition are registered and a computer in which color matching calculation logic using data registered in the database is used.

The color data X of the composition includes color data X1 to Xn, which correspond to the respective blend composition data Y1 to Yn of 1 or more compositions C1 to Cn (n is an integer of 2 or more). The composition data Y includes composition data Y1 to Yn of 1 or more compositions C1 to Cn (n is an integer of 2 or more).

The computerized tinting system and the system for predicting color data of a coating film according to the present invention include a device including the database and the computer for causing the color matching calculation logic to function.

Fig. 1 and 2 are schematic configuration diagrams showing embodiments of an apparatus used in a method for producing a paint by computer toning and a method for predicting color data of a coating film according to the present invention. The device is also used as a device provided in the computer color matching system and the system for predicting color data of a coating film of the present invention.

As shown in fig. 1 and 2, the device D includes a database 1 and a computer 2.

In the present invention, the device D may further include 2 or more databases 1, and may further include 2 or more computers 2. In addition, the database 1 and the computer 2 may be integrated.

The device D used in the present invention may further include 1 or more devices having 1 or more functions, such as the input device 31 (and 32), the output device 41 (and 42), the display device 51 (and 52), the colorimeter 61 (and 62), the imaging device 71 (and 72), and the automatic blending device 81 (and 82), as necessary. In addition, more than 1 of these instruments may be integrated with the database and/or computer.

Here, the database 1 may be a known recording device, a server, or the like. In addition, a personal computer, a portable terminal, a smart phone, or the like, which are commercially available, can be used as the computer 2.

As the input devices 31 (and 32), a known input device such as a keyboard, a touch panel, or a reading device may be used, as the output devices 41 (and 42), a known output device such as a printing device or a data writing device may be used, and as the display devices 51 (and 52), a known display device such as a display may be used.

As the colorimeters 61 (and 62), known colorimeters such as a (multi-angle) spectrophotometer, colorimeter, and colorimeter can be used, as the imaging devices 71 (and 72), known imaging devices such as a CCD camera, a solid-state imaging device, and a (near) infrared spectroscopic imaging device can be used, and as the (automatic) blending devices 81 (and 82), known (automatic) blending devices including an electronic balance device and the like can be used.

In the apparatus used in the present invention, the database, the computer, and the devices are connected to each other by communication means such as wired, wireless, or a combination of these or means by means of a recording medium so as to be mutually transmittable and receivable. Examples of the communication means include a combination of 1 or more of various communication networks such as LAN (local area network), WAN (wide area network), internet and telephone network.

Fig. 1 shows an example of a device in which 1 database 1 and 1 computer 2 are connected to each other so as to be able to transmit and receive data. Any 1 or more of the input device 31, the output device 41, the display device 51, the colorimeter 61, the imaging device 71 and the (automatic) blending machine 81 may be connected to the database 1, and any 1 or more of the input device 32, the output device 42, the display device 52, the colorimeter 62, the imaging device 72 and the (automatic) blending machine 82 may be connected to the computer 2. More than 1 of the colorimeters 61 and 62 and the imaging devices 71 and 72 and the (automatic) blending machines 81 and 82 connected to the database 1 or the computer 2 are measured, blended, and the like according to instructions from the database 1 or the computer 2. The measured data and the like are transmitted to the database 1 or the computer 2, and finally the data can be registered in the database 1.

In the apparatus of fig. 1, the database 1 may be formed as a recording device in the computer 2, and in this case, the apparatus is capable of performing prediction of color data of the color mixing and coating film of the computer independently of the database 1 without performing communication. In addition, the database can be maintained by updating (updating) the data registered in the database at an appropriate timing, as needed, whereby the worker can perform the work based on the latest data.

Fig. 2 shows an example of a device in which 2 or more computers 21 to 2X are connected to 1 database 1. Fig. 2 shows an example in which 4 computers 21 to 24 are connected. The database 1 may be configured such that 2 or more databases 1 are communicably connected. By increasing the number of databases 1, the number of computers connectable to be communicable can be increased.

In fig. 2, the database 1 may be connected to one or more of the input device 31, the output device 41, and the display device 51, and may be connected to one or more of a colorimeter, an imaging device, and an (automatic) blender.

In fig. 2, one or more of the input device 32, the output device 42, the display device 52, the colorimeter 62, the imaging device 72, and the (automatic) blending machine 82 may be connected to the computers 21 to 24.

The device D of fig. 2 corresponds to a plurality of computers 2 connected to the database 1 as a server. For example, the database 1 managed by the paint company or the like may be connected to the computer 2 of the operator (user) via a communication line (for example, an internet line, a telephone line, or the like), so that data communication can be performed.

Fig. 3 is a schematic configuration diagram showing the neural network 9 (activity of nerve cells of the brain reproduced by a program) in the artificial intelligence model in which the fitting composition data Y is inferred from the color data X or the artificial intelligence model in which the color data X is inferred from the fitting composition data Y.

Using the data registered in the database, the learning data is machine-learned while the learning data is input to the computer, thereby generating an artificial intelligence model. As shown in fig. 3, the neural network 9 includes 3 processing layers (3 neuron layers) including an input layer 91, a hidden layer 92, and an output layer 93.

The input layer 91 includes at least 1 to i processing elements called input nodes 911 to 91i, and hidden nodes 921 to 92j connected to the hidden layer 92 of the network.

In the method for producing a paint according to claim 1 of the present invention and the computer color matching system according to claim 4 of the present invention, each unit of the input layer 91 of the neural network 9 used in the artificial intelligence model for estimating and matching the composition data Y from the color data X corresponds to 1 or more feature values according to the color data X.

In the method for producing a paint according to claim 2 of the present invention, the method for predicting color data of a paint film according to claim 3 of the present invention, the computer toning system according to claim 4 of the present invention, and the system for predicting color data of a paint film according to claim 6 of the present invention, each unit of the input layer 91 of the neural network 9 used in the artificial intelligence model for estimating color data X from blending composition data Y corresponds to 1 or more feature amounts related to the blending composition data Y.

The hidden layer 92 has at least 1 to j processing elements called hidden nodes 921 to 92j, and output nodes 931 to 93k connected to the output layer 93 of the network. The hidden layer 92 (hidden nodes 921 to 92 j) exists between the input layer 91 (input nodes 911 to 91 i) and the output layer 93 (output nodes 931 to 93 k). In order to model the complexity of the input-output relationship, the number of hidden nodes 921 to 92j can be changed by increasing or decreasing the number of hidden nodes added to the network function.

The output layer 93 is organized to have at least 1 to k processing elements called output nodes 931 to 93k. The processing elements or nodes are mutually joined so that the relationship between the co-ordinated composition data and the colour data can be calculated at network operation.

In the method for producing a paint according to claim 1 of the present invention and the computer color matching system according to claim 4 of the present invention, each unit of the output layer 93 of the neural network 9 used in the artificial intelligence model for estimating the matching composition data Y from the color data X corresponds to 1 or more kinds of feature amounts related to the matching composition data Y.

In the method for producing a paint according to claim 2 of the present invention, the method for predicting color data of a paint film according to claim 3 of the present invention, the computer toning system according to claim 4 of the present invention, and the system for predicting color data of a paint film according to claim 6 of the present invention, each cell of the output layer 93 of the neural network 9 used in the artificial intelligence model for estimating color data X from the matching composition data Y corresponds to 1 or more feature amounts related to color data X.

The data in the neural network 9 flows in only 1 direction, and each node receives feedback without transmitting a signal to only 1 or more nodes.

In the method for producing a paint according to claim 1 of the present invention and the computer toning system according to claim 4 of the present invention, in the artificial intelligence model for estimating the matching composition data Y from the color data X, the input nodes 911 to 91i in the input layer 91 correspond to 1 input node for 1 input variable (input element; parameter) in each color data. The output nodes 931 to 93k in the output layer 93 correspond to 1 output node for 1 output variable (output element; parameter) in each of the matched constituent data.

In the method for producing a paint according to claim 2 of the present invention, the method for predicting color data of a paint film according to claim 3 of the present invention, the computer color matching system according to claim 4 of the present invention, and the system for predicting color data of a paint film according to claim 6 of the present invention, the input nodes 911 to 91i in the input layer 91 correspond to 1 input variable (input element; parameter) in each of the blended composition data in the artificial intelligence model for estimating color data X from the blended composition data Y. The output nodes 931 to 93k in the output layer 93 correspond to 1 output node for 1 output variable (output element; parameter) in each color data.

In the artificial intelligence model, the number of hidden nodes 921 to 92j in the hidden layer 92 can be increased or decreased according to the complexity of the input-output relationship. The junctions between the input nodes, between the input nodes and the hidden nodes, between the hidden nodes and the output nodes, and between the output nodes of the output layer 93 of the input layer 91 have junction weights related thereto, and may have one or more additional threshold weights for the hidden nodes 921 to 92j and the output nodes 931 to 93k, respectively.

Artificial intelligence using the neural network 9 is particularly advantageous in complex systems or phenomena where the analysis is complex and where it is cumbersome to derive models from human knowledge for use in computer-specific systems.

Also, for example, in the apparatus shown in fig. 2, the neural network 9 may be generated on the server computer side constituting the database. Thus, the connected operator (user) can receive high-quality data at any time without being influenced by geographical importance conditions or the like. In addition, by improving the security of the server computer, data changes due to incorrect connection, damage to the neural network, and the like can be prevented.

In the present invention, at least 1 artificial intelligence model for estimating color data X from matching component data Y may be provided in addition to at least 1 artificial intelligence model for estimating matching component data Y from color data X. In the artificial intelligence model, the neural network is configured in the same manner as the neural network shown in fig. 3, each cell of the input layer 91 corresponds to 1 or more kinds of feature amounts related to the matching composition data Y, and each cell of the output layer 93 corresponds to 1 or more kinds of feature amounts related to the color data X.

Database for storing data

The present invention registers (records) color data in association with matching composition data, thereby constructing a database. In the database of the present invention, various data concerning colors or compositions may be registered in association with each other on the basis of color data and matching composition data. In the present invention, the data registered in the database is preferably very much data (so-called large data). Specifically, 5 thousand or more groups, preferably 1 ten thousand or more groups, and more preferably 2 ten thousand or more groups. These data can be added, changed, and eliminated as desired.

The composition data Y of 1 or more kinds of compositions registered in the database and the corresponding color data X include actual measurement data or include actual measurement data and data calculated from the actual measurement data. Examples of the data calculated from the measured data include various parameter values calculated by a predetermined equation using the measured data, and predicted values calculated from the measured data.

The database may be a single database, or a plurality of databases that are related to each other at least by one common information element or a plurality of databases that are not related to each other.

The database may be provided in a server that is communicable with the computer and is operable remotely. At least a part of the database may be provided in a recording unit (memory or the like) in a computer, a colorimeter, a microscopic brightness measuring device, an automatic teller machine, and a device that obtains or uses data registered in another database.

When there are a plurality of databases, each database may also be connected by a wired or wireless connection. The database may be connected by wire or wirelessly to one or more of a computer, a colorimeter, a microscopic brightness measuring device, an automatic teller machine, and a device that obtains or uses data registered in another database.

Composition and method for producing the same

The composition of the present invention may contain 1 or more coloring materials in an arbitrary amount ratio. The coloring material contained in the composition is, for example, a coloring pigment, a dye, a luminescent pigment (a light-reflective pigment, a light-interference pigment, or the like) or the like, which has a function of coloring the composition.

The composition of the present invention may also be used, for example, as a "primary color paint" to be used in repair. For example, a paint in which raw materials containing 1 or more primary color paint are mixed and colored to a desired color may be used.

The composition of the present invention may further contain various additives and the like used in the fields of coloring pigment pastes, luster pigment pastes, orientation control agents, luster control agents, other coatings and the like.

Color data

In the present invention, the color data registered in the database includes data concerning the appearance characteristics such as texture, brightness, and luster.

These data can be obtained by measuring a coating film obtained from the composition using a colorimeter, a camera, or the like. Further, it may be obtained by using 1 or more pieces of image data of the coating film obtained by an instrument, or by analyzing, converting, correcting, or the like the image data as necessary. At least a part of the various color data obtained by measurement obtained by processing the image data may be calculated by performing arithmetic processing. The data obtained by measurement using the measuring instrument may be data obtained by correcting, if necessary, errors or the like caused by the measuring instrument or measurement fluctuation or the like.

The K value (light absorption coefficient) and S value (light scattering coefficient) of the composition itself may be used as color data. The K value and S value can be obtained by, for example, numerically processing color measurement data of the colors of the composition and the composition to be thinned.

The data related to the color and/or the appearance characteristic is, for example, data obtained directly by measurement using a measurement instrument such as a colorimeter, a multi-angle spectrophotometer, a laser metal-sensor, a variable angle spectrophotometer, a gloss meter, a camera, or a microscopic brightness sensor, or data calculated from the data obtained by the measurement.

In addition, the data related to the color and/or the data related to the appearance characteristics may include data of 1 or more illumination angles, 1 or more observation angles, or images associated with these combinations, and the like.

The measuring instrument for obtaining color data is not particularly limited as long as it can measure the color of a bright coating film (a metal coating film, a pearl coating film, or the like), a pure color coating film, or the like to obtain color data, and a conventionally known measuring instrument may be used without limitation to a measuring principle, a method for calculating color data of a measured value, or the like. For example, 1 or more of a colorimeter such as a single-angle spectrophotometer, a multi-angle spectrophotometer, a colorimeter, a color difference meter, and a variable angle spectrophotometer, an imaging device, a microscopic brightness measuring instrument, and a measuring instrument such as a color sample card may be used. In addition, a data processing device that processes various color data obtained from these measuring instruments can be arbitrarily used.

The color data registered in the database may be, for example, data indicating brightness, saturation, and hue, or data allowing the color to be specified by calculation, as the data related to the color. For example, the data may be based on XYZ color system (X, Y, Z values), RGB color system, lxab color system (Lxab, a, b values), hunterLab color system (L, a, b values), lxac/h color system (L xac, h values) defined by CIE (1994), montel color system (H, V, C values) or the like. In the present invention, the data related to the color among the color data registered in the database may be data based on a color system of 1 or more of these. Preferably, the data based on the LXa Xb color system or the LXC Xh color system is widely used in various fields including the field of repair coating.

The color data registered in the database may be, for example, data concerning appearance characteristics, such as a macroscopic gloss which is a texture perceived under macroscopic observation, and a microscopic gloss, a depth (depth), and a vividness which are textures perceived under microscopic observation, which are textures perceived when observing a surface of a color to be measured such as a coating film containing a luminescent pigment.

Macroscopic shine

The macroscopic brightness may be obtained by irradiating a surface to be measured with uniform light, receiving the reflected light at each angle, and measuring the color of the reflected light. Further, when the surface to be measured is observed at a long distance, there may be exemplified an FF value (trigger value) indicating a trigger phenomenon in which a color (brightness, saturation, hue) changes due to an increase or decrease in illumination and observation angle, an IV value (intensity value) indicating a visual brightness on a strong light side in which an opening angle between the incident light and the regular reflection light is between 10 degrees and 25 degrees, an SV value (scale value) indicating a brightness on a front surface on the strong light side, a cFF value, a metallic feeling index, a depth feeling index, a vividness, a gloss value indicating gloss, and the like.

The macroscopic brightness can be obtained directly by, for example, a multi-angle spectrophotometer, a laser metal-sensing measuring instrument, a variable angle spectrophotometer, a gloss meter, or the like, and can be calculated from these. For example, BYK-Mac i (trade name, manufactured by BYK Co., ltd.), MA-68II (trade name, manufactured by X-Rite Co., ltd.), etc. can be used as the multi-angle spectrophotometer. The FF value, IV value and SV value can also be obtained by using a laser metal sensor ALCOPE (registered trademark) LMR-200 (trade name, manufactured by Guanyi paint Co., ltd.).

The multi-angle spectral reflectance is a spectral reflectance measured by a spectrophotometer capable of measuring colors at multiple angles and is represented by R (x, λ). Here, R is a spectral Reflectance (Reflectance), which is represented by% of the spectral Reflectance corrected by a correction plate attached to the measuring instrument. X is the light receiving angle, and is expressed by the deflection angle with the regular reflection light. Lambda is the wavelength, and the visible light range is 400 to 700nm measured at 10nm intervals (31 wavelengths). The incident angle is-45 degrees as a normal standard.

