JP6958461B2 - Control device, control method, and control program - Google Patents

Description

The present invention relates to control devices, control methods, and control programs.

In recent years, predictive control techniques for predicting a future state and controlling an operation suitable for the predicted future state have been developed in various devices. For example, in Patent Document 1, a control device that controls the operation of a vehicle, a first arithmetic unit that controls the running of the vehicle by speculative execution, and safety even if an unpredictable situation occurs by speculative execution. A control device including a second calculation unit that controls the operation of the vehicle so as to be able to secure the above has been proposed. Specifically, the first calculation unit predicts future input values and controls the operation of the vehicle so as to perform optimum traveling. On the other hand, the second calculation unit controls the operation of the vehicle so as to travel at a speed equal to the target speed and suddenly brake when the distance between the own vehicle and the pedestrian becomes a certain value or less. When the control by the first calculation unit satisfies a certain safety standard, this control device controls the operation of the vehicle by using the calculation result of the first calculation unit, and the control by the first calculation unit is performed. When a certain safety standard is not satisfied, the operation of the vehicle is controlled by using the calculation result of the second calculation unit. As a result, according to Patent Document 1, it is possible to control the operation of the vehicle so as to perform optimum traveling while ensuring a certain level of safety.

Japanese Unexamined Patent Publication No. 2016-192010

The present inventors have found that the following problems may occur in the case of predicting and controlling the operation of a production apparatus that produces a product from a work in a relatively short cycle. That is, in such a production apparatus, the timing of processing the work comes in a relatively short cycle. Therefore, if the arithmetic processing of the predictive control is complicated, the computational resources (for example, a processor, etc.) are preferentially allocated to other processing, and the arithmetic processing of the predictive control is delayed. There is a possibility that the arithmetic processing of the prediction control is not completed by the timing of processing the work. The present inventors have found that if the arithmetic processing of the predictive control is not completed by the timing of processing the work, there arises a problem that the predictive control of the production apparatus cannot be properly executed.

The present invention, on the one hand, has been made in view of such circumstances, and an object of the present invention is to provide a technique for appropriately managing the calculation time of predictive control of a production apparatus.

The present invention employs the following configuration in order to solve the above-mentioned problems.

That is, the control device according to one aspect of the present invention is a control device including a hardware processor and a memory for holding a program executed by the hardware processor, and the hardware processor is included in the program. Calculation of a prediction model constructed to predict a command value from a work to a production device that produces a product based on a series of instructions included and adapted to the production of the product by the production device. The prediction is based on the step of starting the processing, the step of calculating the remaining processing time until the calculation processing of the prediction model is completed after starting the calculation processing of the prediction model, and the calculated remaining processing time. When the step of determining whether or not the determination of the command value based on the output obtained from the model is in time for the control timing for controlling the operation of the production by the production apparatus and the determination of the command value are in time for the control timing. When the determination is made, the output is acquired from the prediction model by continuing until the arithmetic processing of the prediction model is completed, and the command value to the production apparatus is determined based on the output acquired from the prediction model. The step and the step of controlling the operation of the production apparatus based on the determined command value are executed.

In the control device related to the configuration, when predicting the command value to the production device using the prediction model, the step of calculating the remaining processing time and the determination of the command value by the prediction model based on the remaining processing time are performed. The step of determining whether or not the control timing is met is executed. As a result, the control device according to the configuration specifies whether or not the arithmetic processing related to the prediction of the command value using the prediction model is in time for the timing of controlling the operation of the production device using the predicted command value. be able to. Therefore, according to the configuration, it becomes possible to appropriately manage the calculation time of the prediction control of the production apparatus.

In addition, the control device according to the configuration can appropriately execute the prediction control as it is when it is determined that the determination of the command value by the prediction model is in time for the control timing. On the other hand, when the control device according to the configuration determines that the determination of the command value by the prediction model is not in time for the control timing, the prediction control is stopped, the command value is determined by another method, and the like. The operation of the production apparatus can be appropriately controlled by a method other than predictive control. Therefore, according to the configuration, it is possible to prevent the process of controlling the operation of the production apparatus from being delayed because the determination of the command value by the prediction model is not in time for the control timing.

The "production device" is not particularly limited as long as it is a device that can perform some production processing and can be controlled. For example, a press machine, an injection molding machine, an NC lathe, an electric discharge machine, a packaging machine, and the like. It may be a transport machine, a transport mechanism in an inspection machine, or the like. The "work" is not particularly limited as long as it can be a work target of the production apparatus, and may be, for example, a raw material of a product, a product before processing, a part before assembly, or the like. The "product" is a product obtained by performing a production process on the work by the production apparatus, and may include an intermediate product (a product in the process of processing) in addition to the final product. The "prediction model" is not particularly limited as long as it is a model that can predict the command value to the production apparatus at a time point (future time point) before the time point when the prediction process is executed, and depends on the embodiment. It may be appropriately selected. As the "prediction model", for example, a learning model such as a decision tree, a neural network, or a support vector man may be used.

