JP2022107954A - Die management system - Google Patents

Abstract

To make it possible to determine whether cleaning of a die is necessary or not while suppressing degrading of a through-put in a molding process.SOLUTION: A die management system includes an imaging unit 40 and a level estimation unit 30. The imaging unit 40 images a surface of a molded product molded with the die. The level estimation unit 30 includes: an input layer to which a product image being an imaged image of the molded product is input; and an output layer which outputs, according to the inputted product image, a degree of contamination of the die used for molding the molded product being an origin of the product image.SELECTED DRAWING: Figure 3

Description

This specification discloses a mold management system that determines whether or not a mold for molding needs to be cleaned.

2. Description of the Related Art Conventionally, there has been known a management system that uses image recognition to determine the quality or abnormality of a product. For example, in Japanese Patent Laid-Open No. 2002-100000, quality determination is performed based on a captured image of a terminal crimping portion of an electric wire. A neural network using machine learning such as deep learning is used for this pass/fail judgment.

JP 2020-52044 A

By the way, there are cases where it is difficult to directly capture an image of an object to be managed. For example, when trying to image the machined surface of the mold when determining whether or not cleaning of the mold for molding is necessary, the machined surface is imaged, for example, when the male mold and female mold are separated. There is a need. However, even in such a case, this separation distance is sufficient if the molded product can be taken out, and may not be sufficient to image the entire processed surface. In addition, if the male mold and the female mold are separated to the extent that the entire molded product can be imaged, an imaging device is inserted between them, and then a process such as imaging the processed surface is provided, the throughput of the molding process can be improved. (Processing amount per unit time) is reduced.

Therefore, the present specification discloses a mold management system capable of determining whether or not cleaning of the mold is necessary while suppressing a decrease in the throughput of the molding process.

Disclosed herein is a mold management system. This system includes an imaging unit and an estimating unit. The imaging unit images the surface of the molded product molded by the mold. The estimating unit stores an input layer to which a product image, which is a photographed image of a molded product, is input, and the contamination degree of the mold used for molding the molded product that is the basis of the product image, according to the input product image. It has an output layer that outputs

According to the above configuration, it is possible to image the surface of the molded product to which the surface state of the mold is transferred at a desired angle, and it is possible to accurately determine whether or not the mold needs to be cleaned. Moreover, since it is only necessary to image the molded product flowing through the line after molding, the throughput of the molding process is prevented from being lowered.

Further, in the above configuration, the mold management system may include a cleaning work setting section. The cleaning work setting unit arranges the degree of contamination based on a plurality of product images in chronological order to obtain the progress of the degree of contamination, and based on the progress of the degree of contamination and the processes before and after the molding process in which the mold is used, Set the mold cleaning process.

According to the above configuration, it is possible to perform cleaning (advance cleaning) in the previous stage when it is determined that cleaning is necessary at a timing that does not cause a waiting time between the molding process and the previous process, and between the molding process and the subsequent process. becomes.

Further, in the above configuration, the cleaning work setting unit estimates, based on the progress of the degree of contamination, the remaining number of times, which is the number of times of molding until the degree of contamination reaches the threshold contamination degree at which it is determined that cleaning is necessary. The mold cleaning process may be set so as to at least partially overlap with the pause periods set in the process plans of the preceding and succeeding processes within the time required to reach the remaining number of times.

According to the above configuration, the mold cleaning can be performed at the same timing as the pause period of the pre-process or post-process of the molding process, and it is possible to suppress a decrease in throughput.

Further, in the above configuration, a display control unit may be provided that causes the display unit to display the most recent contamination degree output from the output layer of the estimation unit and the history of the contamination degree output from the output layer in parallel.

When the surface of the molded product becomes dirty, the cause can be found in the case where the dirt is transferred to the surface of the mold processed, and the work failure such as gas burning in the molding process. If an operation failure such as the latter occurs suddenly and a high degree of contamination is determined on the basis of this, there is a risk of an erroneous determination that cleaning is required even though the machined surface of the mold is clean. Therefore, by displaying the history of the degree of contamination as in the above configuration, it is possible to confirm whether the occurrence of contamination is sudden or continuous, and to estimate the cause of the contamination.

In the above configuration, the estimating section may include a neural network for image recognition that has been trained using teacher data in which the product image is input data and the degree of contamination is output data.

