JP2017119425A - Molding optimizing method of injection molding machine - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily and quickly optimize molding conditions and to effectively (effective) conduct improvement related to molding quality.SOLUTION: Molding data during a molding process and evaluation information related to molding quality are set in a data treatment part E as constraint conditions Xc ... based on a mathematical programming on input parameters Df ... and output parameters Ds ..., prediction function F based on learning of a neutral network N is set as an objective function Xp and an optimal treatment program Ps for obtaining an optimized molding condition related to the constraint condition Xc ... and the objective function Xp by the mathematical programming is set, the molding data related to the output parameters Ds ... during the molding process is detected in the data treatment part E during production utilization, the optimized molding condition by the optimization treatment program Ps is obtained based on the molding data related to the output parameters Ds ... and existing molding condition is changed based on the obtained molding conditions.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a molding optimization method for an injection molding machine that is suitable for use in realizing reduction of molding failure and ensuring optimum performance in a molded product by optimizing molding conditions.

  Conventionally, as a method of reducing molding defects of a molded product produced by an injection molding machine, the cause is mainly determined by determining whether the molding data of a molded molded product is good or bad and determining that the molding is defective. A method of changing (fine-tuning) one or more molding conditions is employed, and as this type of technique, the molding condition setting method of an injection molding machine disclosed in Patent Document 1 and Patent Document 2 is known. It has been.

  The molding condition setting method of the injection molding machine disclosed in Patent Document 1 is based on a plurality of ranks corresponding to the degree of filling failure of a molded product, and specific molding conditions are changed in a direction corresponding to each rank so as to eliminate filling failure. When setting the molding conditions, select the rank of defective filling of the molded product, change the molding conditions according to the change amount corresponding to the selected rank, and change the molding A re-molding cycle in which re-molding is performed according to conditions is repeated until a non-defective product is molded, and the size of molding conditions when the non-defective product is molded is set.

  Further, the molding condition setting method for an injection molding machine disclosed in Patent Document 2 is based on a database prepared in advance and known data when setting molding conditions for an injection molding machine using a mold whose mold data is unknown. The first setting process to set the molding conditions, the injection pressure, the injection speed, and the injection start position of the screw are molded under preset constant molding conditions, the injection pressure is changed based on the molded product, and the injection is performed. A second setting step for setting molding conditions relating to pressure, injection speed, and measured value, and a third setting step for performing molding under the molding conditions obtained from the step and adjusting the molding conditions based on the molded product. It is a thing.

JP-A-7-112465 JP-A-7-112473

  However, the above-described molding condition setting method for an injection molding machine has the following problems to be solved.

  First, since the setting of the molding conditions is a setting under a certain environment, it is necessary to reset it when the environment changes. That is, when the external environment such as humidity and air temperature fluctuates due to production time, progress, etc., the physical properties of the molding material are affected or the set value drifts. Therefore, even if the molding conditions are set to the optimum values, if there is a change in the environment, molding defects are likely to increase, and if molding defects increase, the molding conditions will correspond to the increase in molding defects or periodically. Needs to be reset. As a result, an additional setting process accompanying the resetting occurs, resulting in a decrease in productivity due to an increase in man-hours and an increase in cost due to a decrease in production efficiency, and in the most recent production until resetting, molding is performed. A decline in quality is inevitable.

  Secondly, when re-setting, while setting under the constraints of limited information and processing time, setting of molding conditions is basically difficult setting process because various factors are related, Setting by this method is not necessarily sufficient from the viewpoint of optimizing the molding conditions. In particular, if processing capacity for a large amount of data is ensured and time constraints are not imposed, it will be possible to optimize the molding conditions to some extent. Since there is a limit from the point of view of realizing the system, it is difficult to use the system in which processing is performed in shot units in an actual production site and applied to subsequent production.

  An object of the present invention is to provide a method for optimizing the molding of an injection molding machine that solves the problems existing in the background art.

  In order to solve the above-described problem, the molding optimization method for the injection molding machine M according to the present invention uses the data processing unit E having a computer function to perform molding data related to the input parameters Df including the molding conditions of the injection molding machine M. When a prediction function F... Based on the learning of the neural network N is obtained from the molding data related to the output parameter Ds... Obtained by the trial molding based on the input parameter Df... And the predetermined molding conditions are optimized by the prediction function F. In advance, in the data processing unit E, molding data at the molding process and evaluation information related to molding quality are set as a constraint condition Xc based on mathematical programming for molding data related to the input parameter Df... And output parameter Ds. And setting the prediction function F ... as the objective function Xp ... Then, an optimization processing program Ps for obtaining an optimized molding condition related to the input parameter Df... Satisfying the constraint condition Xc... And the objective function Xp. The molding data related to the output parameter Ds... During the molding process at the time of shot is detected, and the molding conditions optimized by the optimization processing program Ps are obtained based on the molding data related to the output parameter Ds. The present invention is characterized in that processing for changing existing molding conditions is performed according to conditions.