When the control device according to the one aspect determines that the determination of the command value is not in time for the control timing, the hardware processor may further execute a step of accelerating the arithmetic processing of the prediction model. According to this configuration, by accelerating the arithmetic processing of the prediction model, it is possible to make the determination of the command value in time for the control timing. As a result, the possibility of discontinuing the predictive control can be reduced, and the predictive control can be appropriately executed. It should be noted that the acceleration of the arithmetic processing can be realized by giving priority to the allocation of hardware resources for the arithmetic processing.

In the control device according to the one aspect, the hardware processor may be composed of one or a plurality of cores, and at least one core may be assigned to the arithmetic processing of the prediction model. , The arithmetic processing of the prediction model may be accelerated by increasing the utilization rate of the allocated core with respect to the arithmetic processing of the prediction model. According to this configuration, the arithmetic processing of the prediction model can be appropriately accelerated, whereby the prediction control can be appropriately executed.

In the control device according to the one aspect, the hardware processor may be composed of one or a plurality of cores, and at least one core may be assigned to the arithmetic processing of the prediction model. , The arithmetic processing of the prediction model may be accelerated by increasing the priority of the arithmetic processing of the prediction model in the assigned core. According to this configuration, the arithmetic processing of the prediction model can be appropriately accelerated, whereby the prediction control can be appropriately executed.

In the control device according to the one aspect, the hardware processor may be composed of a plurality of cores, and the first core of the plurality of cores may be assigned to the arithmetic processing of the prediction model. The hardware processor further allocates the second core, which is the second core of the plurality of cores and has not been assigned to the arithmetic processing of the prediction model, to the arithmetic processing of the prediction model. , The arithmetic processing of the prediction model may be accelerated. According to this configuration, the arithmetic processing of the prediction model can be appropriately accelerated, whereby the prediction control can be appropriately executed. The first core and the second core may be one or a plurality of cores, respectively.

In the control device according to the one aspect, the hardware processor may be composed of a plurality of cores, and a third core among the plurality of cores may be assigned to the arithmetic processing of the prediction model. The plurality of cores are fourth cores that are not assigned to the arithmetic processing of the prediction model, and have a free capacity larger than the utilization rate allocated to the arithmetic processing in the third core. The hardware processor may replace the allocation of the prediction model for arithmetic processing from the third core to the fourth core, and allocate to the arithmetic processing in the third core. By allocating a utilization rate larger than the utilization rate to the arithmetic processing in the fourth core, the arithmetic processing of the prediction model may be accelerated. According to this configuration, the arithmetic processing of the prediction model can be appropriately accelerated, whereby the prediction control can be appropriately executed. The third core and the fourth core may be one or a plurality of cores, respectively.

In the control device according to the one aspect, after executing the step of accelerating the arithmetic processing of the prediction model, the hardware processor sets the step of calculating the remaining processing time and the determination of the command value at the control timing. The step of determining whether or not it is in time may be executed again. According to this configuration, as a result of accelerating the arithmetic processing of the prediction model, it is possible to specify whether or not the determination of the command value by the prediction model is in time for the control timing. As a result, it is possible to appropriately manage the calculation time of the prediction control of the production apparatus.

In the control device according to the one aspect, the hardware processor may determine whether or not the arithmetic processing of the prediction model can be accelerated, and it is determined that the arithmetic processing of the prediction model cannot be accelerated. In this case, the arithmetic processing of the prediction model may be stopped, a set value given in advance may be designated as a command value, and the operation of the production apparatus may be controlled based on the designated command value. According to this configuration, when there is no room for accelerating the arithmetic processing of the prediction model and it is determined that the determination of the command value by the prediction model is not in time for the control timing, the operation of the production apparatus is operated by a method other than the prediction control. It can be controlled.

The conditions for determining whether or not the arithmetic processing of the prediction model can be accelerated may be appropriately set according to the embodiment. The case where it is not possible to accelerate the arithmetic processing of the prediction model is, for example, the case where there is no room for preferentially allocating hardware resources to the arithmetic processing. Further, the set value given in advance is, for example, a reference value of a command value and a past command value (for example, a previous value when the repetitive operation is controlled). The past command value may be, for example, the previous value having the same control status.

In the control device according to the one aspect, the prediction model may be configured to output a correction value with respect to the reference value of the command value as a result of prediction of the command value adapted to the production of the product. According to this configuration, the command value to the production apparatus can be appropriately determined by using the correction value obtained from the prediction model.

In the control device according to the one aspect, the prediction model predicts the command value adapted to the production of the product with respect to the input of at least one of the feature amount of the work and the attribute value of the environment in which the product is produced. It may be constructed to do so. According to this configuration, the command value for the production equipment can be determined in consideration of the factors that may affect the production process of the product by the production equipment.