In the present invention, when the incident angle is 45 degrees, the light receiving angle x is 1 or more, preferably 3 or more, of any angle from a strong light (25 degrees, 15 degrees, -15 degrees), a front (face) (45 degrees), to a light shielding surface (shield) (75 degrees, 110 degrees). As shown in fig. 4, the light receiving angles are preferably 6 angles of-15 degrees, 25 degrees, 45 degrees, 75 degrees, and 110 degrees. As shown in fig. 5, the light receiving angle may be 5 angles of 15 degrees, 25 degrees, 45 degrees, 75 degrees, and 110 degrees.

The FF value (trigger value) is a value indicating the degree of change in the L value (luminance) based on the observation angle (light receiving angle). The trigger corresponds to a difference in brightness between the direction of strong light (direction close to the direction of regular reflection of light) and the direction of the light shielding surface (direction far from the direction of regular reflection of light). The larger the FF value is, the larger the change in the L value (luminance) based on the observation angle (light receiving angle) is, indicating excellent triggering.

The FF value can be obtained by measuring the L value (luminance) at the light receiving angle of 15 degrees and the light receiving angle of 110 degrees by irradiating light from an angle of 45 degrees on the surface to be measured using a multi-angle spectrocolorimeter.

FF value = L value of 15 degrees light acceptance angle/L value of 110 degrees light acceptance angle

The IV value is a Y value on an XYZ color system obtained from the spectral reflectance measured in the direction of 15 degrees.

The SV value is a Y value on an XYZ color system obtained from the spectral reflectance measured in the direction of 45 degrees.

The FF value can be obtained by using Y15 as the Y value on the XYZ color system obtained from the spectral reflectance measured in the direction of 15 degrees of the offset angle and Y45 as the Y value on the XYZ color system obtained from the spectral reflectance measured in the direction of 45 degrees of the offset angle.

FF value = 2× (y15—y45)/(y15+y45)

The c value on the Lx c x h color system obtained from the spectral reflectance measured in the direction of 15 degrees of the deflection angle is taken as c x 15, the cFF value is obtained by using c×45 on the color system of lc×h, which is obtained from the spectral reflectance measured in the direction of 45 degrees.

cFF value = 2× (c 15-c 45)/(c 15+ c 45)

The metallic index can be obtained by using the Y value on the XYZ color system obtained from the spectral reflectance measured in the direction of 15 degrees off-angle as Y15 and the FF value as FF.

Metal feel index=y15×ff ²

The depth perception index indicates a depth perception imparted by a lustrous pigment, and the L and c values on the L and c color system obtained from the spectral reflectance at the characteristic angle are respectively obtained as L and c R using the following formulas.

Depth perception index = c R/L R

For example, a depth sensation of a light-shielding surface obtained by measuring c×75, which is a c×75 value of a light-receiving angle 75, and l×75, which is an l×75 value of the light-receiving angle 75, by irradiating light from an angle of 45 degrees using a multi-angle spectrocolorimeter, can be used.

Depth perception of light blocking surface=c×75/l×75

Using L R and c R, the sharpness can be found by the following formula.

Definition = sqrt [ (L x R) ² +(c*R) ² ]

Microscopic light sensation

The microscopic shine is a sensation of luminance unevenness concerning two dimensions such as a sparkle sensation or a particle sensation that appears due to a shiny pigment in a surface to be measured when the surface to be measured is observed at a close distance.

A microscopic shine can be obtained using a microscopic shine meter. Examples of the microscopic shine measuring instrument include a microscopic shine measuring instrument including a light irradiation device for irradiating a shiny coating surface with light, a CCD camera for capturing an image of the irradiated coating surface at an angle at which the irradiated light is incident, and an image analysis device connected to the CCD camera for analyzing the image.

When measuring the microscopic shine of a surface to be measured using a microscopic shine meter, first, a simulated (artificial) sunlight is irradiated onto the surface to be measured. The light irradiation angle to the surface to be measured is usually 5 to 60 degrees, preferably 10 to 20 degrees, relative to the vertical line of the surface to be measured, and particularly preferably about 15 degrees relative to the vertical line. Although the shape of the light irradiation region is not specifically limited, it is generallyCircular, although the illuminated area on the surface to be measured generally corresponds to 1 to 10,000 mm of the surface to be measured ² But not limited thereto. The illuminance of the irradiation light is generally preferably in the range of 100 to 2,000 lux.

The light was irradiated onto the surface to be measured, and the surface to be measured irradiated with the light was photographed by a CCD (Charge Coupled Device) camera at an angle at which the specular reflection light was not incident on the surface to be measured. Although the imaging angle may be an angle at which regular reflected light does not enter, the imaging angle is particularly suitable for the vertical direction with respect to the surface to be measured. In addition, the angle between the photographing direction of the CCD camera and the regular reflection light is preferably in the range of 10 to 60 degrees. The measurement range of the CCD camera on the surface to be measured irradiated with light is not particularly limited as long as it is a range to be uniformly irradiated with light, but generally includes a central portion of the irradiated portion, and the measurement area is preferably 1 to 10,000 mm ² Preferably 10 to 600mm ² Is suitable in the range of (c).

The image captured by the CCD camera is a 2-dimensional image, and is divided into a plurality of (usually 10,000 to 1,000,000) areas (pixels, dots), and the brightness in each area is measured. In the present invention, "brightness" is a digital level representing the gradation value of each region of a 2-dimensional image captured by a CCD camera, and corresponds to the digital amount of brightness of an object. The digital level indicates the brightness of each area outputted from the CCD camera of 8 binary digits resolution, and the digital level is indicated by a value of 0 to 255.

In the 2-dimensional image captured by the CCD camera, the brightness is high because the area corresponding to the portion where the reflected light of the bright pigment is strong is high, and the brightness is low in the area other than the portion. Further, even in a region corresponding to a portion of the luminescent pigment where the reflected light is strong, the brightness varies depending on the size, shape, angle, material, and the like of the luminescent pigment. That is, the luminance of each region can be represented, and in the present invention, the luminance distribution of a 2-dimensional image captured with a CCD camera can be represented three-dimensionally based on the luminance in the respective region. The three-dimensional distribution pattern of the brightness is divided into peaks, valleys and flat portions, the height or size of the peaks indicates the degree of the shininess due to the shiny pigment, the higher the peaks are, the more pronounced the shininess is, and the valleys and flat portions indicate no shininess, indicating that the light is mainly reflected by the colored pigment or the substrate.

The image captured by the CCD camera may be analyzed by an image analysis device connected to the CCD camera. As the image analysis software used in the image analysis device, for example, windorof (trade name: manufactured by Sangu business corporation) and the like are suitable.

In the analysis of an image, the "sparkling feeling" (irregular and fine luminescence feeling due to light that is regularly reflected from the luminescent pigment on the surface of the color to be measured) and the "particle feeling" (feeling due to the fact that the luminescent material contains irregularities and non-directivity (random pattern) due to the orientation and overlapping of the luminescent pigment on the surface of the color to be measured when the surface to be measured is observed as much as possible under the illumination condition where the sparkling feeling is difficult to be found) are quantitatively evaluated, respectively, and since the correlation with the observation evaluation is also high in the visual observation, the unevenness due to the personal difference is small at the time of measurement, and therefore, the method is preferable.

Examples of suitable methods for quantitatively measuring the sparkling sensation include the following measurement methods. A2-dimensional image formed by photographing a surface to be measured irradiated with light by a CCD camera is divided into a plurality of areas, the total of the respective brightnesses of the areas is obtained over the entire span of the areas, a total value is obtained, the total value is divided by the number of the entire areas to obtain an average brightness x, and a threshold value alpha is set to a value equal to or higher than the average brightness x. The preferred threshold α is typically the sum of the average luminance x and y (y is a number from 24 to 40, preferably from 28 to 36, more preferably 32).

Next, the value of the threshold α is subtracted from the brightness of each of the regions, and the subtracted values, which are positive values, are summed up to obtain a total volume V, which is the sum of the subtracted values. The total area S, which is the total number of the areas having the luminance equal to or higher than the threshold value α (the total number of the areas having the luminance equal to or higher than the threshold value α obtained by 2-valued with the threshold value α), is obtained. Since it is believed that the luminance peak can be approximated as a cone, pyramid, by dividing the total volume V by 3 times the total area S, i.e., by the formula

PHavα＝3V/S

The average height of the luminance peak, PHAv alpha, is obtained.

A threshold value β is set which is equal to or higher than the average luminance x and equal to or lower than the threshold value α. The threshold β is preferably equal to or less than the threshold α, and is generally the sum of the average luminance x and z (z is a number of 16 to 32, preferably 20 to 28, more preferably 24).

Next, the value of the threshold β is subtracted from the brightness of each of the regions, and the subtracted values, which are positive values, are summed up to obtain a total volume W, which is the sum of the subtracted values. The total area a, which is the total number of regions having a luminance equal to or greater than the threshold value β (the total number of regions equal to or greater than the threshold value β obtained by 2-valued at the threshold value β), is obtained. Since it is believed that the luminance peak can be approximated as a cone, pyramid, the value obtained by dividing the total volume W by the total area A is 3 times, i.e., by the formula

PHavβ＝3W/A

The average height tav beta of the luminance peaks of the threshold beta can be obtained.

The average particle area of the optical particles can be obtained from the total area a of the threshold value β and the number C of the optical particles indicating the luminance equal to or higher than the threshold value β. In the present invention, "optical particles" means "independent continuum with brightness equal to or higher than a threshold value on a 2-dimensional image". Assuming that the optical particles are round in shape, the diameter D of a circle having the same area as the average particle area is obtained by the following formula.

Number 1

By the following method

PSav＝D/PHavβ

The average width ratio PSav of the luminance peaks was obtained from the PHavβ and D.

From the average height PHAv of the luminance peak value obtained as described above and the average widening rate PSav of the luminance peak value obtained as described above, the average widening rate PSav of the luminance peak value obtained as described above is calculated by the following formula

BV＝PHavα+a·PSav

(wherein a is 300 when PHav alpha is less than 25, 1050 when PHav alpha is greater than 45, and 25 to 45, respectively

a＝300+37.5×(PHavα-25)

The value of the representation

The luminous value BV can be approximated.

In the appropriate method of the present invention, the "sparkle feeling" of the bright coating film can be quantitatively measured by the light emission value BV obtained as described above, and even when the concentration difference and the luminance difference of the bright material in the coating film are large, the correlation between the light emission value BV and the visual observation result of the "sparkle feeling" is relatively high.

The feeling of particle is represented by MGR value. The MGR value is one of the dimensions of the texture at the time of microscopic observation, that is, the microscopic shine, and is a parameter indicating the particle feel at strong light (the multilayer coating film is observed from the vicinity of regular reflection with respect to incident light). The MGR value is a measured value obtained by imaging a coating film of a multilayer coating film with a CCD camera at an incidence angle of 15 degrees/light receiving angle of 0 degrees, performing 2-dimensional fourier transform processing on the obtained digital image data, that is, the 2-dimensional luminance distribution data, extracting only a spatial frequency region corresponding to the particle sensation from the obtained power spectrum image, further extracting values of 0 to 100 from the calculated measurement parameters, and converting the values so as to maintain a linear relationship with the particle sensation. The ratio of 0 when no feeling of particles was found, and the ratio of about 100 when the feeling of particles was found to be the most.

As a suitable method for quantitatively measuring the "particle feel", there is a method in which a bright coating film surface irradiated with light is photographed with a CCD camera to obtain a 2-dimensional image, the energy of a low spatial frequency component is normalized by integration and a direct current component from a spatial frequency spectrum formed by 2-dimensional fourier transformation of the 2-dimensional image to obtain a 2-dimensional power spectrum integral value, and the particle feel of the coating film is quantitatively evaluated from the 2-dimensional power spectrum integral value, as described above.

When the low spatial frequency component is extracted from the image of the spatial frequency spectrum after 2-dimensional fourier transform and the 2-dimensional power spectrum integral value obtained by normalizing the integral and the dc component is measured, the extracted region of the low spatial frequency component extracted from the image of the spatial frequency spectrum is preferably a region having a linear density with a lower limit value of 0 root/mm to an upper limit value of 2 to 13.4 root/mm, and preferably a region of 0 root/mm to 4.4 root/mm, from the viewpoint of a high correlation with the visual observation result of "particle feel". The larger the 2-dimensional power spectrum integration value, the larger the particle feel.

The 2-dimensional power spectrum integration value (hereinafter, may be simply referred to as "IPSL") can be obtained by the following equation.

Number 2

(where v is a spatial frequency, θ is an angle, P is a power spectrum, 0 to L are extracted low spatial frequency regions, and L represents an upper limit of the extracted frequency)

The emission value BV is based on the following one-time expression

MBV＝(BV-50)/2

The calculated MBV value may also be evaluated for "sparkle feel".

MBV is a value with no sparkle as 0 and the most sparkle as about 100, with a more "sparkle" indicating a greater value. MBV values are also sometimes referred to as HB values (Hi-light Brilliant values).

The "particle sensation" may be evaluated by using the MGR value calculated by the following one-time equation based on the 2-dimensional power spectrum integration value (IPSL).

When the PSL value is 0.32 or more,

making MGR= [ (IPSL x 1000) -285]/2,

when the IPSL value is in the range of 0.15 < IPSL < 0.32,

make MGR= [ IPSL× (35/0.17) - (525/17) ]/2

When the IPSL value is 0.15 or less, mgr=0 is made.

The MGR value is a value of 0 for the particle feeling of no bright material and about 100 for the particle feeling of the most bright material, and the more "particle feeling" is, the larger the value is. The MGR value is also sometimes referred to as HG value (Hi-light grade value).

Further, the microscopic brightness calculated by the following formula, which is a combination of the MBV and MGR values, can be evaluated by an exponential value (microscopic brightness index).

Microscopic Brightness index= (MGR+α. MBV)/(1+α)

By examining a surface of a measured color having a large amount of a bright feel, if the α value is preferably 1.80 to 1.40, more preferably 1.63, a result that is better in agreement with the microscopic bright feel of the eye can be obtained. The microscopic shine index is a value of 0 for no shine (neither shininess nor graininess) and about 100 for the most shiny (most shiny and graininess)

Matched composition data

The database is registered with the compounding data of the composition.

The blending composition data includes data concerning each blending component and the blending amount of each of 1 or more coloring materials, binders, additives, and the like contained in the composition. In addition, when a commercially available product is used as the composition, the trade name (product number) itself may be used as the blending composition data, and the blending amount composition of each product may be used as the blending composition data. For example, the present invention is effective in the case of product number management and the like in combination with commercially available products or compositions whose composition data is unknown.

In the present invention, the shape, chemical properties, and the like of each component such as 1 or more coloring materials, binders, additives, and the like contained in the composition may be registered as blending composition data. Examples of the shape include a shape (spherical, scaly, fibrous, etc.) of a coloring material, an average primary particle diameter, an average secondary particle diameter, an average dispersed particle diameter, a particle diameter distribution, an aspect ratio, and a thickness. Examples of the chemical properties include molecular weight, molecular weight distribution, discoloration temperature, and reactivity.

In the present invention, when the composition contains a lustrous pigment such as a light-reflective pigment or a light-interference pigment, the content of the orientation controlling agent for controlling the content of the light-reflective pigment, the content of the light-interference pigment, and the orientation of the lustrous pigment is registered in the database as blending composition data.

The content of each color phase of the colorant, the content of each color phase of the light-reflective pigment, and the content of each color phase of the light-interference pigment contained in the composition are registered in the database as blending composition data.

The hues are defined in the LXCXh color system of the invention based on the LXa-b color system, which was regulated by the International Commission on illumination and was also used in JIS (Japanese Industrial Standard) Z8729 in 1976.