  In this case, according to a preferred embodiment of the invention, the molding data in the molding process set as the constraint condition Xc... Can be selected from the molding data in the metering process and / or the injection process. At least one of injection speed, speed control-pressure control switching position, and holding pressure can be selected. In addition, as evaluation information related to molding quality, information related to molding defects of molded products or information related to performance evaluation of molded products can be selected, and in particular, molding shots of molded products include short shot defects or burrs. Can be selected. A nonlinear function can be applied to the function related to the objective function Xp.

  According to the molding optimizing method of the injection molding machine according to the present invention by such a method, the following remarkable effects can be obtained.

  (1) The molding data related to the output parameter Ds... During the molding process at the time of one or more shots is detected and optimized by the optimization processing program Ps based on the molding data related to the output parameter Ds. Since the molding conditions are determined and the existing molding conditions are changed according to the determined molding conditions, the molding data set during the molding process and the evaluation information (molding defects, etc.) regarding the molding quality set as the constraint condition Xc. In addition, the prediction function F... Set as the objective function Xp... Can optimize the molding conditions easily and quickly, and can effectively (effectively) improve the molding quality.

  (2) Since the prediction function F... Is applied as the objective function Xp..., It is possible to model a certain physical phenomenon by learning a predetermined number of detection data obtained by trial molding and pattern recognition. This makes it possible to easily simulate the molding result from the molding conditions based on the input parameters Df... Without actually operating the injection molding machine.

  (3) After setting (optimizing) the molding conditions, the external environment such as humidity and temperature varies depending on the production time and progress, etc., and the physical properties of the molding material are affected or the set values drift. Even in such a case, since the optimization process can be speeded up, the optimization with respect to the molding conditions can be performed in units of one or more shots. As a result, it is possible to always carry out production in the best condition optimized, such as avoiding a situation where the molding quality temporarily decreases.

  (4) Since optimization with respect to molding conditions is automatically performed in units of one or two or more shots, a separate setting step associated with resetting is unnecessary. As a result, the productivity can be improved along with the reduction of man-hours and the cost can be reduced along with the improvement of the production efficiency, and it can be used as a molding optimization method with high practicality at an actual production site.

  (5) According to a preferred embodiment, the molding data in the metering process and / or the injection process is selected as the molding data in the molding process that satisfies the constraint condition Xc. If one or more control-pressure control switching positions and holding pressures are selected, the molding data related to the behavior of the resin in the injection process can be optimized even if the weighing process is affected by disturbance. This makes it possible to reduce molding defects (short shot defects and / or burr defects) related to the filling amount of the mold.

  (6) According to a preferred embodiment, if information relating to molding defects of molded products or information relating to performance evaluation of molded products is selected as evaluation information relating to molding quality, molding defects or molded product performance largely related to molding quality Therefore, it is possible to implement as an optimum form from the viewpoint of improving the molding quality.

  (7) If a non-linear function is applied to the objective function Xp according to a preferred embodiment, it is possible to cope with various problems expressed as an arbitrary non-linear function, which is most desirable in optimizing the molding conditions of the injection molding machine M. It can be used as an embodiment.

The flowchart for demonstrating the process sequence at the time of the actual production in implementing the shaping | molding optimization method which concerns on suitable embodiment of this invention, A flow chart for explaining a processing procedure before production when carrying out the molding optimization method, Block system diagram showing the processing system of an injection molding machine (experimental machine) and data processing unit capable of implementing the same molding optimization method, An experimental data graph showing the relationship between the number of product defects and the number of revolutions of the material feeder in a short multi-piece molded product to explain the effectiveness of the molding optimization method, System outline diagram for explaining the principle of the neural network used to implement the molding optimization method, Block system diagram for explaining a pre-production processing system when performing the molding optimization method, Molding data table related to input parameters including molding conditions used for implementation of the molding optimization method and molding data table related to output parameters obtained by trial molding, A graph showing an example of a prediction function by a neural network used for implementing the molding optimization method, A data graph showing the relationship between torque (supply amount) and the number of product defects in a multi-piece molded product in a short state to explain the effectiveness of the molding optimization method, An experimental data graph showing the relationship between the number of product defects and the number of revolutions of the material feeder in a short multi-piece molded product to explain the effectiveness of the molding optimization method and when the conditions are changed in the middle , A fluctuation characteristic graph of the VP switching position (actually measured value) for each shot when the rotational speed of the material feeder is changed during molding by the molding optimization method; Fluctuation characteristics graph of holding pressure (actually measured value) for each shot when changing the rotation speed of the material feeder during molding by the molding optimization method, Variation characteristic graph of injection speed (actual measurement value) for each shot when changing the rotation speed of the material feeder during molding by the molding optimization method,

  Next, preferred embodiments according to the present invention will be given and described in detail with reference to the drawings.

  First, configurations of an injection molding machine M (experimental machine) and a data processing unit E that can perform the molding optimization method according to the present embodiment will be described with reference to FIGS.

