JP2023157765A - Machine learning device, machine learning method, and mold maintenance time prediction device - Google Patents

Abstract

To provide a machine learning device and a machine learning method used for maintenance time prediction capable of determining appropriate mold maintenance time, and a mold maintenance time prediction device.SOLUTION: A machine learning device 1 includes: a data acquisition unit 4 for acquiring variations data 3 of an inspection actual measurement value for each cumulative actual number of molds; a maintenance record acquisition unit 6 for acquiring a maintenance record 5 of the mold corresponding to the variations data 3; and a learning unit 7 for constructing a learning model for predicting maintenance time of the mold by performing supervised learning using a set of the variations data 3 acquired by the data acquisition unit 4 and the maintenance record 5 acquired by the maintenance record acquisition unit 6 as teacher data.SELECTED DRAWING: Figure 1

Description

The present invention relates to a machine learning device, a machine learning method, and a mold maintenance time prediction device for use in predicting maintenance time for a mold.

Conventionally, weights related to the degree of mold wear are set according to the material, plate thickness, and processing type of the workpiece material, and the weight is taken into account in the number of mold hits to use a mold that is proportional to the mold wear. It has been proposed to calculate the number of times the mold has been used and to know when it is time to polish the mold when the number of times the mold has been used approaches a maximum usable number predetermined as the time to polish (for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 61-74739

However, in Patent Document 1, since the maximum usable number is uniquely determined, depending on the setting, there is a risk that it may be determined that it is time for maintenance excessively or that the appropriate time for maintenance may be overdue.

The present invention has been made in view of the above circumstances, and provides a machine learning device, a machine learning method, and a mold maintenance time prediction device for use in predicting maintenance time that can predict the appropriate maintenance time for a mold. provide.

In order to solve such problems, the present invention includes a data acquisition unit that acquires variation data of actual inspection values for each cumulative number of molds, and a maintenance record acquisition unit that acquires maintenance records of the mold corresponding to the variation data. and a set of the variation data acquired by the data acquisition unit and the maintenance record acquired by the maintenance record acquisition unit as training data to perform supervised learning to predict the maintenance timing of the mold. A machine learning device is provided, comprising a learning unit that constructs a learning model.

In the machine learning device, the variation data indicates the extent to which the trend of actual inspection values for each cumulative number of molds deviates from a control reference value for processed products to an upper or lower limit value. The maintenance record may indicate a maintenance time when maintenance was actually performed for the deviation of the trend.

In the machine learning device, the variation data may be calculated using actual measurement values of dimensions of the processed product input by the user and the number of actual results using the mold.

The present invention also provides a data acquisition step of acquiring variation data of actual inspection values for each cumulative number of molds, a maintenance record acquisition step of acquiring a maintenance record of the mold corresponding to the variation data, and a maintenance record acquisition step of acquiring the maintenance record of the mold corresponding to the variation data. A learning model for predicting maintenance timing of the mold is constructed by performing supervised learning using a set of the variation data acquired in the acquisition step and the maintenance record acquired in the maintenance record acquisition step as teacher data. A machine learning method is provided, including a learning step.

In the machine learning method, the variation data indicates the extent to which the trend of actual inspection values for each cumulative number of molds deviates from a control reference value for processed products to an upper or lower limit value. The maintenance record may indicate a maintenance time when maintenance was actually performed for the deviation of the trend.

In the machine learning method, the variation data may be calculated using actual measurement values of dimensions of the processed product input by the user and the number of actual results using the mold.

Further, the present invention includes a mold identification information acquisition unit that acquires mold identification information of a mold, and a cumulative performance number acquisition unit that acquires the cumulative performance number of the mold, and the cumulative performance number acquisition unit acquires the cumulative performance number acquisition unit. A learning model constructed by the machine learning device is applied to the cumulative number of results obtained, and a maintenance time for the mold corresponding to the mold identification information acquired by the mold identification information acquisition unit is predicted. , provides a mold maintenance timing prediction device.

According to the machine learning device, machine learning method, and mold maintenance timing prediction device of the present invention, it is possible to grasp signs of mold failure, prevent long-term production stoppages, and reduce mold repair costs. .