The "feature amount of the work" is not particularly limited as long as it can exhibit some features of the work, and may be appropriately selected according to the embodiment. The feature quantities of the work are, for example, hardness, dimensions, material, weight, heat, and the like. Further, the feature amount of the work may directly indicate the feature of the work or may indirectly indicate the feature of the work. To directly show the characteristics of the work is, for example, expressing the hardness (hardness) of the work itself by a numerical value, a class, or the like. On the other hand, indirectly indicating the characteristics of the work means, for example, a secondary index obtained when measuring the hardness (hardness) of the work (for example, a load applied to the work, a torque applied at the time of measurement). Etc.) is expressed by numerical values, classes, etc.

Further, the "attribute value of the environment in which the product is produced" is not particularly limited as long as it can indicate some attribute related to the environment in which the production apparatus operates, and may be appropriately selected according to the embodiment. The attribute values of the environment in which the product is produced may be, for example, the temperature and humidity around the production apparatus, the degree of deterioration of the apparatus (for example, the number of years of processing, the number of times of processing, etc.), vibration, and the like.

As another aspect of the control device according to each of the above modes, an information processing method that realizes each of the above configurations may be used, a program may be used, or a computer or the like that stores such a program. It may be a storage medium that can be read by an apparatus, a machine, or the like. Here, the storage medium that can be read by a computer or the like is a medium that stores information such as a program by electrical, magnetic, optical, mechanical, or chemical action.

For example, the control method according to one aspect of the present invention is such that the computer predicts a command value from a work to a production apparatus that produces a product, which is adapted to the production of the product by the production apparatus. A step of starting the arithmetic processing of the constructed prediction model, a step of calculating the remaining processing time until the arithmetic processing of the prediction model is completed after starting the arithmetic processing of the prediction model, and the calculated residual processing. A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the time is in time for the control timing for controlling the operation of the production by the production apparatus, and the determination of the command value. When it is determined that the control timing is in time, the output is acquired from the prediction model by continuing until the arithmetic processing of the prediction model is completed, and the production apparatus is based on the output acquired from the prediction model. It is an information processing method that executes a step of determining a command value to the user and a step of controlling the operation of the production apparatus based on the determined command value.

Further, for example, the control program according to one aspect of the present invention predicts to a computer a command value from a work to a production apparatus for producing a product, which is adapted to the production of the product by the production apparatus. A step of starting the arithmetic processing of the prediction model constructed as described above, a step of calculating the remaining processing time until the arithmetic processing of the prediction model is completed after the arithmetic processing of the prediction model is started, and the calculated step. A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the remaining processing time is in time for the control timing for controlling the operation of the production by the production apparatus, and the command value. When it is determined that the determination of is in time for the control timing, the output is acquired from the prediction model by continuing until the arithmetic processing of the prediction model is completed, and based on the output acquired from the prediction model, the said It is a program for executing a step of determining a command value to a production apparatus and a step of controlling the operation of the production apparatus based on the determined command value.

According to the present invention, it is possible to appropriately manage the calculation time of the predictive control of the production apparatus.

FIG. 1 schematically illustrates an example of a situation in which the present invention is applied. FIG. 2 schematically illustrates an example of the hardware configuration of the control device according to the embodiment. FIG. 3 schematically illustrates an example of the production apparatus according to the embodiment. FIG. 4A schematically illustrates an example of a production process in the production apparatus of FIG. FIG. 4B schematically illustrates an example of a production process in the production apparatus of FIG. FIG. 4C schematically illustrates an example of a production process in the production apparatus of FIG. FIG. 4D schematically illustrates an example of a production process in the production apparatus of FIG. FIG. 5 schematically illustrates an example of the hardware configuration of the learning device according to the embodiment. FIG. 6 schematically illustrates an example of the software configuration of the control device according to the embodiment. FIG. 7A schematically illustrates an example of the prediction model according to the embodiment. FIG. 7B schematically illustrates the relationship between inputs and outputs to the prediction model. FIG. 8 schematically illustrates an example of the software configuration of the learning device according to the embodiment. FIG. 9 illustrates an example of the processing procedure of the control device according to the embodiment. FIG. 10 schematically illustrates an example of allocation information. FIG. 11 is a diagram for explaining core scheduling. FIG. 12A schematically illustrates an example of the acceleration process. FIG. 12B schematically illustrates an example of the acceleration process. FIG. 12C schematically illustrates an example of the acceleration process. FIG. 12D schematically illustrates an example of acceleration processing. FIG. 13 illustrates an example of the processing procedure of the learning device according to the embodiment.