  In FIG. 3, M is an injection molding machine partly drawn with phantom lines, and includes a machine base Mb, and an injection device Mi and a mold clamping device Mc mounted on the machine base Mb. The injection device Mi includes a heating cylinder 31, and a screw that rotates and advances and retracts inside the heating cylinder 31, and an injection nozzle that does not appear in the drawing at the front end of the heating cylinder 31. On the other hand, the rear portion of the heating cylinder 31 is usually provided with a hopper for supplying a molding material (resin pellet), but when used as an experimental machine, a material supply machine 32 capable of controlling the supply amount is attached instead of this hopper. Used. The material supply machine 32 includes a material container 32 d that stores the molding material, and a rotation supply unit 32 m that supplies the molding material stored in the material container 32 d to the inside of the heating cylinder 31. The rotation supply unit 32m can increase or decrease the supply amount of the molding material supplied to the heating cylinder 31 by variably controlling the rotation speed. The mold clamping device Mc includes a mold 33 that is a combination of a movable mold and a fixed mold. A side panel 34 is provided on the machine base Mb, and a liquid crystal display or the like is used for the side panel 34. A display 35 with a touch panel is provided.

  On the other hand, the injection molding machine M incorporates a molding machine controller 41 having a computer function for performing various control processes and arithmetic processes and performing communication processing with the outside. 32m and display 35 are connected. Furthermore, a sensor group 42 including various sensors for detecting (measuring) various physical quantities such as temperature, rotational speed, electric power, pressure, speed, position, time and the like during operation of the injection molding machine M is provided. Sensor) is connected to an analog input port (or digital input port) of the molding machine controller 41. Therefore, the molding machine controller 41 includes at least a measurement result capture processing function unit Ui that captures a part or all of the measurement results of the sensor group 42 as a function unit used for performing the molding optimization method according to the present embodiment. In addition, a sampling processing function unit Us that samples the measurement result at a constant period to obtain a detection value (molded data) is provided.

  On the other hand, Eo is a general computer system partially drawn with virtual lines, Ed is a display, and Ek is a keyboard. In addition, the display Ed includes a computer main body Em, and the computer main body Em has a general-purpose computer function for performing various control processes and arithmetic processes and for performing communication processes with the outside. The data processing unit E used in the molding optimization method according to the present embodiment is configured.

  Therefore, the computer main body Em includes hardware such as a CPU and an internal memory, and the internal memory is based on a processing program used for performing the molding optimization method according to the present embodiment, that is, a mathematical programming method. An optimization processing program Ps is stored, and a prediction function creation processing program Pn for creating a prediction function F... Based on learning of the neural network N is stored. The computer main body Em and the molding machine controller 41 can communicate with each other by being connected by a LAN system or the like.

  In this computer main body Em, in relation to the optimization processing program Ps, the molding data during the molding process and the evaluation information related to the molding quality, which are the constraint conditions Xc based on the mathematical programming method, are set, and the objective function Xp Has a function of setting a prediction function F.

  In the present embodiment, the molding data in the injection process, specifically, the injection speed, the speed control-pressure control switching position, and the holding pressure are set as the molding data in the molding process that satisfies the constraint condition Xc. Measurement information indicating the measurement state, specifically, measurement torque, was set as molding data. As described above, the molding data in the metering process and / or the injection process is selected as the molding data in the molding process that satisfies the constraint condition Xc. In particular, the injection speed, speed control and pressure control are included in the molding data in the injection process. By selecting one or more switching positions and holding pressures, it is possible to optimize molding data related to resin behavior in the injection process, even when the weighing process is affected by disturbances. There is an advantage that molding defects (short shot defects and / or burr defects, etc.) related to the filling amount of the mold can be reduced.

  In addition, as evaluation information related to molding quality, information related to molding defects of molded products or information related to performance evaluation of molded products can be set. Especially, molding defects of molded products include short shot defects and / or variability. Defects can be included. In the present embodiment, as evaluation information, the number of products (the number of product defects) having a short shot defect when a short state occurs is set in a so-called multi-piece molded product that simultaneously molds a plurality of products. As described above, if the information related to the molding defect of the molded product or the information related to the performance evaluation of the molded product is selected as the evaluation information related to the molding quality, it is effective for the molding defect or the performance of the molded product greatly related to the molding quality. Therefore, there is an advantage that it can be implemented as an optimum form from the viewpoint of improving the molding quality.

  Furthermore, since the prediction function F... Is applied as the objective function Xp..., It is possible to model a certain physical phenomenon by learning a predetermined number of detection data obtained by trial molding and causing pattern recognition. Thus, there is an advantage that the simulation of the molding result can be easily performed from the molding condition by the input parameter Df... Without actually operating the injection molding machine.

  FIG. 4 is a graph illustrating the number of product defects and the number of rotations of the feeder in a short multi-piece molded product when the molding optimization method is not applied to explain the effectiveness of the molding optimization method according to the present embodiment. This is experimental data showing the relevance. Usually, a molding material (resin pellet) is accommodated in the hopper, and the molding material in the hopper is appropriately supplied into the heating cylinder 31 during the molding process. However, when the external environment such as humidity and air temperature fluctuates due to the time of production, progress, etc., these fluctuations act as disturbances on the supply amount supplied to the heating cylinder 31. In particular, when the supply amount to the heating cylinder 31 is reduced due to the influence of the physical properties of the molding material, this reduction directly causes a short circuit.