For example, in an LXCXh color chromaticity diagram calculated from the spectral reflectance at 45 degrees with respect to the specular reflection light of the light irradiating the coating film at 45 degrees, the color of the red color is defined as a color having a hue angle h in the range of-45 degrees or more and less than 45 degrees when the alpha red direction is 0 degrees. Similarly, the orange color is defined as a color having a hue angle h in a range of 45 degrees or more and less than 67.5 degrees when the red direction is 0 degrees, the yellow color is defined as a color having a hue angle h in a range of 67.5 degrees or more and less than 135 degrees when the red direction is 0 degrees, the green color is defined as a color having a hue angle h in a range of 135 degrees or more and less than-135 degrees when the red direction is 0 degrees, and the cyan color is defined as a color having a hue angle h in a range of-135 degrees and less than-45 degrees when the red direction is 0 degrees.

Coating condition data

The database may also be registered with coating condition data K.

The coating condition data K is all data concerning the coating, and examples thereof include coating tool information (type of coating tool, manufacturer, model, etc. of the coating tool), coating condition information (coating temperature, humidity at the time of coating, dry film thickness, coating solid content, coating distance, coating speed, etc.), coater information (name, coating skill, coating tendency, habit, etc.), drying condition information (drying temperature, drying humidity, drying apparatus manufacturer, drying apparatus model, etc.), and the like, which are used in the coating.

Computer with a memory for storing data

The computer included in the device used in the present invention means a computer such as an supercomputer (supercomputer), a desktop computer, a notebook computer, and a portable computer, and an electronic device having a calculation function and an information processing function such as a tablet terminal and a smart phone. The computer may be provided at any place including a work site, or may be carried by an operator or the like.

The computer includes a computing unit and a control unit, and may further include an input/output unit, a communication unit, a recording unit, and the like. In the present invention, a colorimeter or the like having recording, calculation, control, input/output functions and the like is mounted, and the present invention can be realized by being integrated with the colorimeter or the like. In addition, color data and matching composition data can be recorded in a recording part of the computer, and a database can be arranged in the recording part.

When the database is provided outside the computer, the computer is connected to the database by wire or wirelessly.

The computer may be connected to various devices for measuring data related to colors and/or data related to appearance characteristics, an automatic teller machine, other arithmetic devices, input devices such as a keyboard, a mouse, a code reader, a touch panel, an image recognition device, and output devices such as a monitor screen and a printing device, by wire or wirelessly.

The computer may be provided with (install) application software (program) for executing the method of the present invention or for controlling and operating the system of the present invention, as necessary, to perform necessary control and operation.

Automatic blending machine

The apparatus may further include an automatic blending machine for automatically blending each of the blending components based on the blending composition data to blend the colors. The autoregulator may be connected to the computer or database by wire or wirelessly.

The automatic blending machine at least comprises: an electronic balance for automatically weighing the weight or the capacity of the ingredients such as each coloring material; and an injector for injecting the weighed ingredients into the blender.

By using the automatic blending machine, high-precision weighing can be automatically performed, human errors during blending can be reduced, blending can be rapidly performed, and any amount of paint after color blending can be easily prepared. In addition, by recording the reconciliation process, production management can be easily performed. The automatic blending machine may be capable of automating all the operations related to blending, or may be capable of performing some of fine tuning by an operator.

Method for producing paint by computer toning (embodiment 1 of the present invention)

Fig. 6 is a flowchart when a method for producing a computer-controlled paint according to embodiment 1 of the present invention is performed. The flow shown in fig. 6 is just one embodiment of the present invention.

The method for producing a paint by using a computer-controlled paint according to embodiment 1 of the present invention is a method comprising the steps S101 to S111 using a device having a database in which 1 or more kinds of composition data Y1 to Yn (n is an integer of 2 or more) and color data X1 to Xn corresponding to each kind of composition data are registered, and a computer in which color matching calculation logic of data registered in the database is used.

The steps S101 to S111 will be described in detail below.

S101 procedure

The step S101 is a step of inputting learning data to the computer by using the data registered in the database.

In the present invention, it is preferable to separately produce: learning data of composition data and color data of a composition containing 1 or more compositions and not containing a bright pigment; and learning data using composition data and color data of a composition containing 1 or more coloring materials and 1 or more luminescent pigments, and these are inputted separately.

The present inventors made learning data based on the presence or absence of a luminescent pigment, and found that the suitability in computer toning was remarkably improved by inputting the data.

In the present invention, when the composition data of a composition containing 1 or more coloring materials and 1 or more luminescent pigments is used as learning data, it is preferable to use 1 or more data selected from the group consisting of the content of light-reflective pigments, the content of light-interference pigments, the content of orientation controlling agents, and the total of one or more of these as learning data.

In the present invention, when the composition data of a composition containing 1 or more coloring materials and 1 or more luminescent pigments is used as learning data, it is preferable to use content data of each hue of the luminescent pigment as learning data. Specifically, it is preferable to use 1 or more kinds of data selected from the content of each color phase of the light-reflective pigment, the content of each color phase of the light-interference pigment, and the content of each color phase of the colorant in the composition as learning data.

The color of the lustrous pigment is defined in the LXC-Xh color system of the invention based on the LXA-B color system, which was prescribed by the International Commission of illumination and which was also used in JIS (Japanese Industrial Standard) Z8729 in 1976.

For example, in an LXCXh color chromaticity diagram calculated from the spectral reflectance at 45 degrees with respect to the specular reflection light of the light irradiating the coating film at 45 degrees, the color of the red color is defined as a color having a hue angle h in the range of-45 degrees or more and less than 45 degrees when the alpha red direction is 0 degrees. Similarly, the orange color is defined as a color having a hue angle h in a range of 45 degrees or more and less than 67.5 degrees when the red direction is 0 degrees, the yellow color is defined as a color having a hue angle h in a range of 67.5 degrees or more and less than 135 degrees when the red direction is 0 degrees, the green color is defined as a color having a hue angle h in a range of 135 degrees or more and less than-135 degrees when the red direction is 0 degrees, and the cyan color is defined as a color having a hue angle h in a range of-135 degrees and less than-45 degrees when the red direction is 0 degrees.

In the present invention, when the blending composition data of 1 or more kinds of compositions is used as learning data, it is preferable to use the shape data of the coloring material contained in the composition as learning data. Specifically, shape data such as the shape (spherical, scaly, fibrous, etc.) of a coloring material such as a coloring pigment, an average primary particle diameter, an average secondary particle diameter, an average dispersed particle diameter, a particle size distribution, an aspect ratio, and a thickness of the coloring material are preferably used as learning data.

The data may be transmitted by a communication means such as a wired or wireless communication means or a means using a recording medium. Examples of the input using the communication means include a combination of 1 or more of various communication networks such as LAN (local area network), WAN (wide area network), internet, and telephone network. As an input by means of the recording medium, data of the recording medium such as a magnetic recording medium, an optical recording medium, and a paper recording medium can be read by using an appropriate reading means.

S102 procedure

S102 is a process of performing machine learning using the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model in which the matching composition data Y is estimated from the color data X. The artificial intelligence model in the present invention may be composed of 1 or more models selected from decision trees using gradient boosting, linear regression, logistic regression, simple perceptrons, MLPs, neural networks, support vector machines, random forests, gaussian processes, bayesian networks, k-nearest neighbor methods, and other machine learning. In the present invention, it is preferable to use 1 or more artificial intelligence models selected from decision trees using neural networks and gradient boosting, and gaussian processes, and it is particularly preferable to use 1 or more artificial intelligence models selected from decision trees using neural networks and gradient boosting.

In the present invention, by constructing a neural network and learning the neural network using the learning data input in step S101, it is possible to generate a learned artificial intelligence model including at least 1 artificial intelligence model in which the cooperative constituent data Y is estimated from the color data X.

In the present invention, the learned artificial intelligence model generated in the step S102 may further include at least 1 artificial intelligence model for estimating the color data X from the matched composition data Y, in addition to at least 1 artificial intelligence model for estimating the matched composition data Y from the color data X. Here, by performing machine learning using the learning data input in step S101, at least 1 artificial intelligence model for estimating the color data X from the matching composition data Y can be generated. Even in such a case, the neural network can be constructed so as to learn it.

Learning of the artificial intelligence model (neural network) is realized using the learning data input to the computer in step S101. At least the color data X and the composition matching data Y related to more than 1 composition are used as the learning data. As the algorithm of the neural network, a well-known error back propagation algorithm, which is one of teacher learning methods, can be used. The neural network is learned by setting a learning rate (real value between 0 and 1) which is a parameter indicating the learning speed, and an allowable error (real value between 0 and 1) which is an allowable error of the output value during learning. Thus, 1 or more feature values related to the blending composition data Y can be correlated with 1 or more feature values related to the color data X of the coating film of the coating material based on the blending composition data Y. The composition matching data satisfying the color data and the color data based on the composition matching data can be predicted by feedforward calculation using the learned network. The learned network predicts these networks without performing experimental confirmation concerning man-hours such as cost and time.

In the present invention, the step of generating the learned artificial intelligence model preferably includes: (i) A step of learning an artificial intelligence model by using, as learning data, 1 or more kinds of composition data Y and color data X relating to a composition containing no luminescent pigment; and (ii) a step of learning the artificial intelligence model by using, as learning data, 1 or more kinds of the composition data Y and the color data X of the composition containing the lustrous pigment. In this case, the step (ii) preferably includes a step of learning the artificial intelligence model by using, as learning data, 1 or more data selected from the group consisting of the content of the light-reflective pigment, the content of the light-interference pigment, the content of the orientation controlling agent, and the total of one or more of them in the composition.

In the process of learning the artificial intelligence model, the order of the process (i) and the process (ii) is not particularly limited, and the process (ii) may be performed after the process (i) or the process (i) may be performed after the process (ii). In the present invention, learning data is created separately from the viewpoint of whether or not the composition contains a luminescent pigment, and learning the data enables the generation of an artificial intelligence model that can be predicted with higher accuracy.

In the present invention, the step of generating the learned artificial intelligent model preferably includes a step of learning the artificial intelligent model by using, as learning data, 1 or more data selected from the group consisting of the content of the light-reflective pigment, the content of the light-interference pigment, the content of the orientation controlling agent such as amorphous silica, and the total of 1 or more of these. This makes it possible to obtain matching composition data which can be applied to color data of a surface to be measured containing a luminescent pigment with particularly high accuracy.

In the present invention, the step of generating the learned artificial intelligence model preferably includes a step of learning the artificial intelligence model by using, as learning data, 1 or more kinds of data selected from the content of each color phase of the light-reflective pigment, the content of each color phase of the light-interference pigment, and the content of each color phase of the colorant in the composition. This makes it possible to obtain matching composition data suitable for color data of a surface to be measured containing a luminescent pigment with particularly higher accuracy.

In the present invention, the step of generating the learned artificial intelligent model preferably includes a step of learning the artificial intelligent model by using shape data of the coloring material contained in the composition as learning data. This makes it possible to adapt the feeling caused by the particles in the color data of the surface to be measured with higher accuracy.

In the present invention, the coloring material includes not only a usual coloring agent such as an inorganic coloring pigment or an organic coloring pigment, but also a luster pigment such as a particulate or flake-like (scaly) glass, a metal, silica, alumina, or a flake-like glass having interference properties (for example, silica-coated glass flakes, etc.), a light interference pigment such as silica, alumina, or the like. The shape data includes data such as spherical, flake-like, fibrous shape appearance, and the like, and particle diameter, particle size distribution, thickness, aspect ratio, fiber length, fiber diameter, and the like.

S103 step

The step S103 is a step of acquiring color data Xp (hereinafter, sometimes referred to as "target color data Xp") of a target color whose matching composition Yp is unknown.

The target color data Xp may be color data of all colors of a coated article, a molded article, a natural structure, or the like. In particular, the color data is preferable as color data of a painted object.

The present invention enables highly accurate toning even when color data of a coating film containing a bright pigment, which has been difficult to be computer-toned up to date, is used as target color data Xp. Therefore, the target color data Xp in step S103 is preferably color data of a coating film containing a luminescent pigment. Of course, the target color data Xp in step S103 may be color data of a coating film that does not contain a bright pigment.

The elements constituting the target color data Xp may be the same as the elements constituting the color data registered in the database. For example, the color data measured by a measuring instrument or the color data calculated therefrom may be used.

The measuring instrument for obtaining color data is not particularly limited as long as it can measure the color of a bright coating film (a metal coating film, a pearl coating film, or the like), a pure color coating film, or the like to obtain color data, and a conventionally known measuring instrument may be used without limitation to a measuring principle, a method for calculating color data of a measured value, or the like. For example, 1 or more of a single-angle spectrophotometer, a multi-angle spectrophotometer, a colorimeter, a color difference meter, a color change spectrophotometer, and other colorimeter and imaging device, a measuring instrument such as a microscopic brightness measuring instrument, and a measuring instrument such as a color sample card can be used. In addition, a data processing device that processes various color data obtained from these measuring instruments can be arbitrarily used.

The operator can obtain the target color data Xp by directly measuring the object to be measured using various measuring instruments. Further, the measurement data may be automatically acquired by various measuring instruments according to a program or the like. Further, the calculation may be performed based on these colorimetric data.

In the present invention, the target color data Xp is preferably obtained by measuring the surface of the object color using a multi-angle spectrophotometer.

In addition, when the target color data Xp is not data obtained by directly measuring the object to be measured, color data obtained from the trade name or the like of the object to be measured can be used as the target color data Xp. For example, when the target color data Xp is color data on an automobile, the target color data Xp may be set according to paint data obtained by the trade name, model, year, manufacturing number, and the like of the automobile.

S104 procedure

The step S104 is a step of inputting the target color data Xp to the computer.

The input to the computer may be transmitted and received by the computer from various devices for measuring and/or calculating the color data Xp by means of communication means by wire, wireless or a combination of these or by means of a recording medium. Examples of the input using the communication means include a combination of 1 or more of various communication networks such as LAN (local area network), WAN (wide area network), internet, and telephone network. As an input by means of the recording medium, data of the recording medium such as a magnetic recording medium, an optical recording medium, and a paper recording medium can be read by using an appropriate reading means.

In addition, a keyboard, a mouse, a code reader, a touch panel, a voice input device, an image recognition device, or the like may be used to connect to or input by an input means provided in the computer.

S105 procedure

The step S105 is a step of obtaining predicted composition data Ya1 predicted from the color data Xp as composition data containing 1 or more of the compositions C1 to Cn as components by using the learned artificial intelligence model and/or a prediction formula other than the artificial intelligence model.

The learned artificial intelligence model used in the step S105 is the learned artificial intelligence model generated in the step S102, and is at least 1 artificial intelligence model that is inferred from the color data X in cooperation with the composition data Y.

As a method of obtaining predicted cooperative composition data Ya1 predicted from color data Xp as cooperative composition data of 1 or more components of the compositions C1 to Cn using the learned artificial intelligence model, a feature amount of the color data Xp may be inputted to each unit of an input layer in the neural network of the learned artificial intelligence model. The color data Xp input to the input layer is transmitted while being weighted between each node and each layer, and is output as prediction matching component data Ya1 from each cell of the output layer.

The step S105 may include a step of obtaining predicted blend composition data Ya1 predicted from the color data Xp as blend composition data including 1 or more of the above-described compositions C1 to Cn by multi-label classification. Such predicted blend composition data Ya1 may be obtained as the composition of 2 or more types of the compositions C1 to Cn (for example, the amount of C1, the amount of C2, etc., as the composition based on the composition unit), or may be obtained as the respective amounts of the components (pigments, etc.) constituting the compositions C1 to Cn (for example, the amount of the red pigment a, the amount of the red pigment B, etc., as the composition based on the respective components).

Here, the multi-label classification is set so that 2 or more solutions (labels) exist simultaneously for a specific object, or that 2 or more solutions (categories) can be assigned simultaneously.

The number of the matched components to be specific color data is not always 1, but a plurality of the matched components are present. Therefore, by employing the multi-label classification, the prediction matching composition data Ya1 predicted from the color data Xp can be obtained more efficiently. For example, when the blended composition data satisfying the green color data Xg is obtained by the multi-label classification, the blended composition data can be used as a solution for the compositions of the green compositions Cg1 and Cg2, and the blended composition data can be used as a solution for the compositions of the yellow composition Cy1 and the cyan composition Cb 1.