According to the above configuration, an image obtained in a normal molding process can be used as teacher data, and the trouble of separately preparing an image for learning can be saved.

Further, in the above configuration, the estimating unit may learn from teaching data for each of a plurality of molds having the same type of machined surface. In this case, the contamination degree is output for each of the plurality of molds from the output layer.

According to the above configuration, a plurality of molds of the same type can be distinguished from each other, and a mold requiring cleaning can be accurately extracted from the plurality of molds of the same type.

Further, in the above configuration, the output layer of the estimation unit may output the accuracy for each degree of contamination. In this case, the mold management system includes a reference data storage section and a similarity calculation section. The reference data storage unit stores a probability distribution indicating the probability distribution of each degree of contamination when the product image included in the training data is input to the input layer of the estimation unit. The similarity calculation unit obtains the similarity of the probability distribution stored in the reference data storage unit with respect to the probability distribution based on the most recent product image for a plurality of product images of the teacher data, and calculates the distribution of the obtained similarity. , is obtained for each degree of contamination, which is the output data of the teacher data.

Generally, in a neural network, among a plurality of classes (dirtyness here) in the output layer, the class with the highest accuracy is output as the correct answer. In the above configuration, the probability distribution of each class by the output layer is used. Furthermore, the similarity is obtained by comparing the probability distribution based on the most recent product image and the probability distribution when each product image of the training data is input to the neural network. It is obtained for each class (contamination degree) that is data (output data). For example, it is possible to presume that a class that frequently has a high degree of similarity is the class that corresponds to the most recent product image.

According to the mold management system disclosed in this specification, it is possible to determine whether or not the mold needs to be cleaned while suppressing a decrease in the throughput of the molding process.

It is a figure which illustrates the hardware constitutions of the metal mold|die management system which concerns on this embodiment. It is a figure which illustrates the functional block of the metal mold|die management system which concerns on this embodiment. FIG. 4 is a diagram illustrating a neural network configuration of a level estimator; FIG. 4 is a diagram illustrating teacher data for learning a level estimation unit; FIG. 10 is a diagram showing an example of image recognition capable of distinguishing products molded by a plurality of molds of the same type; FIG. 10 is a diagram showing another example of the neural network configuration of the level estimation unit; FIG. 11 is a diagram (1/2) illustrating a history graph of levels (dirtyness) by the level estimating unit; FIG. 11 is a diagram (2/2) exemplifying a history graph of levels (contamination degrees) by the level estimating unit; FIG. 10 is a diagram illustrating calculation of a similarity regarding the probability distribution of the output layer of the level estimator; FIG. 10 is a diagram showing an example of distribution of similarity of probability distributions for each level; It is a figure explaining the schedule setting of forward cleaning of a metal mold|die.

<Overall composition>
FIG. 1 illustrates a mold management system according to this embodiment. As described below, the system outputs the contamination level of the insert mold. A photographed image of a molded product, which is a product molded by the mold, is used to output the contamination level. In other words, in the mold management system according to the present embodiment, the surface of the molded product to which the dirt on the processed surface of the mold is transferred is monitored, and the need for cleaning the mold is determined according to the degree of dirt on the surface of the molded product. be. This mold management system includes a mold management device 10 and an imaging device 40 (imaging unit).

The imaging device 40 (imaging unit) is, for example, a camera device installed at an insert molding work site, and is provided on a transfer line between insert molding and subsequent processes. The conveying line is provided with positioning pins and depressions (not shown), so that a plurality of molded products flowing along the conveying line can be imaged at the same angle. The imaging device 40 includes, for example, an imaging device such as CMOS or CCD, and images the surface of a molded product manufactured by an insert molding apparatus, in other words, molded by a mold.

The mold management apparatus 10 is composed of computer equipment, for example. The mold management apparatus 10 includes a CPU 11 as an arithmetic unit, and a system memory 12 and a storage 13 as storage devices. The storage 13 may be a non-transitory storage device such as a hard disk drive (HDD) or solid state drive (SSD). The mold management apparatus 10 also includes an input unit 14 which is an input device such as a keyboard and a mouse, and an input/output controller 15 which manages input/output of information to/from an external device such as an imaging device 40 .