  Further, in the case of a multi-piece molded product to be molded in the present embodiment, it is not easy to fill the resin evenly in all the product cavities due to the gate balance of the mold. Insufficient filling tends to occur in a part of the molded product, that is, in a part of the product cavities that exist in large numbers. However, good products can be obtained in other product cavities except for product cavities that are underfilled. In the case of such a multi-piece molded product, if insufficient filling is brought close to zero, burrs are likely to occur in other product cavities, and it is often difficult to eliminate molding defects. The short shot defect is generally a molding defect that occurs when the amount of resin filled in the mold cavity (product cavity) is insufficient.

  Therefore, the injection molding machine M of FIG. 3 configured as an experimental machine is used, and as shown in FIG. 4, the rotation speed of the material supply machine 32 (feed machine rotation speed) is increased from 50 [rpm] to 90 [rpm]. The supply amount was increased by increasing the speed of the product, and the number of product defects in the short multi-piece molded product that occurred at this time was confirmed. As a result, as is apparent from FIG. 4, when the rotational speed of the feeder is 50 [rpm], the number of product defects is 8 [pieces] (out of 16 molded products, 8 non-defective products, short shot defective products) However, when the rotational speed of the feeder was increased to 90 [rpm], the supply amount increased and the number of product defects became 0 [pieces]. This result shows that the amount of resin filled in the mold has changed due to changes in the external environment, and even if the molding conditions are constant due to changes in the external environment, short shot defects and burrs have failed. This suggests the possibility of conflicting defects.

  In other words, when a disturbance such as the external environment fluctuates, it can be confirmed that the number of product defects fluctuates due to fluctuations in the filling amount into the product cavity, and the molding conditions related to the supply amount of the molding material can be optimized. For example, the number of product defects can be suppressed by controlling the filling amount into the product cavity. In the present embodiment, even when the supply amount of the molding material supplied to the heating cylinder 31 fluctuates due to disturbance, the prediction function creation processing program Pn and the optimization processing program used in the molding optimization method according to the present embodiment The molding conditions were optimized by Ps, so that the number of product defects per shot was always almost constant.

  Hereinafter, basic processing functions of the prediction function creation processing program Pn and the optimization processing program Ps will be described with reference to FIGS.

  First, the prediction function creation processing program Pn has, as a basic processing function, molding data relating to input parameters Df including the molding conditions of the injection molding machine M, and output parameters Ds obtained by trial molding based on the input parameters Df. A prediction function F... Based on learning of the neural network N is obtained from the molding data relating to the above, and a processing function for predicting a predetermined molding condition by the prediction function F.

FIG. 5 shows a multilayer structure model of the neural network N. A pattern to be processed is input to the input layer Ni, and the input is weighted by the intermediate layer Nm and transmitted to the output layer Mo. The total transmission amount o _i is expressed by [Equation 1]. In [Expression 1], x _i represents an input signal from the intermediate layer Nm, that is, an output signal from the intermediate layer Nm to the output layer No, and wij represents a weight.

  In this case, the intermediate layer Nm receives a signal input from the input layer Ni and outputs a signal to the output layer No. Further, the output layer No that receives signals from the plurality of intermediate layers Nm performs final output in response to each signal. As the output function at this time, the sigmoid function shown in [Expression 2] is used. In addition to the sigmoid function shown as an example, a Gaussian function, a trigonometric function, or the like can be used as the output function.

  FIG. 6 shows a pseudo prediction model for obtaining the prediction function F... By the neural network N using the molding data relating to the weighing process and the injection process. That is, the detection data 52 is obtained by the trial molding 51. The detection data 52 includes molding data related to input parameters Df... (Input layer Ni) including molding conditions of the injection molding machine M, specifically molding data 53 related to the injection speed of the screw, VP switching position. Molding data 54 relating to (speed control-pressure control switching position), molding data 55 relating to the holding pressure applied to the resin in the mold 33 during the pressure holding period, and the like are included, and based on this input parameter Df. Molding data related to the output parameter Ds (output layer No.) obtained by the trial molding 51, specifically, molding data 56 related to the weight of the molded product, molding related to the number of molded product defects (the number of product defects in the embodiment). Data 57 and the like are included. Then, based on each shaping data, a target prediction function F... Is obtained by learning of the neural network N. Further, the input parameter Df... In the present embodiment includes weighing information 58, specifically, a weighing torque that represents a load accompanying the screw rotation of the screw in the weighing process.

  In addition, the injection process which is the object of the present embodiment includes an injection filling period in which the molten resin is injected and filled into the cavity of the mold 33, and the resin in the mold 33 after the injection filling period ends. A holding pressure period in which a predetermined holding pressure is applied is included. In the injection process, the screw advances at a predetermined injection speed, the resin is injected and filled into the mold 33 from the heating cylinder 31 (injection nozzle), and a speed control-pressure control (VP control) switching position is set. Therefore, in this VP control switching position, the injection filling period in which the injection speed is controlled is switched to the pressure holding period in which the holding pressure is controlled.

  7 and 8 show an example in which the prediction function F... Is obtained by the neural network N. FIG. FIG. 7 shows molding data at the time of the injection process that becomes the constraint condition Xc when the rotating speed of the feeder is fixed at 60 [rpm], that is, VP control switching position [mm], holding pressure [MPa], injection The molding data (detection data) of the speed [mm / s] is shown, and the number of product defects and the weight [kg] of the molded product, which are the objective functions Xp. Specifically, in the case of “forming No. 1”, the VP control switching position is “7.4 [mm]”, the holding pressure is “28 [MPa]”, and the injection speed is “195 [mm / s]”. As a result of performing multiple shots under these molding conditions, the average number of product defects is “3.9 [piece]” and the average weight is “0.633 [kg]”. It shows that.