In the present invention, by employing the multi-label classification, the presence amount of each component provided in the predicted blend composition data can be represented with probability, and thus each component composition (both composition per composition and composition per constituent component) in the predicted blend composition data can be predicted with high probability.

Examples of a method using a predictive formula other than the artificial intelligence model used in the step S105 include a method known as computer Color Correction (CCM), that is, calculation based on color matching calculation logic using a computer or calculation based on mathematical optimization.

The step S105 may be a step corresponding to a Computer Color Selection (CCS). For example, among the plurality of color data registered in the database, color data similar to the target color data Xp can be searched for, and after being acquired as search color data Xn1, the matching component data corresponding to the search color data Xn1 can be used as prediction matching component data.

Here, the color data registered in the database is, for example, color data of a known color sample account, color data of a paint sheet produced in the past, or the like, and is associated with matching composition data corresponding to the color data. Thus, by acquiring the search color data Xn1, the corresponding matched composition data can be easily obtained as predicted matched composition data.

Comparing 1 or more of the elements constituting the color data (for example, each value in the Lxab color system, etc.) with the corresponding elements constituting the target color data Xp, the search color data Xn1 can be obtained by elements such as the difference of the search values, the degree of coincidence, the error rate, and the like, which fall within a certain range. The predetermined range may be set by an operator with reference to experience or the like, or may be set by a computer.

In step S105, when the target color data Xp and the search color data Xn1 are compared and whether they are acceptable or not is determined, the comparison may be performed focusing on 1 or more elements constituting the search color data Xn1 and 1 or more elements constituting the target color data Xp, and the comparison may be performed for each corresponding component element. When the pass or fail is determined, for example, a threshold value may be set for the difference, the degree of coincidence, the error rate, or the like in each component, and the pass or fail may be determined by referring to the device or the operator. In this case, the constituent elements may be weighted by reflecting the viewpoint of the skilled worker or the like.

The predicted composition data Ya1 obtained in the step S105 can be used as data of 1 or more of the compositions C1 to Cn. The data may also be used as data of the blending amount ratio of each component such as resin, coloring material, solvent, etc. The data of 1 or more of the above-mentioned compositions C1 to Cn are preferably obtained in consideration of workability and the like at the time of actual blending.

In the predicted blending composition data Ya1 obtained in step S105, the types of the compositions C1 to Cn are not particularly limited, but 15 types or less, preferably 12 types or less, and more preferably 10 types or less can be obtained in consideration of workability at the time of actual blending. In this case, the number of the metal pigment-containing compositions is 5 or less, preferably 3 or less, and the number of the pearlescent pigment-containing compositions is 5 or less, preferably 3 or less.

In the step S105 of the present invention, it is preferable that at least 1 kind of artificial intelligence model of the composition data Y is estimated from the color data X using the learned artificial intelligence model, and the predicted composition data Ya1 predicted from the color data Xp is obtained as the composition data containing 1 or more kinds of the compositions C1 to Cn as components.

S106 procedure

The step S106 is a step of obtaining predicted color data Xa1 predicted from the predicted cooperative component data Ya1 using the learned artificial intelligence model and/or a prediction expression other than the artificial intelligence model, and comparing the obtained predicted color data Xa1 with the color data Xp to determine whether or not the obtained product is acceptable.

As the learned artificial intelligence model used in the method of obtaining the predicted color data Xa1 predicted from the predicted cooperative constituent data Ya1, at least 1 kind of artificial intelligence model that is a learned artificial intelligence model generated in the step S102 and that infers the color data X from the cooperative constituent data Y can be exemplified.

As a method for obtaining the predicted color data Xa1 of the learned artificial intelligence model, the feature values of the prediction matching component data Ya1 may be input to each unit of the input layer in the neural network of the learned artificial intelligence model. The prediction matching component data Ya1 inputted to the input layer is transmitted while being weighted between each node and each layer, and outputted as predicted color data Xa1 from each cell of the output layer.

As a method using a predictive formula other than the artificial intelligence model, which is used in a method of obtaining the predicted color data Xa1 predicted from the predicted complex composition data Ya1, for example, a method known as computer Color Correction (CCM), that is, calculation based on color matching calculation logic using a computer or calculation based on mathematical optimization, is exemplified.

Based on the calculation of the color matching calculation logic using a computer, for example, based on the various color data registered in the database and the composition matching data corresponding thereto, the target color data Xp and the various color data are compared and calculated so that the difference, the degree of coincidence, and the like are within a certain range, whereby one or more matching compositions considered most reasonable are determined as predicted matching composition data Ya1. By using various functions constituting the calculation logic, an arbitrary fitting composition or an approximate fitting composition can be corrected by a small number of repeated steps. In this case, theoretical instructions are generated in a regular pattern, and the calculation speed and the accuracy of the adjustment algorithm can be complemented.

The predicted fit composition data Ya1 is recorded in numbers, for example, with reference toThe color data in the database is calculated by optimizing the mathematical system by searching for components having characteristic information that acts in a direction of reducing the error for each coordinate axis constituting the color data obtained by approximately matching the components. For example, in the color system of LXa b, when the error is defined as DeltaLXa=LXa ₂ -L＊ ₁ 、Δa＊＝a＊ ₂ -a＊ ₁ 、Δb＊＝b＊ ₂ -b＊ ₁ When the error DeltaL on the L-axis is positive, a component having characteristic information acting in a direction of decreasing the L-2 value is searched, and when the error DeltaL on the L-axis is negative, a component having characteristic information of increasing the L-2 value is searched. Similarly, when the error Deltaa on the a-axis is positive, the search is performed to reduce the a-axis ₂ The component of the characteristic information (green) of the value is searched for a value having an increase in a when the error Deltaa in a' axis is negative ₂ The component of the characteristic information (red) of the value is searched for a value having a reduced b when the error Deltab on the b' axis is positive ₂ The component of the characteristic information (green) of the value is searched for a component having an increased b when the error Deltab on the b' axis is negative ₂ The characteristic information (yellow) of the value. In this way, in the approximate composition, the predicted composition data Ya1 of the color close to the target can be obtained by adding a component that acts in the direction of reducing the error, that is, for imparting predetermined characteristic information to each coordinate axis of the color system constituting the color space.

If no component having the characteristic information that acts in the error-reducing direction is retrieved, the candidate matching composition data can be acquired from the target color data Xp after obtaining a new approximate matching composition that is more suitable.

Further, the predicted fit composition data Ya1 may be obtained by performing operator correction (for example, correction of an operator by color data of a known color sample account, color data of a coating plate manufactured in the past, reference to own experience, or the like), correction of the computer, correction of an artificial intelligence model, or the like on the fit composition data obtained by CCM.

In addition, in a job site such as an automobile repair facility, when the available coloring material or composition is limited, the predicted blend composition data Ya1 may be obtained based on only the coloring material or composition available in the job site.

The predictive fit composition data Ya1 can be output by display means, printing means, or the like. The information may be transmitted from the computer to an apparatus or the like for performing the next step, instead of being output.

As a method for obtaining the predicted color data Xa1 using a prediction formula other than the learned artificial intelligence model, various prediction formulas known in the field of color matching using CCM can be used. Examples of such a predictive expression include a method using a predictive expression based on the Kubelka-Munk optical density expression and the Duncan color mixing theory expression, a method using fuzzy inference, and a method of indexing the degree of matching by comparing color data and matching composition data by a computer.

The methods using the Kubelka-Munk optical concentration formula and the Duncan color mixing theory formula are as follows. The respective light scattering coefficients and light absorption coefficients of the respective coloring materials contained in 1 or more of the compositions and the blending ratio of the respective coloring materials were obtained, and "the light absorption coefficient after color mixing/the light scattering coefficient after color mixing" was calculated from the Kubelka-Munk optical density, and the spectral reflectance was obtained from the Duncan color mixing theory using this value. The "light absorption coefficient/light scattering coefficient" can be calculated from the spectral reflectance of the target color, and the blending ratio of each primary color paint such as the coloring material or composition required for color alignment can be obtained. By performing this calculation for each wavelength of the visible spectrum, the pigment blend ratio for producing the target color can be determined.

Here, the optical density formula of Kubelka-Munk is as follows.

Number 3

(K/S) _λ : kubelka-Munk optical concentration function of wavelength λ

K: light absorption coefficient

S: light scattering coefficient

R _λ : reflectance at wavelength λ

Lambda: wavelength of

The Duncan color mixing theory is as follows.

Number 4

K _m : light absorption coefficient after color mixing

S _m : light scattering coefficient after color mixing

K _i : light absorption coefficient of colorant i

S _i : light scattering coefficient of colorant i

P _i : compounding ratio of colorant i

The Kubelka-Munk optical density is calculated from the spectral reflectance and the ratio of the light absorption coefficient to the light scattering coefficient is calculated, and in order to calculate color mixing using the Duncan color mixing theory, it is necessary to first calculate the light absorption coefficient and the light scattering coefficient, respectively. As a method for obtaining the light absorption coefficient and the light scattering coefficient, a known method can be used, and for example, a relative method or an absolute method can be used.

In this case, in order to further improve the prediction accuracy, in order to correct the influence of the internal specular reflection or the refractive index difference generated at the interface between the resin layer forming the paint and the air layer on the measurement of the spectral reflectance, the color mixture calculation may be performed after the reflectance converted into the ideal state by using the sandsen method. In order to adjust the mixing ratio of the colorant to the target color, a repetitive calculation by the newton's iteration method may be used, and in the evaluation of color consistency between the target reflectance and the predicted reflectance, color values XYZ and l×b calculated from the reflectance may be used, or a meta method in which convergence calculation is performed by the newton's iteration method while evaluating the difference between the target value and the predicted value, or an isomeric method in which convergence calculation is performed while evaluating the sum of squares of the difference between the target reflectance and the predicted reflectance may be used.

Fuzzy inference can be employed, with membership functions in fuzzy set theory defining the method of ambiguity. As a specific method of fuzzy inference, various methods have been proposed so far, and any method can be used in the present invention. For example, the fuzzy inference method developed by Mandarin (Mamdani) may be used.

When the predicted color data Xa1 is obtained in step S106, the artificial intelligence model, the predictive expression other than the artificial intelligence model, or both of them may be used in a switching manner.

In the present invention, the step S106 can be performed to obtain the predicted color data Xa1 using only at least 1 artificial intelligence model. In step S106 after the 2 nd time, the predicted color data Xa1 may be obtained without using the artificial intelligence model.

The predicted color data Xa1 that can be obtained in step S106 may be a variety of color data recorded in the database. In the present invention, the predicted color data Xa1 preferably includes multi-angle spectral reflectance and/or a brightness parameter. By including the predicted color data Xa1 with the spectral reflectance and/or the brightness parameter at multiple angles, even for a bright color whose optical characteristics are difficult to predict, it is possible to more accurately perform color matching.

As a criterion for determining whether or not the color data Xa1 is acceptable, for example, the predicted color data Xa1 is equivalent to the color data Xp. As a criterion for determining the same, for example, each element constituting the color data Xp is compared with each element constituting the predicted color data Xa1 individually, and whether or not the difference is within a predetermined range is determined. For example, when the color data Xp and the predicted color data Xa1 include elements using the Lxaa-b color system and elements derived therefrom, the color difference DeltaE can be compared on the basis of the L, a and b respectively to judge whether the product is qualified or not. In this case, a threshold value may be set for the difference, the degree of coincidence, the error rate, etc. among the respective constituent elements, or various correction values may be used, and the operator, the computer, or the instrument may refer to these values, and it is preferable that the computer or the instrument determine whether the difference, the degree of coincidence, the error rate, etc. is acceptable.

In the present invention, when the predicted color data Xa1 is color data related to brightness, it is preferable to determine whether or not the predicted color data is acceptable using spectral reflectance and/or brightness parameters at multiple angles.

When whether or not the predicted color data Xa1 is acceptable is determined, the operator may be notified of an improvement point for bringing the predicted color data Xa1 closer to the target color data Xp, if necessary.

S107 step

In the step S107, when the composition is not acceptable in the step S106, predicted composition data Yai, which is predicted from color data Xp and is different from the predicted composition data up to now, is obtained as composition data containing 1 or more of the compositions C1 to Cn as components, using the learned artificial intelligent model and/or a predictive formula other than the artificial intelligent model, and then, the predicted color data Xai predicted from the predicted composition data Yai is obtained using the predictive formula other than the learned artificial intelligent model, and the step of determining whether or not the composition is acceptable by comparing the composition data Xp is repeated until the composition is acceptable.

In step S107, predicted blend composition data Yai, which is predicted from color data Xp and is different from the predicted blend composition data up to now, is obtained as blend composition data containing 1 or more of the compositions C1 to Cn as components using the learned artificial intelligence model and/or a predictive formula other than the artificial intelligence model, and examples of such methods include the following methods (1) to (5).

(1) By using the multi-label classification, the matching composition data which is not selected in S105 among the plurality of matching composition data satisfying the color data Xp is set as new predicted matching composition data Yai, whereby the matching composition data including 1 or more of the compositions C1 to Cn as components is obtained.

(2) The number of usable types or the like is changed in the compositions C1 to Cn, and the parameters that are not used in the step S105 are introduced to predict from the color data Xp, and the new predicted blend composition data Yai is obtained as blend composition data of 1 or more components of the compositions C1 to Cn.

(3) The predicted blend composition data Yai, which is predicted from the color data Xp and is different from the predicted blend composition data up to now, is obtained as blend composition data containing 1 or more of the above-described compositions C1 to Cn, using the artificial intelligence model that is not used in step S105.

(4) The predicted blend composition data Yai, which is predicted from the color data Xp and is different from the predicted blend composition data up to now, is obtained as blend composition data containing 1 or more of the above-described compositions C1 to Cn, using a predictive formula other than the artificial intelligence model that is not used in the step S105.

(5) Taking into consideration the difference obtained by comparing Xp and Xa1 performed in step S106, the predicted blend composition data Yai is obtained as blend composition data containing 1 or more of the compositions C1 to Cn by using the learned artificial intelligence model and/or a predictive expression other than the artificial intelligence model.

The method of obtaining the predicted composition data Ya1 predicted from the color data Xp as the composition data containing 1 or more of the compositions C1 to Cn as components using the learned artificial intelligence model and/or the predictive expression other than the artificial intelligence model may be substantially the same as the method in the step S105.

In addition, even in the step S107, a procedure is included in which the predicted blend composition data Ya1 predicted from the color data Xp is obtained as blend composition data containing 1 or more of the above-mentioned compositions C1 to Cn by using multi-label classification, similarly to the step S105.

In the step S107, the predictive color data Xai predicted from the predictive cooperative component data Yai is obtained by using a predictive expression other than the learned artificial intelligent model and/or the artificial intelligent model, and whether or not the color data Xp is acceptable is determined by comparison with the color data Xp, and the method is the same as the predictive expression other than the learned artificial intelligent model and/or the artificial intelligent model in the step S106, and the predictive color data Xa1 predicted from the predictive cooperative component data Ya1 is obtained, and whether or not the color data Xp is acceptable is determined by comparison with the color data Xp.

The step S107 may be a step of repeating the micro correction so that the predicted color data Xa1 is equal to the color data Xp.

In the step S107, a switching step (means) may be provided to use only one of the learned artificial intelligence model and a predictive expression other than the artificial intelligence model. In the present invention, a prediction formula different from a prediction formula used when the failure occurs is preferably used. For example, if the predicted color data Xai obtained by using the specific learned artificial intelligence model in the step S106 is not acceptable, it is preferable to switch the use of the predictive expression other than the specific artificial intelligence model in the following step S107.

The predictive expression may be manually switched, or may be automatically switched when a predetermined condition is satisfied. In the present invention, a predictive expression other than the artificial intelligence model is preferably used for at least 1 of the steps S107.

The predicted composition data Yai obtained in the step S107 can be used as data of 1 or more of the compositions C1 to Cn. The data may also be used as data of the blending amount ratio of each component such as resin, coloring material, solvent, etc. The data of 1 or more of the above-mentioned compositions C1 to Cn are preferably obtained in consideration of workability and the like at the time of actual blending.