Hereinafter, embodiments according to one aspect of the present invention (hereinafter, also referred to as “the present embodiment”) will be described with reference to the drawings. However, the embodiments described below are merely examples of the present invention in all respects. Needless to say, various improvements and modifications can be made without departing from the scope of the present invention. That is, in carrying out the present invention, a specific configuration according to the embodiment may be appropriately adopted. The data appearing in the present embodiment are described in natural language, but more specifically, the data is specified in a pseudo language, a command, a parameter, a machine language, etc. that can be recognized by a computer.

§1 Application example First, an example of a situation in which the present invention is applied will be described with reference to FIG. FIG. 1 schematically illustrates an example of a usage scene of the control device 1 and the learning device 2 according to the present embodiment.

As shown in FIG. 1, the control device 1 according to the present embodiment is an information processing device configured to control the operation of the production device 3. In the example of FIG. 1, the production apparatus 3 is a press machine for processing a work. This press machine is an example of the "production apparatus" of the present invention. However, the production device to which the control device 1 can be applied is not limited to such a press machine, and may be appropriately selected according to the embodiment. The production apparatus 3 may be, for example, an injection molding machine, an NC lathe, an electric discharge machine, a packaging machine, a transport machine, a transport mechanism in an inspection machine, or the like, in addition to a press machine.

Specifically, the control device 1 according to the present embodiment first predicts the command value from the work to the production device 3 that produces the product, and is adapted to the production of the product by the production device 3. The arithmetic processing of the prediction model (prediction model 5 described later) constructed in 1) is started. After starting the calculation processing of the prediction model, the control device 1 calculates the remaining processing time until the calculation processing of the prediction model is completed. The control device 1 calculates the remaining processing amount of the calculation processing by checking the progress state of the calculation processing, and calculates the remaining processing time based on the calculated remaining processing amount. This remaining processing time may change depending on the state of computational resources in the control device 1. Next, the control device 1 determines whether or not the determination of the command value based on the output obtained from the prediction model is in time for the control timing for controlling the production operation by the production device 3 based on the calculated remaining processing time. do.

As a result, the control device 1 can specify whether or not the arithmetic processing related to the prediction of the command value using the prediction model is in time for the timing of controlling the operation of the production device 3 by using the predicted command value. can. When it is determined that the determination of the command value is in time for the control timing, the control device 1 acquires the output from the prediction model by continuing until the calculation processing of the prediction model is completed, and obtains the output from the prediction model. Based on this, the command value for the production device 3 is determined. Then, the control device 1 controls the operation of the production device 3 based on the determined command value. On the other hand, when it is determined that the command value determination is not in time for the control timing, the control device 1 accelerates the arithmetic processing of the prediction model so that the command value determination by the prediction model is in time for the control timing. It is possible to stop the arithmetic processing of the prediction model and control the operation of the production apparatus 3 by a method other than the prediction control.

On the other hand, the learning device 2 according to the present embodiment is an information processing device configured to build a prediction model used by the control device 1. In the present embodiment, the prediction model is composed of a learning model (decision tree described later). Therefore, the learning device 2 acquires learning data used for machine learning (learning data set 221 described later), and performs machine learning using the acquired learning data to adapt to the production of products by the production device 3. Build a prediction model that has acquired the ability to predict the command value. The control device 1 can acquire a trained prediction model created by the learning device 2 via a network, for example. The type of network may be appropriately selected from, for example, the Internet, a wireless communication network, a mobile communication network, a telephone network, a dedicated network, and the like.

As described above, in the present embodiment, when the command value to the production apparatus 3 is predicted by using the prediction model, the step of calculating the remaining processing time and the command value by the prediction model based on the remaining processing time. Perform a step to determine if the decision is in time for control timing. Thereby, it is possible to specify whether or not the arithmetic processing related to the prediction of the command value using the prediction model is in time for controlling the operation of the production apparatus 3 by using the predicted command value. Therefore, according to the present embodiment, it is possible to appropriately manage the calculation time of the prediction control of the production apparatus 3.

In particular, the control device 1 can process a plurality of tasks including tasks other than the arithmetic processing of the prediction model. In this case, for example, due to the occurrence of an interrupt task, the computational resources in the control device 1 may not always be available under the same conditions. That is, the state of the computational resources in the control device 1 changes from moment to moment, and the computational resources that can be used for the arithmetic processing of the prediction model may change accordingly. Therefore, the processing time required for the arithmetic processing of the prediction model is not always constant. Therefore, even if the control device 1 is equipped with a prediction model determined to be capable of processing in time for the control timing based on the processing time specified in advance by simulation or the like, the arithmetic processing of the prediction model may not be in time for the control timing. There is sex. On the other hand, in the present embodiment, when the control device 1 executes the arithmetic processing of the prediction model, the control device 1 calculates the remaining processing time required for the arithmetic processing. As a result, even if the state of the computational resources changes from moment to moment, the control device 1 can determine whether or not the arithmetic processing of the prediction model is in time for the control timing. Therefore, according to the present embodiment, it is possible to appropriately manage the calculation time of the prediction control of the production apparatus 3.