Furthermore, the mold management apparatus 10 includes a GPU 16 (Graphics Processing Unit), a frame memory 17, a RAMDAC 18 (Random Access Memory Digital-to-Analog Converter), and a display unit as image processing units for processing the captured image captured by the imaging device 40. A control unit 19 is provided. In addition, the mold management apparatus 10 includes a display section 20 that displays processed images. The mold management apparatus 10 may include a touch panel display in which the input section 14 and the display section 20 are integrated.

The GPU 16 is an arithmetic unit for image processing, and is mainly operated when determining the degree of contamination of a mold based on a molded product, which will be described later. The frame memory 17 is a storage device that stores images captured by the imaging device 40 and processed by the GPU 16 . The RAMDAC 18 converts the image data stored in the frame memory 17 into analog signals for the display unit 20, which is an analog display.

The display control unit 19 causes the display unit 20 to display images processed through the GPU 16 , the frame memory 17 and the RAMDAC 18 . For example, the display control unit 19 can superimpose an image captured by the imaging device 40 on the display unit 20, such as a highlighted display (heat map) that emphasizes the location that is the basis for determining the degree of contamination.

As will be described later, when the level estimating unit 30 (see FIG. 2) outputs the degree of contamination (level) that requires cleaning of the mold for the input captured image, it is received. Then, the display control unit 19 causes the display unit 20 to output a warning display. Further, the process plan including forward cleaning calculated by the cleaning work setting unit 33 is displayed on the display unit 20 by the display control unit 19 .

On the other hand, when the level estimating unit 30 determines that the level of contamination (level) does not require cleaning of the mold, the display control unit 19 receives this and causes the display unit 20 to output an acceptance display.

In FIG. 1, the mold management apparatus 10 includes the display unit 20. For example, the mold management apparatus 10 (computer) including components other than the input unit 14 and the display unit 20 is separated from the assembly site. It may be installed in a separate server room. Also, the display unit 20 may be installed at the work site so as to be viewable by an insert molding operator.

Also, a touch panel display in which the display unit 20 and the input unit 14 are integrated may be installed at the work site. As will be described later, when the level estimator 30 outputs a level requiring cleaning, the molding operation is temporarily stopped. A confirmation button may be displayed on the touch panel display so that the molding operation can be resumed.

Once the molding work is stopped, the worker who performs the molding work in the molding machine visually checks the processed surface of the molded product and the mold, and if there is no problem, that is, there is no problem with the contamination of the processed surface of the mold. When it is determined that the level is reached, the confirmation button is pushed down by the worker.

FIG. 2 exemplifies the functional blocks of the mold management apparatus 10 in a form mixed with the hardware blocks shown in FIG. This functional block diagram is configured by the CPU 11 executing a program stored in, for example, the storage 13 or a non-transitory computer-readable storage medium such as a DVD.

The mold management apparatus 10 includes a level estimation unit 30, a cumulative value calculation unit 31, a similarity calculation unit 32, a cleaning work setting unit 33, and a reference data storage unit 35 as functional blocks. Note that the cumulative value calculation unit 31, the similarity calculation unit 32, the cleaning work setting unit 33, and the reference data storage unit 35 provide an additional provided in

<Level estimation unit>
The level estimator 30 comprises, for example, a neural network for image recognition comprising an input layer, a hidden layer and an output layer. Product image data, which is captured image data of a molded product, is input to the input layer. For example, the number of nodes in the input layer may be equal to the number of pixels in the captured image.

A plurality of nodes are provided in the output layer. For each node, the probability of the class representing the dirtiness (level) is output. The accuracy is output, for example, as a percentage. Further, the class (level) having the highest accuracy among the accuracy of each node is output as the contamination degree corresponding to the input captured image.

For example, as exemplified in FIG. 3, the degree of contamination is divided into a plurality of stages according to the degree of contamination on the surface of the molded product. This degree of contamination indicates the degree of contamination of the mold used for molding the molded product that is the basis of the product image. For example, in FIG. 3, the degree of contamination is divided into nine levels from level 1 to level 9. Here, it is evaluated that the degree of contamination progresses from level 1 to level 9.

A hidden layer is provided between the input layer and the output layer. Since the captured image is included in the input layer, the level estimation unit 30 includes, for example, a convolutional neural network (CNN) for image recognition, and the hidden layer includes a convolution layer and a pooling layer.