  Therefore, the injection speed, the VP control switching position, and the holding pressure are input parameters Df..., And the number of product defects and the weight of the molded product are output parameters Ds. These data are taken into the molding machine controller 41 as predetermined molding data (detection data 52) and sent to the computer main body Em constituting the data processing unit E. In the present embodiment, the above-described measurement information (measurement torque) is also included in the input parameter Df.

  Further, the prediction function creation processing program Pn using the neural network N learns a predetermined number of detection data 52 obtained by trial molding and causes a pattern to be recognized, thereby modeling a certain physical phenomenon in a pseudo manner. Thereby, even if the injection molding machine M is not actually operated, the molding result can be simulated according to the molding conditions based on the input parameters Df.

  FIGS. 8A and 8B show the results of learning by the neural network N, and the pseudo correlation curves La and Lb are obtained by plotting the detection data 52 (the plot is not shown). can get. The correlation curves La and Lb are prediction functions for predicting the output parameter Ds... By each input parameter Df... (Injection speed, VP control switching position). FIG. 8A shows a correlation curve La showing a change in the number of product defects when the injection speed is changed, and FIG. 8B shows a change in the number of product defects when the VP control switching position is changed. Is a correlation curve Lb. 8C is a correlation curve K showing the change in the number of product defects when the measured torque [N · m] included in the input parameter Df...

  On the other hand, a prediction function F (prediction model) in which the detection data in FIGS. 7 and 8 is the input layer Ni (input parameter Df...) And the number of product defects and the weight is the output layer (output parameter Ds...) Is created. Can do. Then, if a learning process is performed by the prediction function creation processing program Pn (neural network N), a correlation curve similar to that in FIG. 8 can be obtained. Specifically, a pseudo correlation curve in which the learned molding data is plotted is obtained, and this correlation curve becomes a target prediction function F (prediction model), and the number of product defects (output parameter Ds) is determined as a target. The molding conditions (input parameters Df...) For making the number can be predicted. This prediction model is used as the objective function Xp of the optimization processing program Ps shown below.

  The optimization processing program Ps has, as a basic processing function, a processing function for obtaining an optimized forming condition related to the input parameter Df... Satisfying the constraint condition Xc... And the objective function Xp by mathematical programming. For the shaping data relating to the input parameter Df... And the output parameter Ds..., The constraint condition Xc... And the objective function Xp set based on the mathematical programming method are used.

  In order to facilitate understanding of the processing function of the optimization processing program Ps, an algorithm for optimizing the molding conditions based on mathematical programming will be described. Various algorithms can be considered as an algorithm to be optimized using the constraint condition Xc... Set based on mathematical programming and the objective function Xp. As an example suitable for this embodiment, one of the interior point methods is used. An algorithm using the trust region method is described.

  Now consider the optimization problem of the nonlinear function shown in [Equation 3]. If a nonlinear function is applied to the objective function Xp, various problems expressed as an arbitrary nonlinear function can be dealt with. Therefore, there is an advantage that it can be used as the most desirable mode in optimizing the molding conditions of the injection molding machine M. is there. F (x) in [Equation 3] indicates one output layer No in the neural network N.

  In the trust region method, optimization can be performed by minimizing or maximizing a quadratic approximation model of the objective function Xp.

First, the search direction is determined by provisionally giving the size (reliability radius) of a region (reliability region) where the quadratic approximation model is considered appropriate. When the confidence radius in the k-th iterative calculation is Δ ^k , the search direction can be obtained as a solution to the subproblem of [Equation 4], considering a quadratic approximation model of f (x).

Next, an initial point x ⁰ and an initial confidence radius Δ ⁰ are given. The parameters 0 <η ₁ ≦ η ₂ <1, 0 <γ ₁ <1 <γ ₂ are determined, and k = 0.

As a result, if it can be determined that x ^k is an approximation of the local optimal solution, the process ends.

On the other hand, p ^k is ^obtained by solving the subproblem [Formula 4]. Thus, if [Equation 5] holds, x ^{k + 1} = x ^k + ^pk is set.

At this time, if [Equation 6] holds, Δ ^{k + 1} ∈ [Δ ^k , γ ₂ Δ ^k ] is used, and the trust region is expanded. Otherwise, Δ ^{k + 1} = Δ ^k Maintain the current status.

If [Equation 5] does not hold, Δ ^{k + 1} ∈ [γ ₁ Δ ^k , Δ ^k ] is used to reduce the trust region.

  Then, k = k + 1 is set and the end condition is checked.

  The above is an algorithm using the trust region method which is one of the interior point methods in mathematical programming, and the data processing unit E uses the above-described prediction function F (prediction function) from the molding data related to the acquired output parameter Ds. The molding conditions optimized by the creation processing program Pn) and the optimization processing program Ps can be obtained.

  Next, a specific processing procedure of the molding optimization method according to the present embodiment will be described according to the flowcharts shown in FIGS. 1 and 2 with reference to FIGS.