§2 Configuration example [Hardware configuration]
<Control device>
Next, an example of the hardware configuration of the control device 1 according to the present embodiment will be described with reference to FIG. FIG. 2 schematically illustrates an example of the hardware configuration of the control device 1 according to the present embodiment.

As shown in FIG. 2, in the control device 1 according to the present embodiment, the control unit 11, the storage unit 12, the communication interface 13, the external interface 14, the input device 15, the output device 16, and the drive 17 are electrically connected. It is a computer. In FIG. 2, the communication interface and the external interface are described as "communication I / F" and "external I / F", respectively.

The control unit 11 includes a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), and the like, which are hardware processors, and is configured to execute information processing based on a program and various data. NS. The storage unit 12 is an example of a memory, and is composed of, for example, a hard disk drive, a solid state drive, or the like. In the present embodiment, the storage unit 12 stores various information such as the control program 81 executed by the control unit 11 (CPU), the allocation data 121, and the learning result data 225.

The control program 81 is a program for causing the control device 1 to execute information processing (FIG. 9) described later that controls the operation of the production device 3, and includes a series of instructions for the information processing. The allocation data 121 indicates the allocation of tasks to computational resources (for example, hardware processor, memory) in the control device 1. The training result data 225 is data for setting the trained prediction model. Details will be described later.

The communication interface 13 is, for example, a wired LAN (Local Area Network) module, a wireless LAN module, or the like, and is an interface for performing wired or wireless communication via a network. By using the communication interface 13, the control device 1 can perform data communication via the network with another information processing device (for example, the learning device 2).

The external interface 14 is, for example, a USB (Universal Serial Bus) port, a dedicated port, or the like, and is an interface for connecting to an external device. The type and number of the external interfaces 14 may be appropriately selected according to the type and number of connected external devices. In the present embodiment, the control device 1 is connected to the production device 3 via the external interface 14. As a result, the control device 1 can control the operation of the production device 3 by transmitting a command value to the production device 3.

The input device 15 is, for example, a device for inputting a mouse, a keyboard, or the like. The output device 16 is, for example, a device for outputting a display, a speaker, or the like. The operator can operate the control device 1 by using the input device 15 and the output device 16.

The drive 17 is, for example, a CD drive, a DVD drive, or the like, and is a drive device for reading a program stored in the storage medium 91. The type of the drive 17 may be appropriately selected according to the type of the storage medium 91. At least one of the control program 81 and the learning result data 225 may be stored in the storage medium 91.

The storage medium 91 transfers the information of the program or the like by electrical, magnetic, optical, mechanical or chemical action so that the computer or other device, the machine or the like can read the information of the recorded program or the like. It is a medium to accumulate. The control device 1 may acquire at least one of the control program 81 and the learning result data 225 from the storage medium 91.

Here, FIG. 2 illustrates a disc-type storage medium such as a CD or DVD as an example of the storage medium 91. However, the type of the storage medium 91 is not limited to the disc type, and may be other than the disc type. Examples of storage media other than the disk type include semiconductor memories such as flash memories.

Regarding the specific hardware configuration of the control device 1, components can be omitted, replaced, or added as appropriate according to the embodiment. For example, the control unit 11 may include a plurality of hardware processors. The hardware processor may be composed of a microprocessor, an FPGA (field-programmable gate array), a DSP (digital signal processor), or the like. The storage unit 12 may be composed of a RAM and a ROM included in the control unit 11. At least one of the communication interface 13, the external interface 14, the input device 15, the output device 16, and the drive 17 may be omitted. The control device 1 may be composed of a plurality of computers. In this case, the hardware configurations of the computers may or may not match. Further, the control device 1 may be a general-purpose controller, a general-purpose server device, a general-purpose desktop PC (Personal Computer), a notebook PC, a tablet PC, or the like, in addition to an information processing device designed exclusively for the provided service. good.

<Production equipment>
Next, an example of the hardware configuration of the production apparatus 3 according to the present embodiment will be described with reference to FIG. FIG. 3 schematically illustrates an example of the hardware configuration of the production apparatus 3 according to the present embodiment.

The production apparatus 3 according to the present embodiment includes a servo driver 31, an upper mold 32, and a lower mold 33. While the lower mold 33 is fixed, the upper mold 32 is configured to be movable in the vertical direction by a servomotor (not shown). As a result, the upper mold 32 can press the work against the lower mold 33 to form the work or move away from the lower mold 33. The servo driver 31 is configured to drive the servomotor of the upper mold 32 based on the command value from the control device 1.