FIG. 4 illustrates training teacher data for the level estimation unit 30 having a neural network. In this example, captured image data of 4500 molded products are used as input data. The molded product molded with the mold immediately after cleaning was designated as No. 1, No. Up to 4500 the surface of the molded product is imaged. Dirt generated on the working surface of the mold is transferred to the surface of the molded product, and No. 1 to No. In the process of reaching 4500, the area of contamination on the surface increases. For example, No. 4001 to No. The surfaces of molded products up to 4500 are highly likely to be judged as defective in the appearance inspection of the product, and cleaning of the working surface of the mold is required.

As for such a product image, for example, when the product is moved to a post-process after molding, the surface of the molded product carried on the conveying line is imaged by the imaging device 40 (imaging unit). These captured images are images obtained in a normal insert molding process, and are obtained without intentionally changing processing conditions and the like for learning of the level estimating section 30 .

No. 1 to No. Captured images of up to 4500 molded products, that is, product images, are divided into 500-sheet contamination levels. That is, the product image No. which is the input data. 1-No. 500, the contamination degree level 1 is given as output data, that is, correct data. Similarly, product image No. 501-No. For each of the 1000 product image no. 1001-No. For each of the 1500, a level of 3 is given. Also, the product image No. 1501-No. 2000 were given a level of 4 and product image no. 2001-No. For each of the 2500, a level of 5 is given. Furthermore, the product image No. 2501-No. 3000 are given a level of 6 and product image no. 3001-No. For each of the 3500, level 7 is given. Finally, product image No. 3501-No. 4000 are given a level of 8 and product image no. 4001-No. For each of the 4500, level 9 is given.

A trained neural network trained using such teacher data is implemented in the level estimation unit 30 . When a captured image (product image) of a molded product actually molded by insert molding is input to the input layer of the level estimating section 30, the accuracy of each contamination level is output from each node of the output layer. . Furthermore, the level with the highest accuracy among them is output as an answer. For example, in the example of FIG. 3, level 2, which has the highest accuracy with respect to the input product image, is output as the final value.

Hereinafter, the manner in which the accuracy for each level is output to each node of the output layer will be referred to as "accuracy output" as appropriate. Also, the manner in which the highest level of accuracy among the accuracy of each level is output as an answer is described as "answer output".

The level values output as answers are sent to the cumulative value calculator 31 (see FIG. 2), the similarity calculator 32 , the cleaning work setting unit 33 , and the display controller 19 . The former three will be discussed later. The display control unit 19 causes the display unit 20 to display the received contamination level. At this time, the display control unit 19 may change the display mode on the display unit 20 according to the level of dirtiness. For example, levels 1 to 6 are displayed on the display unit 20 with a green background as normal display. At levels 7 and 8, the level is displayed on the display unit 20 with a yellow background as a caution display. Furthermore, at level 9, the level is displayed on the display unit 20 with a red background as a warning display. As a result, it is possible for the operator during the molding operation to determine whether or not the mold needs to be cleaned.

<Distinction of molds of the same type>
FIG. 3 illustrates the level estimation unit 30 for estimating the degree of contamination of a single mold based on the surface of the product molded by the mold. is not limited to For example, when there are a plurality of molds having at least the same type of machining surface, the degree of contamination of the mold may be estimated after estimating (identifying) which mold it is.

In this case, 4500 pieces of product image data shown in FIG. 4 are provided for each mold as training data. For example, if there are at least 8 molds with the same type of machined surface, 4500 sheets of product image data are prepared for 8 types, and the contamination level is set as a class for each of them.

FIG. 5 illustrates how the level estimating unit 30 discriminates molds of the same type. FIG. 5 shows the mold No. of the same type. 5 to No. An image in which a heat map is superimposed on a captured image (product image) of a molded product molded by 7 is exemplified.

In this heat map, the parts of interest used as the basis for discriminating each mold are highlighted. The parts of interest are different in each product image, indicating that the molds of the same type can be accurately discriminated. For example, each mold has different surface characteristics such as the casting surface on the machined surface, and this different surface shape is transferred to the molded product. be done.