  First, a processing procedure before production operation when the molding optimization method according to this embodiment is performed will be described with reference to a flowchart shown in FIG.

  The process before the production operation basically performs trial molding, collects the molding data related to the input parameter Df... And the molding data related to the output parameter Ds... And the prediction function F. This is a process of creating the constraint condition Xc... And the objective function Xp.

  In the molding optimization according to the present embodiment, the molding conditions are optimized in order to further improve the molding quality, such as increasing the yield of molded products and ensuring optimum performance. Therefore, first, the current state of the target injection molding machine M is grasped (step S31). That is, input parameters Df... And output parameters Ds... That are control factors related to optimization of the molding conditions are set. Specifically, as described above, the present condition such as “the supply amount of the molding material supplied to the heating cylinder 31 is related to the number of product defects” is grasped and becomes a control factor related to the input parameter Df. The injection speed, the VP control switching position, and the holding pressure are set as molding conditions, and a measurement torque that is measurement information related to the supply amount of the molding material supplied to the heating cylinder 31 is set (step S32). . In this case, more accurate processing can be performed by setting the above-described metering torque as a molding condition. Further, a molding defect type that is a control factor related to the output parameter Ds... Is set (step S33). In the example, short shot defects (number of product defects) were selected as the type. By selecting the short shot defect as a control factor, it is possible to control the fluctuation of the filling amount into the cavity of the mold 33.

  When the control factor setting is completed, trial molding is performed using the input parameters Df... (Steps S34 and S35). At the time of trial molding, the set control factor, that is, molding data relating to the input parameter Df... And molding data relating to the output parameter Ds... Are collected (step S36). Specifically, the measurement result measured by the sensor group 42 is sampled at a constant period to obtain detection data (molded data).

  When the trial molding is performed a predetermined number of times and the molding data collection is completed, the prediction function creation processing program Pn is executed, and the prediction function F... Based on learning of the neural network N is obtained from the collected molding data (steps S37 and S38). ). When the prediction function F... Is obtained, it is incorporated into the optimization processing program Ps as an optimization model (steps S39 and S40).

  On the other hand, the constraint conditions Xc... And the objective function Xp... Based on the mathematical programming necessary for functioning the optimization processing program Ps are set. That is, as the constraint condition Xc..., An injection speed, a VP control switching position, a holding pressure, and a range for product weight related to molding quality, a limit value, and the like, which are molding data during the molding process, are set (step S41). ). Further, a molding defect per shot, that is, an allowable number of product defects (average) in a multi-piece molded product, for example, “2” or more is set (step S42). Furthermore, a prediction function F... For predicting the number of product defects is set as the objective function Xp for the purpose of minimizing the prediction function F. That is, it is possible to set an upper limit of the filling amount that does not generate burrs in the cavity of the mold 33, and to maximize the filling amount within this range.

  FIG. 9 shows the processing principle of the molding optimization method according to the present embodiment when the allowable number of product defects per shot is set to N as the constraint condition Xc. FIG. 9A shows a case where the molding optimization method according to the present embodiment is not applied. As molding conditions, the injection speed is 200 mm / s and the VP control switching position is 7.5 mm. The holding pressure is set to 30 [MPa]. In this case, the Bt point predicts that about 5.6 product defects will occur when the weighing torque is 16.0 [N · m]. FIG. 9A is the same as FIG. 8C.

  On the other hand, when the molding optimization method according to the present embodiment is applied, the injection speed is changed even if the metering torque is 16.0 [N · m], as shown in FIG. 9B. As a result, the number of product defects at the point Bv can be reduced to 2.4. FIG. 9B is the same as FIG. That is, it is shown that the number of product defects can be reduced to N by searching for the optimum values of molding conditions such as the input parameter Df... Injection speed, VP control switching position, holding pressure, metering torque, and the like.

  Next, a processing procedure during actual production operation when the molding optimization method according to the present embodiment is performed will be described with reference to the flowchart shown in FIG.

  First, the operation (operation) of the injection molding machine M is started (step S1). In this case, the injection molding machine M is used as a normal injection molding machine M by removing the material supply machine 32 from the aforementioned experimental machine and attaching a hopper instead of the material supply machine 32. Further, the injection molding machine M is in the automatic operation mode, the mold clamping device Mc is in the mold open state, and the injection device Mi is in the measurement end state.

  On the other hand, a mold clamping process is performed at the start of the molding process (steps S2 and S3). In the mold clamping step, the movable mold in the mold open state moves forward with respect to the fixed mold, and the mold 33 is clamped by the set mold clamping force. In addition, an injection process is performed in the injection device Mi (steps S4 and S5). In the injection process, the screw in the heating cylinder 31 moves forward. As a result, the measured molten resin is injected from the injection nozzle and filled into the cavity of the mold 33. In the injection process, an injection filling process is performed in which molten resin is injected and filled into the cavity of the mold 33 by the forward movement of the screw, and the pressure holding process is performed when the screw reaches the VP control switching position. Is called. That is, at the VP control switching position, the speed control is switched to the pressure control, and a predetermined holding pressure is applied to the resin in the mold 33.