Next, an example of the production process in the production apparatus 3 is schematically illustrated with reference to FIGS. 4A to 4D. The production apparatus 3 is arranged on the production line, for example. As shown in FIG. 4A, in the initial state, the upper mold 32 is arranged at a standby position away from the lower mold 33, and waits until the work 40 is conveyed to the lower mold 33. The work 40 is, for example, a metal plate material. However, the work 40 is not limited to such an example, and may be appropriately selected depending on the type of the production apparatus 3. The work 40 may be, for example, a raw material of a product, a product before processing, a part before assembly, or the like.

After the work 40 is arranged at a predetermined position of the lower mold 33, the production apparatus 3 drives the servomotor of the upper mold 32 by the servo driver 31 as shown in FIG. 4B to push the upper mold 32. Place it at the molding start position. The molding start position is, for example, a position where the tip of the upper mold 32 comes into contact with the work 40 or immediately before the work 40.

Then, as shown in FIG. 4C, the production apparatus 3 further drives the servomotor of the upper mold 32 by the servo driver 31, moves the upper mold 32 to the target position (bottom dead point), and moves the upper mold 32 to the target position (bottom dead point). The work 40 is formed by the lower mold 33 and the lower mold 33. As a result, the production apparatus 3 can produce the product 41 from the work 40. The product 41 is not particularly limited as long as it is obtained by performing the production process on the work 40 by the production apparatus 3, and may be a final product or an intermediate product (during processing). It may be).

After the molding is completed, the production apparatus 3 drives the servomotor of the upper mold 32 by the servo driver 31 as shown in FIG. 4D, and moves the upper mold 32 to the standby position. Then, the product 41 obtained by molding the work 40 is conveyed from the production apparatus 3 by a belt conveyor (not shown) or the like. As a result, a series of production steps for producing the product 41 from the work 40 is completed.

In this production process, if the pressing time in FIG. 4C is insufficient or the servomotor is not driven until the upper die 32 reaches the bottom dead center, the quality of the obtained product 41 deteriorates. .. Therefore, conventionally, on-site workers regularly check the quality of products and adjust the operation settings of the production equipment to suppress the occurrence of defective products. On the other hand, the control device 1 according to the present embodiment predicts an appropriate command value to the production device 3 by using the prediction model so that a defect does not occur in the production process. As a result, the control device 1 automatically adjusts the operation of the production device 3 so as to suppress the occurrence of defective products.

<Learning device>
Next, an example of the hardware configuration of the learning device 2 according to the present embodiment will be described with reference to FIG. FIG. 5 schematically illustrates an example of the hardware configuration of the learning device 2 according to the present embodiment.

As shown in FIG. 5, the learning device 2 according to the present embodiment is a computer to which the control unit 21, the storage unit 22, the communication interface 23, the input device 24, the output device 25, and the drive 26 are electrically connected. .. In FIG. 5, the communication interface is described as "communication I / F".

Like the control unit 11, the control unit 21 includes a CPU, RAM, ROM, etc., which are hardware processors, and is configured to execute information processing based on a program and various data. The storage unit 22 is composed of, for example, a hard disk drive, a solid state drive, or the like. The storage unit 22 includes various types of learning program 82 executed by the control unit 21 (CPU), learning data set 221 used for machine learning of the prediction model, learning result data 225 created by executing the learning program 82, and the like. Memorize information.

The learning program 82 is a program for causing the learning device 2 to execute information processing (FIG. 13) described later in machine learning of the prediction model, and to generate learning result data 225 as a result of the machine learning. The learning program 82 includes a series of instructions for the information processing. The learning data set 221 is data used for machine learning for constructing a prediction model that has acquired the ability to predict a command value adapted to the production of a product by the production apparatus 3. Details will be described later.

The communication interface 23 is, for example, a wired LAN module, a wireless LAN module, or the like, and is an interface for performing wired or wireless communication via a network. By using the communication interface 23, the learning device 2 can perform data communication via the network with another information processing device (for example, the control device 1). Further, the learning device 2 can distribute the created learning result data 225 to an external device by using the communication interface 23.

The input device 24 is, for example, a device for inputting a mouse, a keyboard, or the like. The output device 25 is, for example, a device for outputting a display, a speaker, or the like. The operator can operate the learning device 2 via the input device 24 and the output device 25.

The drive 26 is, for example, a CD drive, a DVD drive, or the like, and is a drive device for reading a program stored in the storage medium 92. The type of the drive 26 may be appropriately selected according to the type of the storage medium 92. At least one of the learning program 82 and the learning data set 221 may be stored in the storage medium 92.

The storage medium 92 stores the information of the program or the like by electrical, magnetic, optical, mechanical or chemical action so that the information of the program or the like recorded by the computer or other device, the machine or the like can be read. It is a medium to do. The learning device 2 may acquire at least one of the learning program 82 and the learning data set 221 from the storage medium 92.