FIG. 6 exemplifies such a neural network that outputs the degree of contamination level for each of a plurality of molds of the same type. In this neural network, the number of nodes in the output layer is the number of molds of the same type times the number of levels. For example, if there are eight molds of the same type and the levels are divided into nine stages, the number of nodes in the output layer is 72.

In this mode, the contamination level and the mold number are output from the level estimation section 30 to the display control section 19 . In response to this, the display control unit 19 causes the display unit 20 to display the mold number in addition to the contamination level.

<Use level history>
Contamination on the surface of a molded product may be caused by the transfer of contamination on the surface of the mold to the molded product, or by so-called gas burning due to evaporation of resin during molding. Since the latter does not require cleaning of the mold, it is preferable to avoid displaying a warning on the display unit 20 when the level of contamination is estimated to be high due to the latter cause.

Therefore, historical information on the degree of contamination may be used to estimate the cause of the contamination of the molded product. The cumulative value calculation unit 31 stores the history of the contamination levels output from the level estimation unit 30 in chronological order. Then, the cumulative value calculator 31 transmits to the display controller 19, for example, the most recent contamination level value and past level values of a plurality of points (for example, 99 points) retroactively therefrom.

The display control unit 19 creates a histogram as illustrated in FIG. 7 from the contamination degree level values received from the cumulative value calculation unit 31 . In this histogram, the horizontal axis indicates the level of the degree of contamination, and the vertical axis indicates the number of data points (the number of product images) having the corresponding level value.

For example, in FIG. 7, one point of data of contamination level 5 is plotted, and the rest are distributed to levels 1 and 2. FIG. From this, it can be inferred that the cause of the contamination of the molded product determined to be contamination level 5 is sudden and not due to the continuous contamination of the mold processing surface. For example, the display control unit 19 causes the display unit 20 to display a histogram based on the level value history of FIG.

Here, when the dirt on the molded product is caused by the dirt on the working surface of the mold, the dirt is constantly transferred to the molded product. Therefore, as shown in FIG. 8, the histogram based on the level values traces a transition such that the pillars of the histogram grow and move toward higher level values. The contamination level may be set based on such histogram changes. For example, the level at which the frequency exceeds a threshold (eg, 50) may be set as the contamination level corresponding to the most recent product image.

If the contamination level output by the level estimating unit 30 is different from the contamination level based on the histogram of the level values, the contamination of the molded product used as the criterion for determining the former contamination level is the same as that of the mold. It may not be caused by dirt. In such a case, the display control unit 19 causes the display unit 20 to display in parallel the most recent contamination level output by the level estimating unit 30 and the contamination level based on the histograms of FIGS. , to display a warning such as blinking the background to prompt the operator to confirm.

<Similarity calculation>
In the above embodiment, the contamination level of the mold is estimated using the response output by the level estimation unit 30. However, using the probability distribution output to each node of the output layer of the level estimation unit 30, A contamination level may be estimated.

Referring to FIG. 2, teacher data of level estimation unit 30 is stored in reference data storage unit 35 . The teacher data is stored in pairs of product images as input data and levels as output data. When a product image of teacher data is input to the level estimation unit 30, the probability is output to each node of the output layer. Among these, the level stored as the output data in the teacher data obtains the highest accuracy.

The accuracy distribution of the contamination level output from each node of the output layer when the product image included in the training data is input to the input layer of the level estimation unit, that is, the accuracy distribution is stored in the reference data storage unit 35. stored in The similarity calculation unit 32 extracts the probability distribution data for all the teacher data from the reference data storage unit 35, for example. Further, the similarity calculation unit 32 is transmitted from the level estimation unit 30 with the probability distribution data output with respect to the most recent product image.

The similarity calculator 32 obtains the similarity between the probability distribution of the most recent product image and the probability distribution of the teacher data. For example, cosine similarity is used as similarity. For example, referring to FIG. 9, the probability distribution of the most recent product image is (2.5, 99.8, . is replaced by a vector with the components of Similarly, the probability distribution of the teacher data is also replaced by a vector having components of the number of levels, such as (2.3, 99.9, . . . , 1.6, 0.5).

Here, the cosine similarity between two vectors a and b having n components is obtained by Equation 1 below.

When the value obtained by Equation 1 is close to 1 (Cos=0°, parallel), the similarity is high, and the closer it is to -1, the lower the similarity. The similarity calculation unit 32 obtains such similarity for the probability distribution for the most recent product image and all probability distributions in the teacher data.