  A metering process is performed by the completion | finish of an injection process (step S6, S7). In the measuring step, the screw in the heating cylinder 31 rotates at the set rotation speed, and the molding material (resin pellet) in the hopper is taken into the heating cylinder 31. As a result, the molding material is plasticized and melted by the temperature of the heating cylinder 31 and the rotation of the screw, and is transferred to the front to be measured and accumulated in the front part of the heating cylinder 31.

  Upon completion of the weighing process, the mold clamping device Mc performs a mold opening process (step S8). In the mold opening process, after the mold 33 is cooled, the movable mold in the mold clamping state moves backward from the fixed mold to the mold opening position, and the mold 33 is opened. When the mold opening process is completed, a molded product removing process is performed (step S9). In the molded product removal step, the molded product attached to the fixed mold is taken out by an ejector mechanism attached to the mold 33.

  Thus, the molding process (molding cycle) for one shot of the injection molding machine M is completed. Thereafter, the same molding process is repeated until the production (production plan) is completed (steps S10, S2,...). If the planned production is completed, the operation (operation) of the injection molding machine M is stopped (steps S10 and S11).

  On the other hand, in each molding cycle, the molding optimization process is performed according to the molding optimization method according to the present embodiment.

  First, when the weighing process is started, weighing information including weighing torque and the like is acquired (steps S6 and S12). This measurement information is used to determine whether or not the molding condition needs to be changed by predicting the filling amount into the cavity of the mold 33 from the state of the measured resin. That is, the number of product defects in the next shot is predicted using the obtained weighing information and the prediction model predicted from the trial molding (step S13). The prediction result is compared with the target value (illustrated is “2”) to determine whether the molding conditions are necessary (step S14). At this time, if the number of product defects is “2”, the automatic operation is continued without changing the molding conditions. On the other hand, when the number of product defects exceeds or falls below “2”, optimization processing for the molding condition (input parameter Df...) Is executed.

  In this case, first, data relating to the predicted number of product defects is transmitted to the data processing unit E (computer system Em) (step S15). In the data processing unit E, the received data is input to the optimization model stored in the optimization processing program Ps to perform optimization processing. That is, first, the molding condition relating to the input parameter Df... Is obtained from the received data by the optimization model (prediction function F), and the molding condition satisfying the constraint condition Xc... And the objective function Xp. (Steps S16 and S17). By repeating this process for the corresponding molding conditions, the optimized molding conditions can be obtained (step S18). Then, the optimized molding condition is transmitted from the data processing unit E to the molding machine controller 41 (step S19). Thereby, the molding machine controller 41 performs a process of changing the preset molding conditions according to the received molding conditions (step S20). In this case, the execution timing of the change process is performed before the mold clamping process in the next shot is started.

  FIG. 11 shows experimental data for confirming the effectiveness (potentiality) of the molding optimization method according to the present embodiment. This experiment was conducted using the injection molding machine (experimental machine) M shown in FIG. 3 and shows the relationship between the number of product defects and the rotation speed of the material feeder 32. In this case, the molding optimization method according to the present embodiment is applied, that is, the processing procedure shown in FIGS. 1 and 2 is followed, and the material supply machine 32 is controlled to control the supply amount of the molding material. This is the result of a specially emphasized disturbance.

  As shown in the experimental results shown in FIG. 4, when the molding optimization method according to the present embodiment is not applied, the supply amount decreases if the supply machine speed is relatively low (50 [rpm]). When the number of product defects is “8” and the rotational speed of the feeder is relatively high (90 [rpm]), the supply amount is increased and the number of product defects is “0”.

  On the other hand, when the molding optimization method according to the present embodiment is applied, as shown in FIG. 11, the molding conditions are optimized by the molding optimization method in both the former stage and the latter stage. It can be confirmed that the number of product defects is approximately “1”, and the number of product defects in the subsequent stage is approximately “4”. That is, it shows that the filling amount of the mold 33 into the cavity can be controlled against the occurrence of disturbance. In FIG. 11, the rotation speed of the feeder is increased stepwise from 60 → 70 → 80 [rpm] in the former stage, and the condition is changed at the time point Kc. Thereafter, in the latter stage, (80) → In this example, the number of product defects is set to “1” in the former stage, and the number of product defects is set to “4” in the latter stage.

  On the other hand, FIG. 12 to FIG. 14 show actual fluctuation states of the optimized molding conditions in FIG. 11, FIG. 12 shows fluctuation characteristics obtained by actually measuring the VP switching position, and FIG. FIG. 14 shows variation characteristics obtained by actually measuring the pressure, and FIG. 14 shows variation characteristics obtained by actually measuring the injection speed. As is apparent from FIGS. 12 to 14, it can be confirmed that the optimization process is always performed in accordance with the disturbance state or the like in any case.