Here, in FIG. 5, similarly to FIG. 2 above, as an example of the storage medium 92, a disc-type storage medium such as a CD or DVD is illustrated. However, the type of the storage medium 92 is not limited to the disc type, and may be other than the disc type. Examples of storage media other than the disk type include semiconductor memories such as flash memories.

Regarding the specific hardware configuration of the learning device 2, the components can be omitted, replaced, or added as appropriate according to the embodiment. For example, the control unit 21 may include a plurality of hardware processors. The hardware processor may be composed of a microprocessor, FPGA, ECU and the like. The learning device 2 may be composed of a plurality of information processing devices. Further, as the learning device 2, in addition to an information processing device designed exclusively for the provided service, a general-purpose server device, a PC, or the like may be used.

[Software configuration]
<Control device>
Next, an example of the software configuration of the control device 1 according to the present embodiment will be described with reference to FIG. FIG. 6 schematically illustrates an example of the software configuration of the control device 1 according to the present embodiment.

The control unit 11 of the control device 1 expands the control program 81 stored in the storage unit 12 into the RAM. Then, the control unit 11 interprets and executes the control program 81 expanded in the RAM by the CPU, and controls each component based on a series of instructions included in the control program 81. As a result, as shown in FIG. 6, the control device 1 according to the present embodiment includes the data acquisition unit 111, the prediction calculation unit 112, the remaining processing time calculation unit 113, the delay determination unit 114, the acceleration processing unit 115, and the operation control. It operates as a computer including unit 116 as a software module. That is, in this embodiment, each software module is realized by the control unit 11 (CPU).

The data acquisition unit 111 acquires data used by the prediction model 5 to predict a command value adapted to the production of the product 41 by the production apparatus 3. In the present embodiment, the prediction model 5 is constructed so as to predict the command value adapted to the production of the product 41 with respect to the input of the feature amount 61 of the work 40 and the attribute value 62 of the environment for producing the product 41. .. Therefore, the data acquisition unit 111 acquires the feature amount 61 of the work 40 and the attribute value 62 of the environment for producing the product 41.

The prediction calculation unit 112 uses a prediction model 5 that is a command value from the work 40 to the production device 3 that produces the product 41 and is constructed so as to predict a command value adapted to the production of the product 41 by the production device 3. I have. In order to carry out the prediction control, the prediction calculation unit 112 starts the calculation processing of the prediction model 5. The remaining processing time calculation unit 113 calculates the remaining processing time from the start of the calculation processing of the prediction model 5 to the completion of the calculation processing of the prediction model 5. Based on the calculated remaining processing time, the delay determination unit 114 determines whether or not the determination of the command value based on the output obtained from the prediction model 5 is in time for the control timing for controlling the operation of production by the production apparatus 3. ..

When it is determined that the determination of the command value is in time for the control timing, the prediction calculation unit 112 acquires an output from the prediction model 5 by continuing until the calculation processing of the prediction model 5 is completed. The operation control unit 116 determines the command value to the production device 3 based on the output acquired from the prediction model 5. Then, the operation control unit 116 controls the operation of the production device 3 based on the determined command value.

In the present embodiment, the prediction model 5 is configured to output a correction value 63 with respect to the reference value 60 of the command value as an output value corresponding to the prediction result of the command value adapted to the production of the product 41. Therefore, the prediction calculation unit 112 acquires the output value corresponding to the correction value 63 from the prediction model 5 by completing the calculation processing of the prediction model 5. In response to this, the operation control unit 116 determines the value obtained by correcting the reference value 60 with the acquired correction value 63 as the command value, and operates the production device 3 based on the determined command value. Control.

On the other hand, in the present embodiment, when it is determined that the determination of the command value is not in time for the control timing, the acceleration processing unit 115 accelerates the arithmetic processing of the prediction model 5. The acceleration processing unit 115 accelerates the arithmetic processing of the prediction model 5, for example, by preferentially allocating the computational resources of the hardware to the arithmetic processing of the prediction model 5. If it is determined that the calculation process of the prediction model 5 cannot be accelerated and the command value is not determined in time for the control timing, the prediction calculation unit 112 stops the calculation process of the prediction model 5. Then, the operation control unit 116 designates a set value given in advance as a command value, and controls the operation of the production apparatus 3 based on the designated command value.

(Prediction model)
Next, the prediction model 5 according to the present embodiment will be described with reference to FIGS. 7A and 7B. FIG. 7A schematically illustrates an example of the configuration of the prediction model 5 according to the present embodiment. Further, FIG. 7B schematically illustrates the relationship between the input and the output with respect to the prediction model 5.

As shown in FIG. 7A, the prediction model 5 according to the present embodiment is composed of a decision tree (specifically, a regression tree). The prediction model 5 (decision tree) includes a root node R, leaf nodes L1 to L5, and intermediate nodes N1 to N3 arranged between the root node R and the leaf nodes L1 to L5. A link is provided between each node. In the example of FIG. 7A, between the root node R and the intermediate node (N1, N2), between the intermediate node N1 and each leaf node (L1, L2), between the intermediate node N2 and the leaf node L3 and the intermediate node N3. A link is provided between the intermediate node N3 and each leaf node (L4, L5).