Further, the similarity calculation unit 32 summarizes the obtained similarity distribution (similarity distribution) for each contamination level. This contamination degree level corresponds to the correct data (class) which is the output data of the teacher data. FIG. 10 exemplifies a histogram of similarity distribution for each contamination level. In this histogram, the horizontal axis indicates the degree of similarity, and the vertical axis indicates the number of images (frequency). In this example, the degree of similarity is obtained to the first decimal place so that the horizontal axis is a discrete value.

For each similarity, the frequency of each level is plotted. For example, at a high degree of similarity (0.8 to 0.9), the frequency of training data with dirtiness level 1 is high. In this case, the most recent product image and the training data with the contamination degree level 1 acquire a high degree of similarity with high frequency. From this point of view, for example, the contamination level included in a region (selected region) with a high similarity (for example, 0.8 or higher) and a high frequency (for example, 50 or higher), such as the hatched region in FIG. is chosen as the contamination level given to the most recent product image.

The similarity calculation unit 32 creates a similarity distribution histogram as shown in FIG. Then, the contamination level is transmitted to the display control unit 19 . The display control unit 19 causes the display unit 20 to display the contamination level based on the response output from the level estimation unit 30, the similarity histogram from the similarity calculation unit 32, and the contamination level included in the selected region.

When the contamination level output from the level estimation unit 30 and the contamination level output from the similarity calculation unit 32 match, the level value is determined as the contamination level for the most recent product image. be.

On the other hand, if the contamination level output from the level estimation unit 30 and the contamination level output from the similarity calculation unit 32 are different, there is a possibility that sudden contamination has occurred due to, for example, gas burning. be. In this case, in order to prompt the operator to check the molded product, for example, the display control unit 19 compares the contamination level output from the level estimation unit 30 and the contamination level output from the similarity calculation unit 32. Along with parallel display on the display unit 20, warning display such as blinking of the background image is performed.

As described above, in addition to the contamination level response output by the level estimation unit 30, the history information of the level values by the cumulative value calculation unit 31 is arbitrarily used, and the similarity distribution by the similarity calculation unit 32 is used. By utilizing , it is possible to estimate the degree of contamination with higher accuracy.

<Set forward cleaning plan>
For example, when the contamination level required for mold cleaning is set to level 9, before reaching level 9, the cleaning process is carried out ahead of schedule based on the process plan for the processes before and after the insert molding process. can suppress a decrease in the throughput of the entire process.

For example, as shown in FIGS. 7 and 8 described above, based on the history of the contamination level, it is possible to predict the future transition of the contamination level, that is, the progress of the contamination level. For example, the contamination level based on a plurality of product images is arranged in chronological order to obtain the progress of the contamination level. Specifically, the cumulative number of molded products from a certain contamination level to the next level can be obtained from the contamination level history information. For example, as described above, in the teaching data of FIG. 4, the contamination level was set for every 500 molded products. It can be estimated that

Further, as the progress of the contamination level, for example, based on the number of molded products at each level from contamination level 1 to level 4, the number of molded products at each level from level 5 to level 9 can be estimated. . For example, if the contamination degree level 9 is the threshold contamination degree that requires mold cleaning, based on the contamination degree level for the most recent product image and the transition estimation of the contamination degree level, until the contamination degree threshold value (level 9) is reached, It is possible to obtain the remaining number of times, which is the number of times of molding (number of shots).

The cleaning work setting unit 33 (see FIG. 2) obtains the remaining number of times of such molding processes, and refers to the process plan of the processes before and after the insert molding process using the mold to be cleaned, thereby determining the mold. Plan for forward cleaning.

FIG. 11 exemplifies the insert molding process and process plans for the pre-process and post-process. In this figure, the number of molded products is counted as "Lot". For example, Lot 1 to Lot 500 indicate the 1st to 500th molded products using cleaned molds.

Also, in this example, the contamination level transition estimation is performed at the time when the contamination level is switched from level 3 to level 4. FIG. This estimation is performed using linear approximation, for example. By transition estimation, the remaining number of times from the time of estimation until level 9, which is the threshold dirtiness, is reached is obtained.