The predicted composition data Yai obtained in the step S107 can be used as data of 1 or more of the compositions C1 to Cn. The data may also be used as data of the blending amount ratio of each component such as resin, coloring material, solvent, etc. The data of 1 or more of the above-mentioned compositions C1 to Cn are preferably obtained in consideration of workability and the like at the time of actual blending.

In the predicted blending composition data Yai obtained in step S107, the types of the compositions C1 to Cn are not particularly limited, but 15 types or less, preferably 12 types or less, and more preferably 10 types or less are possible in consideration of workability at the time of actual blending. In this case, the number of the metal pigment-containing compositions is 5 or less, preferably 3 or less, and the number of the pearlescent pigment-containing compositions is 5 or less, preferably 3 or less.

S108 step

The step S108 is a step of acquiring the matching composition data Yap1 that is acceptable when the matching composition data is acceptable in either the step S106 or the step S107. In the present invention, the qualified mating composition data Yap1 may be output, or the data may be transmitted without being output.

The acceptable formulation data Yap1 may comprise formulation data of a coating composition obtained by a tinting process. For example, there may be mentioned data such as the blending ratio of a plurality of commercially available tinting paints and the blending ratio of coloring material components such as pigments to the tinting paint and the blending ratio of 1 or more coloring materials.

In addition, the acceptable blended composition data Yap1 may contain components required for eliminating the difference between the acceptable blended composition and the predicted blended composition and/or data concerning the blending amount thereof. For example, 1 or more data such as the difference of 1 or more components when comparing the fit composition with the predicted fit composition can be exemplified. These differential data correspond to fine adjustment color matching composition data used when fine adjustment is performed from a specific matching composition, and contribute to simplification of the color adjustment operation.

When the qualified match composition data Yap1 is output, a portable telephone such as a monitor, a display, a portable terminal device, a smart phone, or any other output device that can display or output information or an image according to a signal may be used. In addition, an output device such as a printing device that can display information or an image on an appropriate medium such as paper or plastic according to a signal may be used.

The output of the fit-together component data Yap1 may be an output inside the computer, and at this time, the fit-together component data Yap1 output inside the computer is transmitted to an automatic fit device, a terminal device, a data recording medium, or the like by a communication means or the like.

In addition, the acceptable match-up composition data Yap1 may not be output, but may be transmitted to an automatic match-up device, a terminal device, a data recording medium, or the like through a communication means or the like.

S109 step

The step S109 is a step of modulating the actual candidate paint CMap1 based on the acceptable composition data Yap1 to obtain a coated sheet of the actual candidate paint CMap1 and obtaining the actual measurement color data Xap 1.

The method of preparing the actual candidate paint CMap1 is not particularly limited, and may be performed by a known method in preparing paint. For example, the components constituting the actual candidate paint CMap1 may be placed in a mixing container, and mixed by a stirring device, a dispersing device, or the like as necessary, to prepare the paint. Further, 1 or more commercially available compositions may be mixed as a primary color paint to prepare the paint.

In the present invention, the candidate paint CMap1 may be prepared by transmitting the acceptable component data Yap1 calculated by the computer to an automatic blending machine having an electronic balance or the like via a wired or wireless network. Thus, even if the operator is not a skilled person, the actual candidate paint CMap1 can be easily prepared.

The method for obtaining the coated sheet of the actual candidate paint CMap1 is not limited, and may be performed by a known method in preparing the coated sheet. For example, a method of forming a coated sheet is exemplified as follows, in which a coating film of 1 or more layers of the color-adjusting paint is formed on a base material so as to cover a film thickness of not less than 1 layer, and a coating film of the clear paint is formed on the uppermost layer so as to cover a film thickness of 10 to 100 μm of the dry film thickness. When forming each coating film, the coating film may be dried and cured by heating as needed. When drying and curing are performed by heating, the drying and curing may be performed uniformly after all coating films have been formed, or may be performed every time a coating film is formed. The actual candidate paint CMap1 may be produced fully automatically using an automatic paint applicator such as a robot, or by performing a part of the steps by the operator.

The substrate used in the method of the present invention is not particularly limited, and a substrate used for producing a test coated plate for color matching may be used. Examples thereof include metal plates, papers, and plastic films. As long as the size of the base material is such that color measurement can be performed and the degree of color tone can be visually confirmed. The length of the side 1 is usually about 5 to 20cm, for example.

The step of obtaining the actual measurement color data Xap1 by measuring the color of the coated sheet of the actual candidate paint CMap1 may be directly obtained by measurement using a measuring instrument such as a colorimeter, a multi-angle spectrophotometer, a laser metal-sensor, a variable angle spectrophotometer, a gloss meter, or a microscopic brightness sensor, or may be obtained by calculation using data obtained by measurement.

S110 procedure

The step S110 is a step of determining whether or not the paint is acceptable by comparing the color data Xp with the actual measurement color data Xap1 and/or comparing the target color with the color of the paint sheet of the actual candidate paint CMap 1. Whether the test is qualified or not is judged by an operator, a computer or an instrument.

By visually comparing the color of the article (coated sheet) having the target color data Xp with the color of the coated sheet of the actual candidate paint CMap1, the operator can determine whether the article is acceptable or not. For example, the target color data Xp or each element constituting the target color data Xp may be compared with color data obtained by measuring the actual candidate paint CMap1 on a colorimeter or the like or each element constituting the target color data Xp or each element may be determined to be acceptable or not by an operator, a computer, or an instrument. In this case, as in the determination of pass or fail in the steps S106 and S107, a threshold value may be set for the difference, the degree of coincidence, the error rate, etc. among the constituent elements, or various correction values may be used, and pass or fail may be determined by an operator, a computer, or an instrument with reference to these values.

In the present invention, when the computer determines whether it is qualified or not, for example, machine learning may be employed. For example, more than 1 model selected from the group consisting of gradient lifted decision trees, linear regression, logistic regression, simple perceptrons, MLP, neural networks, support vector machines, random forests, gaussian processes, bayesian networks, k-nearest neighbor methods, other machine learning can be used.

In the present invention, 1 or more selected from the group consisting of a neural network, a gradient-lifted decision tree, and a gaussian process is preferably used, and 1 or more selected from the group consisting of a neural network and a gradient-lifted decision tree is more preferably used.

In the invention, SOM (Self-organization Map) in a neural network can be adopted.

In this case, the determination accuracy can be improved for the data used in the learning by improving the SOM by adding an independent data interpretation (for example, setting a vector on the SOM network map and defining the determination direction).

Further, it is preferable to add an improvement to the algorithm of the SOM itself, and in a node of the SOM where data does not exist, a vector estimated from surrounding nodes is defined in advance, whereby it is possible to cope with unknown data. Thus, by arranging the nodes (colors) of similar colors and metallic sensations in the vicinity of the SOM network map by SOM, it is possible to accurately determine whether or not the matching degree is acceptable with respect to the colors (colors and metallic sensations) having learning data and unknown data.

In the present invention, it is preferable that whether the operator is qualified or not is determined by a computer or an instrument in order to reduce the burden on the operator.

In the present invention, the determination of whether or not the operator is qualified based on the visual inspection may be combined with the determination of whether or not the operator, the computer, or the instrument is qualified based on the color data or each element constituting the color data.

When the acceptance or rejection is determined, the operator can be notified of the improvement point of the composition of the actual candidate paint CMap1, if necessary, regardless of whether the acceptance or rejection is established.

When the paint is acceptable in step S110, the paint may be prepared according to the blended composition of the actual candidate paint CMap 1. Further, if necessary, a process of performing fine color matching by an operator may be added without using a computer, and the color of the image may be further brought closer to the target color.

S111 procedure

The step S111 is a step of repeating the steps S105 to S110 or the steps S107 to S110 until the failure in the step S110 is reached.

In the present invention, even when the pass or fail in the step S110 is determined to be 2 or more times, the steps S105 to S110 or the steps S107 to S110 may be repeated until the pass or fail in the step S110 is determined to be a pass.

When the steps S105 to S110 or the steps S107 to S110 are repeated, the predicted color data Xa1 or Xai can be obtained in the steps S105 to S107 by using the learned artificial intelligence model and/or a predictive expression other than the artificial intelligence model as in the previous step. Further, a switching step (means) may be provided to use only one of the learned artificial intelligence model and a predictive expression other than the artificial intelligence model. In the present invention, when the predicted color data Xa1 or Xai obtained by using at least 1 learned artificial intelligence model is not acceptable, it is preferable to switch to a predictive formula other than the artificial intelligence model to be used next.

When the step S105 to S110 or the step S107 to S110 is repeated without being qualified in the step S111, the step S105 to S107 may be switched to use a learned artificial intelligence model and/or a predictive expression other than the learned artificial intelligence model, which have not been used the last time, when the predicted color data Xa1 or Xai is obtained in the step S105 to S107. The switching may be performed manually, or may be performed automatically when a predetermined condition is satisfied. In the present invention, a predictive expression other than the artificial intelligence model is preferably used for at least 1 of the steps S105 to S107.

When the color data is not acceptable in the step S111, it is preferable that the difference Δ between the predicted color data Xa1 or Xai and the actually measured color data is inputted to the computer as the correction coefficient α, and then the steps S105 to S110 or the steps S107 to S110 are repeated.

Thus, a paint that can obtain a desired color can be prepared within 5 times, preferably within 3 times, more preferably within 2 times, of the actual candidate paint preparation.

Method for producing paint by computer toning (embodiment 2 of the present invention)

Fig. 7 is a flowchart when a method for producing a computer-tinting paint according to embodiment 2 of the present invention is performed. The flow shown in fig. 7 is only one embodiment of the present invention.

The method for producing a paint by using a computer-controlled paint according to claim 2 of the present invention is a method comprising the steps S201 to S211 described below using a device having a database in which at least 1 or more kinds of color data X and matching composition data Y of a composition are registered, and a computer in which color matching calculation logic using data registered in the database is used.

The steps S201 to S211 will be described in detail below.

S201 step

The step S201 is a step of inputting learning data to the computer by using the data registered in the database.

The learning data used in step S201 may be the same as the learning data used in step S101 in the method for producing a paint by computer toning according to embodiment 1 of the present invention.

The means for inputting to the computer in the step S201 may be the same as the means for inputting to the computer in the step S101 in the method for producing a paint by computer-based color matching according to the embodiment 1 of the present invention.

S202 procedure

The step S202 is a step of performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating the color data X from the matching composition data Y. The artificial intelligence model in the present invention may be composed of 1 or more models selected from decision trees using gradient boosting, linear regression, logistic regression, simple perceptrons, MLPs, neural networks, support vector machines, random forests, gaussian processes, bayesian networks, k-nearest neighbor methods, and other machine learning. In the present invention, it is preferable to use 1 or more artificial intelligence models selected from decision trees using neural networks and gradient boosting, and gaussian processes, and it is particularly preferable to use 1 or more artificial intelligence models selected from decision trees using neural networks and gradient boosting.

In the present invention, the neural network is learned by using learning data input in the step S201, which is composed of 1 or more selected from the group consisting of decision trees using a neural network, gradient boosting, and gaussian processes, whereby a learned artificial intelligence model including at least 1 of the learned artificial intelligence models in which color data X is inferred from the fitting composition data Y can be generated.

Learning of the artificial intelligence model (neural network) is realized using the learning data input to the computer in step S201. At least the color data X and the composition matching data Y related to more than 1 composition are used as the learning data. As the algorithm of the neural network, a well-known error back propagation algorithm, which is one of teacher learning methods, can be used. The neural network is learned by setting a learning rate (real value between 0 and 1) which is a parameter indicating the learning speed, and an allowable error (real value between 0 and 1) which is an allowable error of the output value during learning. Thus, 1 or more feature values related to the blending composition data Y can be correlated with 1 or more feature values related to the color data X of the coating film of the coating material based on the blending composition data Y. The composition matching data satisfying the color data and the color data based on the composition matching data can be predicted by feedforward calculation using the learned network. The learned network predicts these networks without performing experimental confirmation concerning man-hours such as cost and time.

In the present invention, at least 1 of the steps for generating the learned artificial intelligence model may be the same as at least 1 of the steps for generating the learned artificial intelligence model in step S102 in the method for producing a computer-based paint according to the 1 st aspect of the present invention.

S203 step

S203 is to obtain target color data X of a target color _t Is a step of (a) a step of (b).

As target color data X _t Examples of the material include a coated article, a molded article, a natural structure, and the likeColor data of all colors. In particular, the color data is preferable as color data of a painted object.

The present invention uses the color data of a coating film containing a bright pigment, which has been difficult to be computer-toned so far, as target color data X _t The color can be adjusted with high accuracy. Therefore, the target color data X in step S203 is preferable _t Color data of a coating film containing a lustrous pigment. Of course, the target color data X in step S203 _t Color data of a coating film containing no luminescent pigment may also be used.

Composing target color data X _t The elements constituting the color data registered in the database may be the same as the elements constituting the color data. For example, the color data measured by a measuring instrument or the color data calculated therefrom may be used.

The measuring instrument for obtaining color data is not particularly limited as long as it can measure the color of a bright coating film (a metal coating film, a pearl coating film, or the like), a pure color coating film, or the like to obtain color data, and a conventionally known measuring instrument may be used without limitation to a measuring principle, a method for calculating color data of a measured value, or the like. For example, 1 or more of a single-angle spectrophotometer, a multi-angle spectrophotometer, a colorimeter, a color difference meter, a color change spectrophotometer, and other colorimeter and imaging device, a measuring instrument such as a microscopic brightness measuring instrument, and a measuring instrument such as a color sample card can be used. In addition, a data processing device that processes various color data obtained from these measuring instruments can be arbitrarily used.

The operator can obtain the target color data X by directly measuring the measured color object by using various measuring instruments _t . Further, the measurement data may be automatically acquired by various measuring instruments according to a program or the like. Further, the calculation may be performed based on these colorimetric data.

In the present invention, it is preferable to obtain the target color data X by measuring the surface of the color to be measured using a multi-angle spectrophotometer _t 。

In addition, when the target color data X _t Rather than directly measuring the colour measuredIn the case of the data obtained from the object, color data obtained from the trade name of the object to be measured or the like can be used as target color data X _t And is used. For example, when the target color data X _t When color data on an automobile is used, target color data X can be set based on paint data obtained by the trade name, model, year, manufacturing number, etc. of the automobile _t 。

S204 procedure

S204, inputting the target color data X to the computer _t Is a step of (a) a step of (b).

The means for inputting to the computer in the step S204 may be the same as the means for inputting to the computer in the step S104 in the method for producing a paint by computer-based color matching according to the embodiment 1 of the present invention.

S205 step

Step S205 is to obtain the color data X approximate to the target color data by searching using a computer _t Is to search for color data X _n1 Corresponding to the retrieved color data X _n1 Is approximately matched to form data Y _n1 At the same time for the target color data X _t And the search color data X _n1 And comparing to judge whether the product is qualified or not.

The step S205 may be a step corresponding to a Computer Color Selection (CCS) capable of searching for the approximate target color data X among the plurality of color data registered in the database _t As the search color data X _n1 And is obtained.

Here, the color data registered in the database is, for example, color data of a known color sample account, color data of a paint sheet produced in the past, or the like, and is associated with matching composition data corresponding to the color data. Thus, by acquiring the search color data X _n1 The corresponding matched composition data is approximately matched composition data Y _n1 Can also be easily obtained.

For 1 or more of the elements constituting the color data (for example, each value in the Lxab color system, etc.), and the target color data X are respectively formed _t Can be compared with corresponding elements of the formulaElements of the search value, such as difference, consistency, error rate, etc., within a certain range are obtained to obtain the search color data X _n1 . The predetermined range may be set by an operator with reference to experience or the like, or may be set by a computer.

In step S205, when comparing the target color data X _t And the search color data X _n1 When the color data X is judged to be acceptable or acceptable, the composition of the search color data X can be focused on _n1 More than 1 element of (1) and the composition target color data X _t The comparison is performed for each corresponding component of the above-mentioned components 1 or more. When the pass or fail is determined, for example, a threshold value may be set for the difference, the degree of coincidence, the error rate, or the like in each component, and the pass or fail may be determined by referring to the device or the operator. In this case, the constituent elements may be weighted by reflecting the viewpoint of the skilled worker or the like.