  Therefore, according to the molding optimization method of the injection molding machine according to the present embodiment, as a basic processing mode, in the data processing unit E, the molding data related to the input parameter Df... And the output parameter Ds. On the other hand, evaluation data related to molding data and molding quality at the molding process is set as a constraint condition Xc ... based on mathematical programming, and a prediction function F ... is set as an objective function Xp ... An optimization processing program Ps for obtaining an optimized molding condition related to the input parameter Df... Satisfying the constraint condition Xc... And the objective function Xp. The molding data relating to the output parameter Ds during the molding process is detected and the output parameter Ds. Since the molding conditions optimized by the optimization processing program Ps are obtained based on the molding data, and the existing molding conditions are changed according to the obtained molding conditions, the molding process is set as the constraint condition Xc. The molding data and the evaluation information (molding failure, etc.) related to the molding quality and the prediction function F ... set as the objective function Xp ... allow the molding conditions to be optimized easily and quickly, and also related to the molding quality. Improvements can be made effectively (effectively). Further, since the prediction function F... Is applied as the objective function Xp..., It is possible to model a certain physical phenomenon by learning a predetermined number of detection data obtained by trial molding and causing pattern recognition. Thus, the simulation of the molding result can be easily performed from the molding conditions based on the input parameter Df... Without actually operating the injection molding machine.

  Moreover, after setting (optimizing) the molding conditions, the external environment such as humidity and temperature varies depending on the production time and progress, etc., and the physical properties of the molding material are affected or the set values drift. Even so, since the optimization process can be speeded up, the optimization with respect to the molding conditions can be performed in units of one or more shots. As a result, it is possible to always carry out production in the best condition optimized, such as avoiding a situation where the molding quality temporarily decreases. Furthermore, since optimization for molding conditions is automatically performed in units of one or two or more shots, a separate setting step associated with resetting is not necessary. As a result, the productivity can be improved along with the reduction of man-hours and the cost can be reduced along with the improvement of the production efficiency, and it can be used as a molding optimization method with high practicality at an actual production site.

  As mentioned above, although preferred embodiment was described in detail, this invention is not limited to such embodiment, It does not deviate from the summary of this invention in a detailed structure, a shape, a raw material, quantity, a numerical value, etc. It can be changed, added, or deleted arbitrarily.

  For example, the molding data in the injection process is selected as the molding data in the molding process set as the constraint condition Xc... And the injection speed, speed control-pressure control switching position, and holding pressure are selected as the molding data in the injection process. Although the case has been shown, one or more of these may be selected, or other control factors of the injection process may be selected. Furthermore, as the molding data, molding data in the metering process or molding data in both the metering process and the injection process as in the embodiment may be selected. Moreover, although the case where the information regarding the molding defect of a molded product is applied as the evaluation information related to the molding quality is exemplified, the information related to the performance evaluation of the molded product may be applied. In addition, the case where the number of product defects in the multi-piece molded product in a short state is applied as the molding defect of the molded product is exemplified, but the present invention can be similarly applied or applied to other molding defects. On the other hand, the data processing unit E shows the case where the computer system Em separately prepared for the injection molding machine M is used, but it may be configured integrally with the molding machine controller 41. Moreover, although the trust region method which is one of the interior point methods is exemplified as the mathematical programming method, various other well-known mathematical programming methods such as linear programming and branch and bound methods can be used. Furthermore, in the embodiment, the case where the optimization is performed with respect to the molding conditions in units of one shot on the premise of a multi-piece molded product is shown, but the number of defects is determined in units of two or more shots on the premise of a single molded product. Thus, optimization for the molding conditions may be performed.

  The molding optimization method according to the present invention can be used for various injection molding machines having a function of optimizing molding conditions using a neural network.

  M: injection molding machine, E: data processing unit, Df ...: input parameters, Ds ...: output parameters, N: neural network, Xc ...: constraints, Xp ...: objective function, Ps: optimization processing program

Claims (6)

  A data processing unit with a computer function obtains a prediction function based on learning of the neural network from molding data related to input parameters including molding conditions of the injection molding machine and molding data related to output parameters obtained by trial molding based on these input parameters. An injection molding machine molding optimization method for optimizing a predetermined molding condition by using the prediction function, wherein a mathematical programming method is applied to the data processing unit in advance for molding data relating to the input parameter and the output parameter. The evaluation parameters related to molding data and molding quality at the molding process are set as constraints based on the above, and the prediction function is set as an objective function, and the input parameters satisfying the constraints and objective function by the mathematical programming method The optimized molding conditions related to In the production operation, the data processing unit detects molding data related to the output parameter during the molding process at the time of one or more shots, and molding related to the output parameter. A molding optimization method for an injection molding machine, characterized in that a molding condition optimized by the optimization processing program is obtained based on data, and a process for changing an existing molding condition according to the obtained molding condition is performed.   2. The molding optimization method for an injection molding machine according to claim 1, wherein molding data in the weighing process and / or the injection process is selected as molding data for the molding process set as the constraint condition.   3. The method of optimizing molding of an injection molding machine according to claim 2, wherein at least one of injection speed, speed control-pressure control switching position, and holding pressure is selected as molding data in the injection process. .   4. The optimum molding of an injection molding machine according to claim 1, 2 or 3, wherein the evaluation information related to the molding quality is selected from information related to molding defects of the molded product or information related to performance evaluation of the molded product. Method.   5. The method for optimizing molding of an injection molding machine according to claim 4, wherein a short shot defect and / or a burr defect is selected as the molding defect of the molded product.   2. The molding optimization method for an injection molding machine according to claim 1, wherein the objective function is a nonlinear function.

Images