The cleaning work setting unit 33 confirms whether or not a maintenance period (rest period) is planned for the previous process or the subsequent process during the remaining number of times period, which is the time from the most recent molding process until the remaining number of times is reached. . For example, in FIG. 11, maintenance is planned for the post-process during the remaining number of times period. Therefore, the cleaning work setting unit 33 sets the cleaning work setting unit 33 so that at least a part of the maintenance period of the post-process is overlapped with the maintenance of the post-process when the contamination level reaches level 9, which is presumed to be level 7. In this way, the process plan of the insert molding process is changed so as to carry out front-loaded cleaning for cleaning the mold. The changed process plan is displayed on the display unit 20 via the display control unit 19 . Further, the data to be displayed on the display unit 20 may include contamination degree transition estimation data illustrated in FIG. 11 .

A worker who is performing an insert molding operation and views the display unit 20 cleans the mold based on the process plan including the forward cleaning process set by the cleaning operation setting unit 33 . Also, the process plan set by the cleaning work setting unit 33 may be edited by the operator.

As described above, in the mold management system according to the present embodiment, it is possible to image the surface of the molded product to which the surface condition of the mold is transferred at a desired angle, and to accurately determine whether the mold needs to be cleaned. It becomes possible. Moreover, since it is only necessary to image the molded product flowing through the line after molding, the throughput of the molding process is prevented from being lowered.

In addition to contamination level estimation by the level estimating unit 30, the contamination level history data by the cumulative value calculating unit 31 and the similarity distribution by the similarity calculating unit 32 are used to estimate the contamination level. Estimation can be performed with high accuracy. Furthermore, by executing forward cleaning by the cleaning work setting unit 33, mold cleaning can be performed in parallel with at least one of the pre-process and post-process maintenance process, and a series of processes from the pre-process to the insert molding process and the post-process are possible. It is possible to suppress the decrease in throughput in

10 mold management device, 19 display control unit, 20 display unit, 30 level estimation unit (estimation unit), 31 cumulative value calculation unit, 32 similarity calculation unit, 33 cleaning work setting unit, 35 reference data storage unit, 40 imaging device (imaging unit).

Claims (7)

an imaging unit that captures an image of the surface of the molded product molded by the mold;
An input layer to which a product image, which is a photographed image of a molded product, is input, and according to the input product image, the degree of contamination of the mold used for molding the molded product that is the basis of the product image is output. an estimator comprising an output layer that
A mold management system. The mold management system according to claim 1,
The degree of contamination based on the plurality of product images is arranged in chronological order to determine the progress of the degree of contamination, and based on the progress of the degree of contamination and the processes before and after the molding process in which the mold is used, the metal mold is obtained. A mold management system comprising a cleaning work setting unit for setting a mold cleaning process. The mold management system according to claim 2,
The cleaning work setting unit estimates, based on the progress of the contamination degree, the remaining number of times, which is the number of times of molding until the contamination degree reaches a threshold contamination degree at which it is determined that cleaning is necessary, and the estimated remaining number of times. setting the cleaning process of the mold so as to at least partially overlap with the rest period set in the process plan of the preceding and following processes within the time required to reach the remaining number of times;
Mold management system. The mold management system according to any one of claims 1 to 3,
A mold management system comprising a display control unit that causes a display unit to display in parallel the most recent contamination degree output from the output layer of the estimation unit and the history of the contamination degree output from the output layer. . The mold management system according to any one of claims 1 to 4,
The estimating unit includes a neural network for image recognition trained by teacher data having the product image as input data and the contamination level as output data,
Mold management system. The mold management system according to claim 5,
The estimating unit is a mold management system, wherein the estimating unit learns from the training data for each of the plurality of molds having the same type of machined surface, and outputs the contamination degree for each of the plurality of molds from the output layer. The mold management system according to claim 5 or 6,
The output layer of the estimating unit outputs an accuracy for each of the contamination degrees,
a reference data storage unit storing a probability distribution indicating the probability distribution of each of the contamination degrees when the product image included in the training data is input to the input layer of the estimation unit;
The similarity of the probability distribution stored in the reference data storage unit to the probability distribution based on the most recent product image is obtained for the plurality of product images of the teacher data, and the distribution of the obtained similarity is obtained for each of the contamination levels that are the output data of the teacher data,
Mold management system.

Images