Additionally, by making maintenance timing visible, it becomes possible to pass on the experience of experienced engineers to less experienced engineers.

Furthermore, appropriate mold management through such visualization will also contribute to achieving the SDGs (``8. Decent work and economic growth'', ``9. Build a foundation for industry and technological innovation,'' etc.).

1 is a block diagram showing a schematic configuration of a machine learning device and a mold maintenance timing prediction device according to a preferred embodiment of the present invention. 2 is a flowchart showing the flow of learning model construction according to a preferred embodiment of the present invention. FIG. 2 is an explanatory diagram showing variation data in a preferred embodiment of the present invention. FIG. 7 is an explanatory diagram illustrating an example of how the trend of variation data fluctuates in a preferred embodiment of the present invention. FIG. 7 is an explanatory diagram illustrating another example of the degree of deviation in the trend of variation data in a preferred embodiment of the present invention. 3 is a flowchart showing the operation of the mold maintenance time prediction device according to the preferred embodiment of the present invention.

Hereinafter, preferred embodiments of a machine learning device, a machine learning method, and a mold maintenance time prediction device of the present invention will be described with reference to the drawings. However, the present invention is not limited to the following description, and those skilled in the art may make various modifications based on the gist of the invention described in the claims or disclosed in the detailed description. or change is possible. Such modifications and changes are also included within the scope of the present invention.

FIG. 1 is a block diagram showing a schematic configuration of a machine learning device 1 and a mold maintenance time prediction device 2 according to a preferred embodiment of the present invention. The machine learning device 1 includes a data acquisition unit 4 that acquires variation data 3 of actual inspection values for each cumulative number of molds (not shown), and a maintenance record 5 of the mold corresponding to the variation data 3. The maintenance time of the mold is predicted by performing supervised learning using the maintenance record acquisition unit 6, the set of the variation data 3 acquired by the data acquisition unit 4, and the maintenance record 5 acquired by the maintenance record acquisition unit 6 as teacher data. and a learning section 7 that constructs a learning model for the purpose of learning.

The learning model constructed by the learning section 7 is stored in the learning model storage section 8. When new training data is obtained after building a learning model, supervised learning is added to the learning model stored in the learning model storage unit 8, and the learning model that has been built is updated to create a new one. The learning model is stored in the learning model storage unit 8.

In daily inspections, the user measures, for example, the item code, item name, applicable location (inspection location), standard value of managed dimensions, lower limit value of managed dimensions, upper limit value of managed dimensions, and dimensions of processed products. The actual measured values, the average value of the actual measured values, the mold code, the mold name, the number of actual results using the mold, etc. are sent to an input management system such as a master system that is communicably connected via a cloud server. Enter and manage. Note that the same mold may be used for different items.

The mold maintenance timing prediction device 2 includes a mold identification information acquisition unit 9 that acquires mold identification information of the mold, and a cumulative performance number acquisition unit 10 that acquires the cumulative performance number of the mold. The learning model constructed by the machine learning device 1 is applied to the cumulative performance number acquired by the number acquisition unit 10, and the maintenance period of the mold corresponding to the mold identification information acquired by the mold identification information acquisition unit 9 is determined. Predict.

The mold maintenance timing prediction device 2 also includes a storage unit 11 that stores prediction results, and a display unit 12 that displays the prediction results.

FIG. 2 is a flowchart showing the flow of learning model construction according to a preferred embodiment of the present invention.

First, in step S11, the data acquisition unit 4 acquires the variation data 3. The data acquisition unit 4 outputs the variation data 3 to the learning unit 7. The variation data 3 indicates the degree to which the trend of the actual inspection value for each cumulative number of results deviates from the reference value of the dimension managed by the processed product to either the upper or lower limit value.

FIG. 3 shows reference values and upper and lower limits of dimensions to be managed in a histogram and probability distribution curve of actual inspection values for each cumulative number of results. The actual inspection values for each cumulative number of results are such that, in addition to cases where the peak of variation approaches the standard value as shown in Figure 3, there are also cases where the peak of variation tends to move toward the lower limit value as shown in Figure 4, and cases where the peak of variation approaches the reference value as shown in Figure 4. As shown in Fig. 5, there are cases where the peak of the dispersion tends to deviate to the upper limit value.