S206 procedure

S206, when the color data is not qualified in the S205, obtaining the predicted target color data X by using a computer _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni And comparing to judge whether the product is qualified or not.

Obtaining predictions as providing target color data X using a computer _t Candidate match composition data Y _ni For example, a method known as computer Color Correction (CCM), that is, a calculation based on color matching calculation logic using a computer or a calculation based on mathematical optimization, may be mentioned.

Based on the calculation of color matching calculation logic using a computer, for example, based on various color data registered in the database and composition matching data corresponding thereto, the target color data X _t Comparing with the various color data to be in a state of difference, consistency and the likeBy performing calculation in a fixed range, one or more fitting compositions considered most reasonable are determined as candidate fitting composition data Y _ni . By using various functions constituting the calculation logic, an arbitrary fitting composition or an approximate fitting composition can be corrected by a small number of repeated steps. In this case, theoretical instructions are generated in a regular pattern, and the calculation speed and the accuracy of the adjustment algorithm can be complemented.

Candidate matching composition data Y _ni For example, by referring to color data recorded in a database, components having characteristic information that acts in a direction of reducing an error are searched for each coordinate axis constituting color data obtained by approximately matching the components, whereby calculation by mathematical optimization can be obtained. For example, in the color system of LXa b, when the error is defined as DeltaLXa=LXa ₂ -L＊ ₁ 、Δa＊＝a＊ ₂ -a＊ ₁ 、Δb＊＝b＊ ₂ -b＊ ₁ When the error DeltaL on the L-axis is positive, a component having characteristic information acting in a direction of decreasing the L-2 value is searched, and when the error DeltaL on the L-axis is negative, a component having characteristic information of increasing the L-2 value is searched. Similarly, when the error Deltaa on the a-axis is positive, the search is performed to reduce the a-axis ₂ The component of the characteristic information (green) of the value is searched for a value having an increase in a when the error Deltaa in a' axis is negative ₂ The component of the characteristic information (red) of the value is searched for a value having a reduced b when the error Deltab on the b' axis is positive ₂ The component of the characteristic information (green) of the value is searched for a component having an increased b when the error Deltab on the b' axis is negative ₂ The characteristic information (yellow) of the value. In this way, in the approximate composition, by adding a component that acts in the direction of reducing the error, that is, that imparts predetermined characteristic information to the coordinate axes of the color system constituting the color space, candidate composition data Y of the color close to the target can be obtained _ni 。

It is assumed that, when a component having characteristic information functioning in the direction of reducing the error is not retrieved, the color data X is calculated from the target color _t Can obtain more suitable newCandidate matching composition data is acquired after the approximate matching composition.

Further, the candidate matching composition data Y may be obtained by performing operator correction (for example, correction of a known color sample account, color data of a coating plate manufactured in the past, correction of an operator by referring to own experience, or the like), correction of the computer, correction of an artificial intelligence model, or the like on the matching composition data obtained by CCM _ni 。

In addition, in a job site such as an automobile repair facility, when the available coloring material or composition is limited, the candidate blending composition data Y may be obtained based on only the coloring material or composition available in the job site _ni 。

Can be combined into data Y by display means or printing means _ni . The information may be transmitted from the computer to an apparatus or the like for performing the next step, instead of being output.

When the predicted color data X is obtained in the S206 process _ni In this case, the artificial intelligence model and/or predictive formulas other than the artificial intelligence model may be used. The artificial intelligence model or predictive formulas other than the artificial intelligence model may be used interchangeably.

In the present invention, the step S206 can be performed to obtain the predicted color data X using only at least 1 artificial intelligence model _ni Is a step of (a) a step of (b). In step S206 after the 2 nd time, the predicted color data X may be obtained without using the artificial intelligence model _ni Is a step of (a) a step of (b).

As predicted color data X available in the S206 process _ni The database may be exemplified by various color data recorded therein. In the present invention, the predicted color data X is preferably _ni Including multi-angle spectral reflectance and/or shine parameters. By making the predicted color data X _ni The color can be adjusted with higher accuracy even for a bright color whose optical characteristics are difficult to predict by including the spectral reflectance and/or the brightness parameter at multiple angles.

In step S206, predicted color data X is obtained as a result of using the learned artificial intelligence model _ni In the method of (2), candidate match composition data Y is input to each unit of an input layer in a neural network of an artificial intelligence model which has been learned _ni The feature quantity of (2) may be set. Candidate matching composition data Y input to input layer _ni The color data is transmitted while being weighted between each node and each layer, and is outputted from each cell of the output layer as color data.

In step S206, predicted color data X is obtained as a result of using a prediction other than the learned artificial intelligence model _ni Various predictive formulas known in the field of color matching using CCM can be used. Examples of such a predictive expression include a method using a predictive expression based on the Kubelka-Munk optical density expression and the Duncan color mixing theory expression, a method using fuzzy inference, and a method of indexing the degree of matching by comparing color data and matching composition data by a computer.

The method using the Kubelka-Munk optical density formula and the Duncan color mixing theory formula can be the same as the method using the Kubelka-Munk optical density formula and the Duncan color mixing theory formula described in step S106 in the method for producing a computer-based paint according to embodiment 1 of the present invention.

The method of using fuzzy inference may be the same as the method using fuzzy inference described in step S106 in the method of producing a paint by computer toning according to embodiment 1 of the present invention.

For example, by separately composing the color data X _t Each element and constitution of predicted color data X _ni By comparing the elements of (a) with each other, whether or not the product is acceptable can be determined. For example, when color data X _t Predicting color data X _ni When the element of the Lxalbumin color system is used and the element obtained therefrom is included, the color difference DeltaE can be compared on the basis of the L, a and b respectively to judge whether the product is qualified or not. In this case, a threshold value may be set for the difference, the degree of coincidence, the error rate, and the like in each component, and whether or not the component is acceptable may be determined by referring to the threshold value.

In the invention, whenThe predicted color data X _ni In the case of color data concerning brightness, it is preferable to determine whether or not the color data is acceptable using spectral reflectance and/or brightness parameters at multiple angles.

When the judgment of the pass or fail is made, the operator can be notified of the pass or fail of the predicted color data X as needed _ni Near target color data X _t Improvement points of (c) and the like.

S207 step

S207 step of obtaining predicted supply target color data X using a computer when the color data X is not acceptable in the S206 step _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni And repeating the step of judging whether the product is qualified or not until the product is qualified by comparing the product with the product.

In step S207, the target color data X predicted to be supplied is obtained by using a computer _t Candidate match composition data Y _ni The method of (2) can be used to obtain the target color data X predicted to be provided by the computer in the step S206 _t Candidate match composition data Y _ni The same method as in (a).

In the step S207, the method of determining whether or not the test is acceptable is the same as the method of determining whether or not the test is acceptable in the step S206.

In step S207, the slave candidate matching composition data Y is acquired _ni Predicted color data X for prediction _ni At this time, at least 1 learned artificial intelligence model and/or predictive formulas other than the artificial intelligence model, which are the same as those of the step S206, may be used. Further, a switching step (means) may be provided to use only at least 1 learned artificial intelligence model or any one of predictive formulas other than the artificial intelligence model. In the present invention, a prediction formula different from a prediction formula used when the failure occurs is preferably used. For example, when at least 1 learned person is used in the S206 process Predicted color data X obtained by an artificial intelligence model _ni If the model is not acceptable, the predictive expression other than the artificial intelligence model is preferably switched to be used in the next step S207.

The predictive expression may be manually switched, or may be automatically switched when a predetermined condition is satisfied. In the present invention, a predictive expression other than the artificial intelligence model is preferably used for at least 1 of the steps S207.

S208 procedure

In the step S208, if the matching composition data Y is acceptable in any of the steps S205 to S207, acceptable matching composition data Y is obtained _C1 Is a step of (a) a step of (b). In the invention, the qualified matched composition data Y can be output _C1 Data may be transmitted without being output.

Qualified match composition data Y _C1 Compounding composition data of the coating composition obtained by the toning method may be included. For example, there may be mentioned data such as the blending ratio of a plurality of commercially available tinting paints and the blending ratio of coloring material components such as pigments to the tinting paint and the blending ratio of 1 or more coloring materials.

In addition, the qualified match composition data Y _C1 The composition and/or data on the blending amount thereof required for eliminating the difference between the acceptable blending composition and the approximate blending composition and/or the candidate blending composition may be included. For example, there may be mentioned 1 or more data such as a difference in the blending components of 1 or more when the matching composition is compared with the approximate blending composition or the candidate blending composition, and a difference in the blending components of 1 or more when the approximate blending composition is compared with the candidate blending composition. These differential data correspond to fine adjustment color matching composition data used when fine adjustment is performed from a specific matching composition, and contribute to simplification of the color adjustment operation.

When the output is qualified, the data Y is formed by matching _C1 In this case, a mobile phone such as a monitor, a display, a mobile terminal device, and a smart phone may be used, and any output device capable of displaying or outputting information or an image according to a signal may be used. In addition, a printing device capable of displaying information or images on an appropriate medium such as paper or plastic according to a signal can be usedAnd an output device.

Qualified match composition data Y _C1 The output of (2) may also be the output of the computer, in which case the qualified match composition data Y is output in the computer _C1 And transmitted to an automatic matching device, a terminal device, a data recording medium, etc. through a communication means, etc.

In addition, the qualified matching composition data Y may not be outputted _C1 But is transmitted to an automation device, a terminal device, a data recording medium, etc. through a communication means, etc.

S209 step

S209 is based on the qualified matching composition data Y _C1 Preparation of actual candidate paint CM _Ci Obtaining the actual candidate paint CM _Ci Obtaining measured color data X by coating a plate _Ci Is a step of (a) a step of (b).

The actual candidate paint CM is not specifically limited _Ci The preparation method of (2) may be performed by a known method in the preparation of a paint. For example, the actual candidate paint CM may be constituted _Ci The components of (a) are placed in a mixing vessel, and mixed by a stirring device, a dispersing device, or the like as needed, to prepare a product.

In the invention, the qualified matching composition data Y calculated by the computer can also be used _C1 To an automatic regulator equipped with an electronic balance via a wired or wireless network, thereby modulating the actual candidate paint CM _Ci . Thus, even if the operator is not a skilled person, the actual candidate paint CM can be easily prepared _Ci 。

With respect to obtaining actual candidate paint CM _Ci The method of coating the plate is not particularly limited, and may be carried out by a known method in the case of preparing a coated plate. For example, a method of forming a coated sheet is exemplified as follows, in which a coating film of 1 or more layers of the color-adjusting paint is formed on a base material so as to cover a film thickness of not less than 1 layer, and a coating film of the clear paint is formed on the uppermost layer so as to cover a film thickness of 10 to 100 μm of the dry film thickness. When forming each coating film, the coating film may be dried and cured by heating as needed. When it is solidified by heatingShi Gansao, curing may be performed after all coating films have been formed, or may be performed every time a coating film is formed. Actual candidate paint CM _Ci The coated sheet (c) may be produced fully automatically using an automatic coating apparatus such as a robot, or may be produced by performing a part of the steps by an operator.

The substrate used in the method of the present invention is not particularly limited, and a substrate used for producing a test coated plate for color matching may be used. Examples thereof include metal plates, papers, and plastic films. As long as the size of the base material is such that color measurement can be performed and the degree of color tone can be visually confirmed. The length of the side 1 is usually about 5 to 20cm, for example.

For actual candidate paint CM _Ci The color of the coated plate is measured to obtain measured color data X _Ci The step (c) may be obtained directly by measurement using a measuring instrument such as a colorimeter, a multi-angle spectrophotometer, a laser metal-sensing measuring instrument, a variable angle spectrophotometer, a gloss meter, or a microscopic brightness-sensing measuring instrument, or may be obtained by calculation using data obtained by measurement.

S210 procedure

The S210 process is to use the color data X _t And the measured color data X _Ci And/or as a comparison of the target color with the actual candidate paint CM _Ci And (3) comparing the colors of the coated plates to judge whether the plates are qualified or not. Whether the test is qualified or not is judged by an operator, a computer or an instrument.

Visual observation of the color X as a target by an operator _t With the actual candidate paint CM _Ci The color of the coated plate is compared, thereby being capable of judging whether the coated plate is qualified or not. In addition, for example, the target color data X may be individually and separately recorded _t Or each element constituting the paint and the actual candidate paint CM _Ci The color data obtained by measurement with a colorimeter or the like of the coated sheet of (a) or each element constituting the same is compared, and whether or not the coated sheet is acceptable is determined by an operator, a computer or an instrument. In this case, the respective components may be determined as to whether or not the components are acceptable in the steps S206 and S207The difference, the degree of coincidence, the error rate, etc. in the element are set as thresholds or various corrections are used, and the operator, the computer, or the instrument refers to these to determine whether or not the element is acceptable.

In the present invention, when the pass or fail is determined by the computer, the pass or fail determination described in step S110 in the method for producing a paint by computer toning according to embodiment 1 of the present invention may be the same as the pass or fail determination described in the step.

In the present invention, it is preferable that whether the operator is qualified or not is determined by a computer or an instrument in order to reduce the burden on the operator.

In the present invention, the determination of whether or not the operator is qualified based on the visual inspection may be combined with the determination of whether or not the operator, the computer, or the instrument is qualified based on the color data or each element constituting the color data.

When the paint is judged to be acceptable, the operator can be notified of the actual candidate paint CM, as required, regardless of whether the paint is acceptable or not _Ci And the improvement point of the matching composition.

When the paint is qualified in the S210 process, the paint CM can be selected according to the actual candidate _Ci Is matched with the components to form the modulated paint. Further, if necessary, a process of performing fine color matching by an operator may be added without using a computer, and the color of the image may be further brought closer to the target color.

S211 step

The step S211 is a step of repeating the steps S206 to S210 when the judgment of whether the product is acceptable or not in the step S210 is not acceptable.

In the present invention, even when the pass or fail is determined 2 times or more in the step S210, the steps S206 to S210 may be repeated until the pass or fail is determined in the step S210.

When the steps S206 to S210 are repeated, the predicted color data X may be obtained in the step S206 and/or the step S207 by using at least 1 learned artificial intelligence model and/or a prediction model other than the artificial intelligence model, as in the previous step _ni . In addition, a switching step (means) may be provided to cause onlyAt least 1 learned artificial intelligence model or any one of predictive formulas other than the artificial intelligence model is used. In the present invention, the predicted color data X obtained when at least 1 learned artificial intelligence model is used _ni If the model is not acceptable, the model is preferably switched to a predictive expression other than the artificial intelligence model.

If the step S211 is not acceptable and the steps S206 to S210 are repeated, the predicted color data X is obtained in the step S206 and/or the step S207 _ni In this case, the method may be switched to use a prediction formula which was not used at the last time among at least 1 learned artificial intelligence models and/or prediction formulas other than artificial intelligence models. The predictive expression may be manually switched, or may be automatically switched when a predetermined condition is satisfied. In the present invention, predictive formulas other than the artificial intelligence model are preferably used for at least 1 of the steps S206 and/or S207.

In addition, when the color data X is not acceptable in the step S211, the predicted color data X is preferably obtained _ni And actually measured color data X _Ci The difference Δ of (a) is inputted to a computer as a correction coefficient α, and then steps S206 to S211 are repeated.

Thereby, the actual candidate paint CM in step S209 _Ci The number of times of the preparation is 5 or less, preferably 3 or less, more preferably 2 or less, and a paint which can obtain a desired color can be prepared.

Application of preparation method of paint based on computer color matching

The method for producing a computer-tinting paint according to embodiments 1 and 2 of the present invention can be used for preparing a paint for obtaining a desired color. In addition, the method can be used for identification of the matching composition of the paint or correction of the matching composition.

In particular, the method for producing a computer-controlled paint according to the first and second embodiments of the present invention can be used for the preparation of a paint for repairing a colored article, for example, a vehicle such as an automobile or a motorcycle, a part thereof, a large vehicle such as a truck, a bus, a trolley, a monorail, a part thereof, or other industrial products. The colored article may in particular also be an article having a single-layer or multi-layer coating film. In particular, the effects of the present invention can be maximally exhibited when a multilayer coating film including a transparent coating film is provided on a coating film containing a bright pigment such as a metallic paint or a true pearl luster.

Method for predicting color data of coating film (embodiment 3 of the present invention)

Fig. 8 is a flowchart when a method of predicting color data of a coating film according to embodiment 3 of the present invention is performed. The flow shown in fig. 8 is only one embodiment of the present invention.

The method for predicting color data of a coating film according to claim 3 of the present invention is a method comprising the steps S301 to S307 of using a computerized color matching device having a database in which at least color data X and matching composition data Y of 1 or more compositions are registered and a computer in which color matching calculation logic using data registered in the database functions.

Here, the steps S301 and S302 are the same as the steps S101 and S102, respectively. The steps S303 to S307 will be described in detail below.

S303 step

S303 is a paint CM for obtaining color data of a predicted paint film _t Is matched with the composition data Y _CM Is a step of (a) a step of (b).

Paint CM to predict color data of coating film _t The paint can be used as a paint for which color or color data is desired, without actually preparing the paint. Thus, for example, when a plurality of colors are produced by trial production at a time, it is possible to easily obtain color or color data without performing the steps of preparing a paint, preparing a coating film by coating, and measuring color data of the coating film.

Matched composition data Y _CM The data on the respective types (trade name, product number, etc.) of the binder, the coloring pigment, the additive component, etc. contained in the coating material are related to the blending amount. Specifically, the composition and the amount of the composition data to be registered in the database may be the same as those of the composition data.

S304 step

S304, the matching composition data Y is input to the computer _CM Is a step of (a) a step of (b).

The data may be input by inputting the target color data X to the computer in the step S104 _t The same means as those of (a).

S305 step

Step S305 is to acquire the matching composition data Y by searching with a computer as needed _CM Is to search for color data X _n1 Is a step of (a) a step of (b).

The step S305 may be a step of searching for a matching component data similar to the matching component data Y from a plurality of matching component data registered in a database in the same step as a Computer Color Selection (CCS) _CM Is used as the searching color data X _n1 And is obtained.

The matching composition data registered in the database is, for example, matching composition data of a paint sold in the market, matching composition data of a paint produced in the past, or the like, and is associated with color data corresponding to the matching composition data. Thus, by obtaining the approximate fitting composition data Y _CM Is used as the searching color data X _n1 And can be easily obtained.

When retrieving the approximate fitting composition data Y from the plurality of fitting composition data _CM In the case of the composition data of (a) 1 or more elements constituting the composition data (for example, the content of pigment of a specific color, etc.), the composition data can be obtained by comparing the composition data with the corresponding elements, and obtaining the composition data by using the elements having a certain range of the difference in the search value, the consistency, the error rate, etc. The predetermined range may be set by an operator with reference to experience or the like, or may be set by a computer.

In the present invention, the step S305 is performed as needed, and there is no problem even if it is not performed.

S306 procedure

S306 is performed whenIn the step S305, the corresponding search color data X is not searched _n1 At the time, or when the step S305 is not performed, data Y is composed from the fit using the at least 1 learned artificial intelligence model or the at least 1 learned artificial intelligence model and predictive formulas other than the artificial intelligence model _CM Obtaining predicted color data X _m1 Is a step of (a) a step of (b).

In the present invention, at least 1 artificial intelligence model is used to form data Y from coordination _CM Obtaining predicted color data X _m1 . In addition, the predicted color data X can be obtained by using at least 1 learned artificial intelligence model and a prediction model other than the artificial intelligence model _m1 . Here, the method using the artificial intelligence model, the predictive expression other than the artificial intelligence model, the method using the same, and the like may be the same as those described in the step S106.

S307 step

S307 step of obtaining the CM coated with the paint according to need _t Measured color data X of coated board _CM And the predicted color data X _m1 And a step of comparing.

By performing the step S307 and feeding back the result, the color data of the coating film can be predicted with higher accuracy.

For example, when actually measuring color data X _CM And predictive color data X _m1 When the deviation of the color data is large, the predicted color data is obtained again by using only the prediction model other than the artificial intelligence model, and the predicted color data is simultaneously compared with the actually measured color data X _CM The comparison is performed, whereby feedback can be performed.

In addition, the predicted color data X may be also _m1 And actually measured color data X _CM The difference Δ of (a) is inputted to a computer as a correction coefficient β, and then steps S305 to S307 are repeated.

Application of method for predicting color data of coating film

For example, when coating such as vehicle coating is performed, the method for predicting color data of a coating film according to the present invention can be used for predicting the color tone of the coating film when preparing the coating.

By using the method of predicting color data of a coating film of the present invention, it is not necessary to generate a specific paint with respect to an individual paint distribution composition, and colors with respect to a plurality of paint distribution compositions specified can be predicted with high accuracy. Among the plurality of paint distribution composition candidates, a paint distribution composition having the smallest color deviation from the specified color can be easily selected. Thus, it is not necessary to prepare each of the paint for the plurality of paint dispensing compositions, and it is not necessary to measure the paint after actually applying the paint to the material to be coated to prepare a coated sheet, and it is possible to obtain corresponding color data.

Computer color matching system (embodiment 4 of the present invention)

The computer toning system according to the 4 th aspect of the present invention is a computer toning system including a database in which 1 or more kinds of composition C1 to Cn (n is an integer of 2 or more) are registered, and a computer in which color matching calculation logic using data registered in the database functions, and includes means S401 to S411.

(S401) means for inputting learning data into the computer by using the data registered in the database.

(S402) means for performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model matching the composition data Y from the color data X.

(S403) means for acquiring color data Xp of a target color whose matching composition Yp is unknown.

(S404) means for inputting the color data Xp to the computer.

(S405) means for obtaining predicted composition data Ya1 predicted from the color data Xp as composition data containing 1 or more of the compositions C1 to Cn as components by using the learned artificial intelligence model and/or a prediction formula other than the artificial intelligence model.

(S406) means for obtaining predicted color data Xa1 predicted from the predicted cooperative constituent data Ya1 using the learned artificial intelligence model and/or a prediction expression other than the artificial intelligence model, and comparing the predicted color data Xa1 with the color data Xp to determine whether or not the image is acceptable.

(S407) when the composition is not acceptable in the means S406, using the learned artificial intelligence model and/or a predictive expression other than the artificial intelligence model, obtaining predicted composition data Yai which is predicted from color data Xp and is different from the predicted composition data up to that, as composition data containing 1 or more of the compositions C1 to Cn as components, and then obtaining predicted color data Xai predicted from predicted composition data Yai using the learned artificial intelligence model and/or a predictive expression other than the artificial intelligence model, and repeating the means of determining whether or not the composition is acceptable by comparing the composition data with the color data Xp until the composition is acceptable.

(S408) means for acquiring the matching composition data Yap1 when the matching is qualified in either of the means S406 or S407.

(S409) means for modulating the actual candidate paint CMap1 based on the acceptable composition data Yap1 to obtain a coated sheet of the actual candidate paint CMap1 and obtaining actual measurement color data Xap 1.

(S410) means for determining whether or not the paint is acceptable by comparing the color data Xp with the actual measurement color data Xap1 and/or comparing the target color with the color of the paint panel of the actual candidate paint CMap 1.

(S411) means for repeating the means S405 to S410 or the means S407 to S410 until the pass is achieved when the pass is not achieved in the means S410.

Here, "1 or more compositions", "databases in which 1 or more compositions C1 to Cn (n is an integer of 2 or more) are registered, and color data X1 to Xn corresponding to the respective composition data" and "a computer that functions by the color matching calculation logic of the data registered in the databases" can be substantially the same as "1 or more compositions", "databases" and "computers" in the apparatus used in the method for producing a paint by computer color matching according to embodiment 1 of the present invention.

In addition, the means S401 to S411 correspond to the means for performing the steps S101 to S111 in the method for producing a paint by computer toning according to the embodiment 1 of the present invention, and thus may be substantially the same as the means described in relation to the steps S101 to S111. Further, even in the automatic blending means, the same automatic blending means as that realized by the automatic blending machine used in the method for producing a paint by computer-based color matching according to embodiment 1 of the present invention can be used.

Computer color matching system (embodiment 5 of the present invention)

The computer toning system according to embodiment 5 of the present invention is a computer toning system including a database in which color data X and matching composition data Y of 1 or more kinds of compositions are registered and a computer in which color matching calculation logic using data registered in the database functions, and includes means S501 to S511.

(S501) means for inputting learning data to the computer by using the data registered in the database.

(S502) means for performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the matching composition data Y.

(S503) obtaining target color data X of a target color _t Means of (3).

(S504) inputting the target color data X to the computer _t Means of (3).

(S505) obtaining the color data X by searching using a computer _t Is to search for color data X _n1 Corresponding to the retrieved color data X _n1 Is approximately matched to form data Y _n1 At the same time for the target color data X _t And the search color data X _n1 Comparing to determine whether it is qualified or notMeans of the method.

(S506) when the result is not acceptable in the means S505, the target color data X is predicted to be provided by using the computer _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni Means for comparing and judging whether the test is acceptable or not.

(S507) when the result is not acceptable in the means S506, obtaining the predicted target color data X by using the computer _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni And means for repeating the means for determining whether the test is acceptable or not until the test is acceptable.

(S508) when the matching composition data Y is qualified in any one of the means S505 to S507, the qualified matching composition data Y is obtained _C1 Means of (3).

(S509) is based on the qualified matched composition data Y _C1 Preparation of actual candidate paint CM _Ci Obtaining the actual candidate paint CM _Ci Obtaining measured color data X by coating a plate _Ci Means of (3).

(S510) is by the color data X _t And the measured color data X _Ci And/or as a comparison of the target color with the actual candidate paint CM _Ci A means for comparing the colors of the coated plates to determine whether the coated plates are acceptable or not.

(S511) means for repeating the above means S506 to S510 when the means S510 is not acceptable.

Here, "1 or more compositions", "database in which color data X and composition data Y of 1 or more compositions are registered", and "computer that functions by color matching calculation logic of data registered in the database" may be substantially the same as "1 or more compositions", "database", and "computer" in the apparatus used in the method for producing a computer-based color-tunable paint according to embodiment 2 of the present invention.

In addition, the means S501 to S511 correspond to the means for performing the steps S201 to S211 in the method for producing a paint by computer toning according to the embodiment 2 of the present invention, and thus may be substantially the same as the means described in relation to the steps S201 to S211. Further, even in the case of the automatic blending means, the same automatic blending means as that achieved by the automatic blending machine used in the method for producing a paint by computer-based color matching according to embodiment 2 of the present invention can be used.

System for predicting color data of coating film (embodiment 6 of the present invention)

The system for predicting color data of a coating film according to claim 6 of the present invention is a system for predicting color data of a coating film, which includes a database in which color data X and matching composition data Y of 1 or more compositions are registered and a computer in which color matching calculation logic using data registered in the database is used, and includes the following means S601 to S607.

(S601) means for inputting learning data into the computer by using the data registered in the database.

(S602) means for performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the matching composition data Y.

(S603) paint CM for obtaining color data of predicted paint film _t Is matched with the composition data Y _CM Means of (3).

(S604) inputting the matched composition data Y into the computer _CM Means of (3).

(S605) if necessary, by causingRetrieving, by computer, the data Y corresponding to the coordinated composition _CM Is to search for color data X _n1 Means of (3).

(S606) when the corresponding search color data X is not searched in the means S605 _n1 At the time, or when the means S605 is not performed, data Y is composed from the fit using the at least 1 learned artificial intelligence model or the at least 1 learned artificial intelligence model and predictive formulas other than the artificial intelligence model _CM Obtaining predicted color data X _m1 Means of (3).

(S607) obtaining a CM coated with the paint according to need _t Measured color data X of coated board _CM And the predicted color data X _m1 Means for comparing.

Here, "1 or more compositions", "database in which color data X of 1 or more compositions and composition data Y are registered", and "computer functioning by color matching calculation logic of data registered in the database" may be substantially the same as "1 or more compositions", "database", and "computer" in the apparatus used in the method for predicting color data of a coating film according to embodiment 3 of the present invention.

In addition, the means S601 to S607 correspond to the means for performing the steps S301 to S307 in the method for predicting color data of a coating film according to embodiment 3 of the present invention, and thus may be substantially the same as the means described in relation to the steps S301 to S307. Further, even in the case of the automatic blending means, the same automatic blending means as that realized by the automatic blending machine used in the method for predicting color data of a coating film according to embodiment 3 of the present invention can be used.

Application software (embodiment 7 of the invention)

The present invention also relates to application software for controlling and operating the computer toning system according to the 4 th and 5 th aspects of the present invention and/or the system for predicting color data of a coating film according to the 6 th aspect of the present invention.

The application software according to claim 7 of the present invention is application software for controlling and operating the computer toning system according to embodiments 4 and 5 of the present invention and/or the system for predicting color data of a coating film according to claim 6 of the present invention to function, thereby functioning to execute the method of the present invention.

The application software according to claim 7 of the present invention may be stored in advance in a recording device such as HDD (Hard Disk Drive) or a flash memory, and the recording device such as HDD (Hard Disk Drive) or a flash memory is provided by a system for predicting color data of a coating film according to claim 4 and 5 of the present invention and/or a system for predicting color data of a coating film according to claim 6 of the present invention, or an apparatus constituting each means of the system. The device may be attached (installed) to an apparatus or the like by using a detachable recording medium or the like such as a wireless or wired communication means, DVD, CD-ROM, USB memory, or the like.

Examples

Hereinafter, the present invention will be described more specifically by way of examples. However, the present invention is not limited to the following examples.

Example 1

The total of 86 color primary colors, metallic primary colors and pearlescent primary colors were selected from the RETAN PG80, RETAN PG HYBRIDECO, RETAN WB ECO EV and RETAN ECO FLEET (all trade names manufactured by Guangxi paint Co., ltd.). The matching composition data and color data are acquired and registered in a database. The color data used 155 reflection spectrum data at 5 angles of 400 to 700nm measured by a multi-angle spectrophotometer (incidence angle 45 degrees, light receiving angle 15 degrees, 25 degrees, face (face) 45 degrees, shading (shading) 75 degrees, 110 degrees). Learning data created using the data registered in the database is input to a computer, and machine learning is performed through a neural network, whereby a learned artificial intelligence model including at least 1 artificial intelligence model obtained by estimating the fitting composition data Y from the color data X is generated.

100 coated sheets (100 colors, each having a different composition of a coating film, and a metal and/or pearlescent coating film containing about 14%) coated with a coating material having unknown composition data were prepared, and the respective color data were obtained and input to a computer color correction device (manufactured by the Guanyi coating company) having the learned artificial intelligence model and a predictive model other than the artificial intelligence model mounted thereon at the same time, and a color matching operation was performed using the learned artificial intelligence model. The color matching load during the color matching operation was evaluated by the color matching accuracy and the reduction effect. The proportion of the pass of the number of times of toning was found to be within 2 times, and the toning accuracy was evaluated in the following stages 1 to 5. The reduction effect represents the man-hour (time) required to obtain the final acceptable color matching, with the ratio of comparative example 1 being 100. The results are shown in Table 1. The toning burden in the toning operation was 40 in example 1 when comparative example 1 was taken as 100, and a 60% reduction effect was obtained from the conventional toning operation.

5: the qualification rate is more than 85 percent

4: the qualification rate is more than 75 percent and less than 85 percent

3: the qualification rate is more than 65 percent and less than 75 percent

2: the qualification rate is more than 55 percent and less than 65 percent

1: the qualification rate is less than 55 percent

Comparative example 1

86 kinds of blended composition data and color data were obtained in total from the same color primary color, metal primary color and pearlescent primary color as in example 1, and registered in a computer color correction device (manufactured by Guanyi paint Co., ltd.) of a conventional model, which was not equipped with an artificial intelligence model.

For 100 sheets Tu Zhuangban similar to example 1, candidate matching compositions were obtained using the above-described computer color correction device, and the toning operation was repeated until a skilled person of 5 years or more had acquired a satisfactory matching, and the toning burden was evaluated in the same manner as example 1. The results are shown in Table 1.