The degree of deviation in the trend can be expressed numerically, for example, by calculating the difference between the average value of the measured values or the value at the maximum point of the probability distribution curve and the reference value or upper and lower limit values of the dimension to be managed. Based on the sign of the difference, it can be determined whether the tendency of the actual inspection value for each cumulative number of results of the variation data 3 deviates from the reference value of the managed dimension to the upper or lower limit value. Depending on the magnitude of the difference, it is possible to determine how much the tendency of the actual inspection value for each cumulative number of results of the variation data 3 deviates from the reference value of the managed dimension to the upper and lower limit values.

Each time a user inspects a processed product, the user inputs the actual measured values of the dimensions of the processed product and the actual number of times the mold has been used from the previous inspection to this inspection into the master system, and the above difference and cumulative value are input into the master system. The number of actual results is calculated and recorded, and the variation data 3 is calculated as the degree of deviation in the trend of the actual inspection value for each cumulative number of results.

Next, in step S12, the maintenance record acquisition unit 6 acquires the maintenance record 5 of the mold corresponding to the variation data 3. The maintenance record acquisition unit 6 outputs the maintenance record 5 to the learning unit 7 in association with the variation data 3. In addition, the two steps of step S11 and step S12 may be performed in parallel, or step S12 may be performed first.

The maintenance record 5 indicates, for example, the maintenance timing when maintenance was actually performed for the degree of deviation in the trend of the above-mentioned variation data 3. The maintenance period has a correlation with the degree of deviation in the trend of variation data 3, and by using this correlation as training data and performing supervised learning to construct a learning model, it is possible to predict the appropriate maintenance period for the mold. becomes.

To explain more specific correlations, for example, as mentioned above, when the peak of dispersion tends to shift toward the lower limit due to an increase in the number of cumulative results, or when the peak of dispersion shows a tendency to shift toward the upper limit, There is a correlation between this and the need for more maintenance. By collecting a large amount of data on the degree of trend deviation and data on the corresponding maintenance timing, it becomes possible to understand, for example, the degree of deviation that requires a large number of maintenance operations.

In order to apply such correlations to supervised machine learning, for example, as mentioned above, the average value of the actual measured values or the value of the maximum point of the probability distribution curve, and the reference value and upper and lower limit values of the dimensions to be managed. The deviation of the trend is expressed numerically by the difference between the two values, and the relationship between the numerical value and the maintenance period is used as the training data.

The maintenance record 5 is input into the master system by the user each time maintenance is performed, and is recorded in association with other data and variation data 3. Note that the timing of maintenance used for the teacher data is preferably determined by the user based on experience.

Next, in step S13, the learning unit 7 generates teacher data by combining the variation data 3 and each data of the maintenance record 5 corresponding thereto.

Next, in step S14, the learning unit 7 performs machine learning based on the teacher data generated in step S13. This machine learning is supervised learning.

Next, in step S15, the learning unit 7 determines whether to end machine learning. The condition for determining completion may be, for example, that supervised learning has been performed a predetermined number of times.

If the conditions for ending machine learning are not met, the process returns to step S11 and machine learning is repeated. If the conditions for terminating machine learning are met, the process proceeds to step S16, and the constructed learning model is stored in the learning model storage unit 8 for preservation.

FIG. 6 is a flowchart showing the operation of the mold maintenance timing prediction device 2 according to the preferred embodiment of the present invention.

First, in step S21, the mold maintenance time prediction device 2 uses the mold identification information acquisition unit 9 to acquire mold identification information of a mold that predicts the maintenance time from input by the user, for example. As the mold identification information, for example, a mold code, a mold name, etc. can be used.

Next, in step S22, the mold maintenance time prediction device 2 uses the cumulative performance number acquisition unit 10 to acquire the cumulative performance number of the mold corresponding to the mold identification information acquired in step S21. The cumulative number of results of molds can be obtained, for example, based on the number of results using molds input into the master system described above. Alternatively, the user may input the cumulative number of molds. The two steps of step S21 and step S22 may be performed in parallel, with step S22 performed first or at the same time.

Next, in step S23, the mold maintenance timing prediction device 2 inputs the mold identification information and the cumulative number of molds, and calculates the mold maintenance timing based on the learning model constructed by the above method. Predict.