Example 2

The total of 86 color primary colors and pearlescent primary colors were selected from the RETAN PG80, RETAN PG HYBRIDECO, RETAN WB ECO EV and RETAN ECO FLEET (all trade names manufactured by Guangxi paint Co., ltd.). The matching composition data and color data are acquired and registered in a database. The color data used 155 reflection spectrum data at 5 angles of 400 to 700nm measured by a multi-angle spectrophotometer (incidence angle 45 degrees, light receiving angle 15 degrees, 25 degrees, face (face) 45 degrees, shading (shading) 75 degrees, 110 degrees). Learning data created using data registered in a database is input to a computer, and machine learning is performed through a neural network, thereby creating an artificial intelligence model for estimating color data X from matching composition data Y.

100 coated panels coated with a coating material having unknown composition (100 colors including different colors and coating film compositions, and about 14% of metallic and/or pearlescent coating films) were prepared, and the respective color data were obtained and input to a computer color correction device (manufactured by Guanyi coating corporation) having a learned artificial intelligence model and a prediction type other than the artificial intelligence model simultaneously mounted thereon, to perform a color matching operation using the artificial intelligence model. The color matching load during the color matching operation was evaluated by the color matching accuracy and the reduction effect in the same manner as in example 1. The results are shown in Table 1. The toning burden in the toning operation was 40 in example 2 when comparative example 2 was taken as 100, and a 60% reduction effect was obtained from the conventional toning operation.

Comparative example 2

86 kinds of blended composition data and color data were obtained in total from the same color primary colors and pearlescent primary colors as in example 2, and registered in a conventional type of computer color correction device (manufactured by Guanyi paint Co., ltd.) not equipped with an artificial intelligence model.

For 100 sheets Tu Zhuangban similar to example 2, candidate matching compositions were obtained by using the above-mentioned computer color correction device, and the toning operation was repeated until a person skilled in 5 years or more had acquired a satisfactory matching, and the toning burden was evaluated in the same manner as example 2. The results are shown in Table 1.

Example 3

An artificial intelligence model for estimating color data X from the cooperative constituent data Y was generated in the same manner as in example 2, except that instead of the machine learning by the neural network, the machine learning using the learning data was made into the machine learning using the gradient-lifted decision tree. As in example 2, the artificial intelligent model was used to perform the toning operation, and the toning burden during the toning operation was evaluated by the toning accuracy and the reduction effect. The results are shown in Table 1.

TABLE 1

/>

Claims (17)

1. A method for producing a paint, based on computer toning using a device having a database and a computer,

in the database, 1 or more kinds of composition data Y1 to Yn of compositions C1 to Cn each having n of 2 or more integers are registered, color data X1 to Xn corresponding to each composition data,

the color matching calculation logic using the data registered in the database functions in the computer, characterized in that,

comprises the following steps S101-S111,

s101, using the data registered in the database, inputting learning data into the computer,

s102 is a process of performing machine learning using the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model matching the composition data Y from the color data X,

S103 is a step of acquiring color data Xp of a target color whose matching composition Yp is unknown,

s104 is a step of inputting the color data Xp to the computer,

s105 is a step of obtaining predicted composition data Ya1 predicted from the color data Xp as composition data containing 1 or more of the compositions C1 to Cn as components by using the learned artificial intelligence model and/or a prediction formula other than the artificial intelligence model,

s106 is a step of obtaining predicted color data Xa1 predicted from the predicted cooperative constituent data Ya1 by using the learned artificial intelligence model and/or a predictive expression other than the artificial intelligence model, and comparing the predicted color data Xa with the color data Xp to determine whether the color data Xa is acceptable or not,

s107, when the composition is not acceptable in the step S106, obtaining predicted composition data Yai which is predicted from color data Xp and is different from the predicted composition data up to now as composition data containing 1 or more of the compositions C1 to Cn by using a predictive expression other than the learned artificial intelligence model and/or artificial intelligence model, obtaining predicted color data Xai predicted from predicted composition data Yai by using a predictive expression other than the learned artificial intelligence model, and repeating a step of judging whether or not the composition is acceptable by comparing the composition data with the color data Xp until the composition is acceptable,

S108 is a step of acquiring the fit composition data Yap1 when the data is qualified in either the step S106 or the step S107,

s109 is a step of modulating the actual candidate paint CMap1 based on the acceptable composition data Yap1 to obtain a coated sheet of the actual candidate paint CMap1 and obtaining actual measurement color data Xap1,

s110 is a step of determining whether or not the paint is acceptable by comparing the color data Xp with the actual measurement color data Xap1 and/or comparing the target color with the color of the paint sheet of the actual candidate paint CMap1,

s111 is a step of repeating the steps S105 to S110 or the steps S107 to S110 until the failure in the step S110 is reached.

2. A method for producing a paint, which comprises the steps of coloring a paint by a computer using a device having a database and a computer, wherein at least 1 or more of composition data Y and corresponding color data X are registered in the database, and wherein the computer is operated by a color matching calculation logic of the data registered in the database,

comprises the following steps S201 to S211,

s201 is a step of inputting learning data to the computer by using the data registered in the database,

S202 is a step of performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the matching composition data Y,

s203 is obtaining target color data X of the target color _t In the process of (a) and (b),

s204 is to input the target color data X to the computer _t In the process of (a) and (b),

s205, obtaining the color data approximate to the target color data X by searching with a computer _t Is to search for color data X _n1 Corresponding to the retrieved color data X _n1 Is approximately matched to form data Y _n1 At the same time for the target color data X _t And the search color data X _n1 A step of comparing the results to determine whether the test result is acceptable or not,

s206, when the color data is not qualified in the step S205, obtaining the predicted target color data X by using the computer _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni A step of comparing the results to determine whether the test result is acceptable or not,

s207, when the color data is not acceptable in the step S206, obtaining the predicted target color data X by using the computer _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni Comparing, repeating the step of judging whether the product is qualified or not until the product is qualified,

s208 is to obtain the qualified matching composition data Y when the product is qualified in any one of the steps S205 to S207 _C1 In the process of (a) and (b),

s209 is to compose data Y according to the qualified coordination _C1 Preparation of actual candidate paint CM _Ci Obtaining the actual candidate paint CM _Ci Obtaining measured color data X by coating a plate _Ci In the process of (a) and (b),

s210 is a color data X _t And the measured color data X _Ci And/or as a comparison of the target color with the actual candidate paint CM _Ci A step of comparing the colors of the coated plates to determine whether the plates are acceptable or not,

s211 is a step of repeating the steps S206 to S210 when the step is not acceptable in the step S210.

3. A method for predicting color data of a coating film, comprising using a device having a database in which at least 1 or more composition data Y and corresponding color data X are registered, and a computer in which color matching calculation logic using the data registered in the database is operated,

The method includes the following steps S301 to S309,

s301 is a step of inputting learning data to the computer by using the data registered in the database,

s302 is a step of performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the matching composition data Y,

s303 is paint CM for obtaining color data of predicted paint film _t Is matched with the composition data Y _CM In the process of (a) and (b),

s304 is to input the matched composition data Y to the computer _CM In the process of (a) and (b),

s305, according to need, obtaining the matching composition data Y by searching with a computer _CM Is to search for color data X _n1 In the process of (a) and (b),

s306 is a step of, when the corresponding search color data X is not searched in the step S305 _n1 At the time, or when the step S305 is not performed, data Y is composed from the fit using the at least 1 learned artificial intelligence model or the at least 1 learned artificial intelligence model and predictive formulas other than the artificial intelligence model _CM Obtaining predicted color data X _m1 In the process of (a) and (b),

s307 is to obtain the CM coated with the paint according to the need _t Measured color data X of coated board _CM And the predicted color data X _m1 And a step of comparing.

4. The method of producing a paint according to claim 1, wherein the step S105 and/or the step S107 includes a step of obtaining predicted blend composition data Ya1 and/or Yai predicted from the color data Xp as blend composition data containing 1 or more of the compositions C1 to Cn as components by using multi-label classification.

5. The method of producing a paint according to claim 1, wherein the predicted blend composition data Ya1 obtained in the step S105 and/or the predicted blend composition data Yai obtained in the step S107 are blend composition data containing 15 or less of the compositions C1 to Cn as components, and the metal pigment-containing composition is 5 or less and the pearlescent pigment-containing composition is 5 or less.

6. The method of producing a paint according to claim 2, wherein when the paint is not acceptable in the step S211, the predicted color data X is obtained by _ni And the measured color data X _Ci The difference Δ of (a) is inputted to a computer as a correction coefficient α, and the steps S206 to S211 are repeated.

7. The method for producing a paint according to any one of claims 1, 2, and 4 to 6 or the method for predicting color data of a coating film according to claim 3, wherein the composition data Y and the corresponding color data X of 1 or more compositions registered in the database include actual measurement data or include actual measurement data and data calculated from the actual measurement data.

8. The method for producing a paint according to any one of claims 1, 2 and 4 to 6, wherein,

in the step S102 or the step S202, the step of generating the learned artificial intelligence model includes:

(i) A step of learning an artificial intelligence model by using, as learning data, 1 or more kinds of composition data Y and color data X relating to a composition containing no luminescent pigment;

and (ii) a step of learning the artificial intelligence model by using, as learning data, 1 or more kinds of blending composition data Y and color data X of the composition containing the lustrous pigment.

9. The method for producing a paint according to any one of claims 1, 2 and 4 to 6, wherein,

in the step S102 or the step S202, the step of generating a learned artificial intelligence model includes,

and learning the artificial intelligence model by using, as learning data, data of 1 or more kinds selected from the group consisting of the content of the light-reflective pigment in the composition, the content of the light-interference pigment, the content of each color phase of the light-reflective pigment in the composition, the content of each color phase of the light-interference pigment, the content of each color phase of the colorant, and 2 or more total of these contents, and/or shape data of the coloring material contained in the composition.

10. The method of producing a paint according to any one of claims 1, 2 and 4 to 6, wherein the color data Xp in the step S103 or the target color data X in the step S203 _t Color number of coating film containing lustrous pigmentAccording to the above.

11. The method of producing a paint according to any one of claims 1, 2, and 4 to 6, wherein the step of switching to a predictive mode other than the artificial intelligence model is included in the repeating step when the paint fails in the step of determining.

12. The method of producing a paint according to any one of claims 1, 2, and 4 to 6, wherein a computer is used for the determination in the step of determining.

13. The method for producing a paint according to any one of claims 1, 2, and 4 to 6 or the method for predicting color data of a coating film according to claim 3, which is used for repair coating of a vehicle.

14. A computer toning system is provided with:

a database in which 1 or more compositions C1 to Cn each having n as an integer of 2 or more are registered, and color data X1 to Xn corresponding to each composition;

And a computer which functions by using color matching calculation logic of the data registered in the database, wherein,

the system comprises the following means S401-S411,

s401 is means for inputting learning data to the computer by using the data registered in the database,

s402 is means for performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model matching the composition data Y from the color data X,

s403 is a means for acquiring color data Xp of a target color whose matching composition Yp is unknown,

s404 is means for inputting the color data Xp to the computer,

s405 is a means for obtaining predicted composition data Ya1 predicted from color data Xp using the learned artificial intelligence model and/or a prediction formula other than the artificial intelligence model as composition data containing 1 or more of the compositions C1 to Cn as components,

s406 is means for obtaining predicted color data Xa1 predicted from the predicted cooperative constituent data Ya1 by using the learned artificial intelligence model and/or a predictive expression other than the artificial intelligence model, and comparing the predicted color data Xa with the color data Xp to determine whether the color data Xa is acceptable or not,

S407 is a means for obtaining predicted fit composition data Yai predicted from color data Xp and other than the predicted fit composition data up to now as fit composition data containing 1 or more of the compositions C1 to Cn as components by using a predictive expression other than the learned artificial intelligence model and/or artificial intelligence model when the means S406 fails, then obtaining predicted color data Xai predicted from predicted fit composition data Yai by using a predictive expression other than the learned artificial intelligence model and repeatedly judging whether or not the composition data is acceptable by comparing the predicted fit composition data with the color data Xp until the composition data is acceptable,

s408 is a means for acquiring the fit composition data Yap1 when the fit is satisfied in either the means S406 or S407,

s409 is means for modulating the actual candidate paint CMap1 based on the acceptable composition data Yap1 to obtain a coated sheet of the actual candidate paint CMap1 and obtaining actual measurement color data Xap1,

s410 is means for determining whether or not the paint is acceptable by comparing the color data Xp with the actual measurement color data Xap1 and/or comparing the target color with the color of the paint sheet of the actual candidate paint CMap1,

S411 is a means for repeating the means S405 to S410 or S407 to S410 until the pass is reached when the pass is not made in the means S410.

15. A computer toning system is provided with:

a database in which at least 1 or more composition matching composition data Y and corresponding color data X are registered; and a computer which functions by using color matching calculation logic of the data registered in the database, wherein,

the system comprises the following means S501-S511,

s501 is means for inputting learning data to the computer by using the data registered in the database,

s502 is means for performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the matching composition data Y,

s503 is target color data X for acquiring a target color _t Is characterized in that,

s504 is to input the target color data X to the computer _t Is characterized in that,

s505 is to obtain the color data X approximate to the target color data by searching using a computer _t Is to search for color data X _n1 Corresponding to the retrieved color data X _n1 Is approximately matched to form data Y _n1 At the same time for the target color data X _t And the search color data X _n1 Means for comparing the results to determine whether the test result is acceptable or not,

s506 is to use the computer to obtain the prediction to provide the target color data X when the color data is not qualified in the step S505 _t Candidate match composition data Y _ni Thereafter, the candidate matching composition data Y is obtained using the at least 1 learned artificial intelligence model and/or a predictive model other than the artificial intelligence model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni Means for comparing the results to determine whether the test result is acceptable or not,

s507 is to use a computer to obtain the predicted target color data X when the color data X is not qualified in the step S506 _t Candidate match composition data Y _ni Thereafter using the at least 1 learned artificial intelligence model and/or the artificial intelligenceCan obtain the candidate matching composition data Y by prediction outside the model _ni Predicted color data X for prediction _ni At the same time for the color data X _t And the predicted color data X _ni Means for repeating the means for determining whether the test is acceptable or not until the test is acceptable,

s508 is to obtain the qualified matching composition data Y when the data is qualified in any one of the means S505 to S507 _C1 Is characterized in that,

s509 is the composition data Y according to the qualified coordination _C1 Preparation of actual candidate paint CM _Ci Obtaining the actual candidate paint CM _Ci Obtaining measured color data X by coating a plate _Ci Is characterized in that,

s510 is the color data X _t And the measured color data X _Ci And/or as a comparison of the target color with the actual candidate paint CM _Ci A means for comparing the colors of the coated plates to determine whether the plates are acceptable or not,

s511 is a means for repeating the means S506 to S510 when the means S510 fails.

16. A system for predicting color data of a coating film, comprising:

a database in which at least 1 or more composition matching composition data Y and corresponding color data X are registered; and a computer which functions by using color matching calculation logic of the data registered in the database, wherein,

the system comprises the following means S601-S609,

s601 is a means for inputting learning data to the computer by using the data registered in the database,

s602 is means for performing machine learning on the learning data to generate a learned artificial intelligence model including at least 1 artificial intelligence model for estimating color data X from the cooperative constituent data Y,

S603 is a paint CM for obtaining color data of a predicted paint film _t Is matched with the composition data Y _CM Is characterized in that,

s604 is to input the matched composition data Y to the computer _CM Is characterized in that,

s605, according to need, obtaining the matching composition data Y by searching with a computer _CM Is to search for color data X _n1 Is characterized in that,

s606 is when no corresponding search color data X is searched in the means S605 _n1 At the time, or when the means S605 is not performed, data Y is composed from the fit using the at least 1 learned artificial intelligence model or the at least 1 learned artificial intelligence model and predictive formulas other than the artificial intelligence model _CM Obtaining predicted color data X _m1 Is characterized in that,

s607, obtaining the CM coated with the paint according to the need _t Measured color data X of coated board _CM And the predicted color data X _m1 Means for comparing.

17. The computerized tinting system according to claim 14 or 15 or the system for predicting color data of a paint film according to claim 16, wherein the system is provided with automatic reconciliation means for achieving tinting cooperation by performing automatic reconciliation based on the acquired cooperation composition data.