Finally, in step S24, the mold maintenance time prediction device 2 stores the prediction result in step S23 in the storage unit 11.

Supervised learning can use analysis techniques such as regression analysis (least squares method) and decision trees, and may also be ensemble learning that combines a plurality of these analysis techniques. Alternatively, the machine learning device 1 may be able to detect the feature amount by itself using a neural network such as deep learning provided with an intermediate layer, and can produce a complex output that more closely matches the teacher data.

The machine learning device 1 can be realized, for example, by installing an application software program on a personal computer or a server device. That is, the machine learning device 1 includes at least an arithmetic processing device such as a CPU (Central Processing Unit), a storage device that stores an application software program, and the like. Since supervised learning requires a large amount of calculation, the machine learning device 1 may further include, for example, a GPU (Graphics Processing Unit).

The mold maintenance time prediction device 2 can be realized, for example, by installing an application software program on a personal computer or a server device. That is, the mold maintenance time prediction device 2 includes at least an arithmetic processing device such as a CPU (Central Processing Unit), a storage device that stores an application software program, and the like. The mold maintenance time prediction device 2 predicts the maintenance time of the mold by calculating the mold identification information, the cumulative number of molds, and the learning model constructed by the machine learning device 1 using an application software program as input. The prediction is made, and the prediction result is output and displayed on the display unit 12, such as a display. Users can predict the timing of mold maintenance by inputting the actual measurements of the dimensions of processed products and the number of times the mold has been used each time an inspection is performed. This can prevent long-term production stoppages due to damage and reduce repair costs through regular maintenance. By preventing long-term production stoppages, it is possible to reduce lost sales opportunities and reduce the time it takes to revise production plans.

1 Machine learning device 2 Mold maintenance timing prediction device 3 Variation data 4 Data acquisition section 5 Maintenance record 6 Maintenance record acquisition section 7 Learning section 8 Learning model storage section 9 Mold identification information acquisition section
10 Cumulative results acquisition section
11 Storage section
12 Display section

Claims (7)

a data acquisition unit that acquires variation data of actual inspection values for each cumulative number of molds;
a maintenance record acquisition unit that acquires a maintenance record of the mold corresponding to the variation data;
A learning model for predicting maintenance timing of the mold is created by performing supervised learning using the set of the variation data acquired by the data acquisition unit and the maintenance record acquired by the maintenance record acquisition unit as training data. A learning department to be built,
A machine learning device equipped with The variation data indicates the extent to which the trend of the actual inspection value for each cumulative number of molds deviates from the control reference value of the processed product to the upper or lower limit value, and The machine learning device according to claim 1, wherein the record indicates a maintenance time when maintenance was actually performed for the degree of deviation of the trend. The machine learning device according to claim 1, wherein the variation data is calculated using actual measurement values of dimensions of processed products input by the user and the number of results using a mold. a data acquisition step of acquiring variation data of actual inspection values for each cumulative number of molds;
a maintenance record acquisition step of acquiring a maintenance record of the mold corresponding to the variation data;
By performing supervised learning using a set of the variation data acquired in the data acquisition step and the maintenance record acquired in the maintenance record acquisition step as teacher data, a learning model for predicting the maintenance period of the mold is created. learning steps to build,
machine learning methods, including; The variation data indicates the extent to which the trend of the actual inspection value for each cumulative number of molds deviates from the control reference value of the processed product to the upper or lower limit value, and 5. The machine learning method according to claim 4, wherein the record indicates a maintenance time when maintenance was actually performed for the degree of deviation of the trend. 5. The machine learning method according to claim 4, wherein the variation data is calculated using actual measurement values of dimensions of processed products input by the user and the number of actual results using a mold. a mold identification information acquisition unit that acquires mold identification information of the mold;
and a cumulative performance number acquisition unit that acquires the cumulative performance number of the mold,
A learning model constructed by the machine learning device according to any one of claims 1 to 3 is applied to the cumulative performance number acquired by the cumulative performance number acquisition unit, and the mold identification information acquisition unit A mold maintenance time prediction device that predicts a maintenance time for the mold corresponding to the mold identification information acquired in the mold identification information.

Images