In the example of FIG. 7A, the depth of the decision tree is 4, the number of intermediate nodes is 3, and the number of leaf nodes is 5. However, the depth of the decision tree, the number of intermediate nodes, and the number of leaf nodes need not be limited to such an example, and may be appropriately determined according to the embodiment. Further, in the example of FIG. 7A, no link is provided from the root node R to each leaf node L1 to L5. However, the composition of the decision tree does not have to be limited to such an example, and there may be a leaf node connected to the link from the root node.

Such arithmetic processing of the prediction model 5 is a search processing that follows a link from the root node R of the determination tree to the leaf nodes L1 to L5. That is, a branching condition is associated with the route from the root node R to the leaf nodes L1 to L5 (in the example of FIG. 7A, the root node R and the intermediate nodes N1 to N3). In the example of FIG. 7A, the root node R has a branch condition of "x0 <2500", the intermediate node N1 has a branch condition of "x1 <20", and the intermediate node N2 has a branch condition of "x1 <35". , The branch condition of "x0 <3500" is associated with the intermediate node N3. On the other hand, the final results (classes C1 to C5) of the arithmetic processing of the prediction model 5 are associated with the leaf nodes L1 to L5. In the present embodiment, each leaf node L1 to L5 (classes C1 to C5) is associated with a correction value 63 corresponding to the input feature amount 61 and the attribute value 62. In the present embodiment, the prediction calculation unit 112 reaches any of the leaf nodes L1 to L5 by starting the search process from the root node R and repeating the determination of whether or not the input data satisfies the branching condition. We will proceed with the search to deeper nodes.

In the example of FIG. 7A, the input x0 corresponds to the feature amount 61, and the input x1 corresponds to the attribute value 62. FIG. 7B illustrates the relationship between each input (x0, x1) and the classes C1 to C5 associated with the arriving leaf nodes L1 to L5. For example, it is assumed that the input x0 is 2000 and the input x1 is 30. In this case, the prediction calculation unit 112 determines whether or not the input x0 satisfies the branch condition set in the root node R as the calculation processing (search processing) of the first layer of the prediction model 5. In the example of FIG. 7A, the branch condition set in the root node R is "x0 <2500" and the input x0 is 2000. Therefore, the prediction calculation unit 112 inputs the branch condition set in the root node R x0. Is determined to satisfy, and the search proceeds to the intermediate node N1 in the next hierarchy.

Next, the prediction calculation unit 112 determines whether or not the input x1 satisfies the branch condition set in the intermediate node N1 as the second-layer calculation processing of the prediction model 5. In the example of FIG. 7A, the branch condition set in the intermediate node N1 is "x1 <20" and the input x1 is 30, so that the prediction calculation unit 112 inputs the branch condition set in the intermediate node N1 x1. Is not satisfied, and the process proceeds to the leaf node L2 in the next hierarchy. As a result, the search process of the decision tree reaches the leaf node L2, so that the arithmetic process of the prediction model 5 is completed. The prediction calculation unit 112 can acquire the correction value 63 associated with the class C2 of the leaf node L2 as the final result of the calculation processing of the prediction model 5.

The method of acquiring the correction value 63 associated with each of the classes C1 to C5 may be appropriately determined according to the embodiment. For example, correction values may be directly associated with each class C1 to C5. Further, for example, the control device 1 may hold reference information such as a table format indicating the correspondence between the classes C1 to C5 and the correction value in the storage unit 12. This reference information may be included in the learning result data 225. In this case, after arriving at any leaf node, the prediction calculation unit 112 collates the class of the reached leaf node with the reference information, and as a final result of the calculation processing of the prediction model 5, the reference value of the command value. A correction value 63 for 60 can be obtained.

Information indicating the configuration of the prediction model 5 (decision tree) and each branching condition is included in the learning result data 225. By referring to the learning result data 225, the prediction calculation unit 112 can set the trained prediction model 5 used in the process of predicting the command value adapted to the production of the product 41.

<Learning device>
Next, an example of the software configuration of the learning device 2 according to the present embodiment will be described with reference to FIG. FIG. 8 schematically illustrates an example of the software configuration of the learning device 2 according to the present embodiment.

The control unit 21 of the learning device 2 expands the learning program 82 stored in the storage unit 22 into the RAM. Then, the control unit 21 interprets and executes the learning program 82 expanded in the RAM by the CPU, and controls each component based on a series of instructions included in the learning program 82. As a result, as shown in FIG. 8, the learning device 2 according to the present embodiment is configured as a computer including the learning data acquisition unit 211 and the learning processing unit 212 as software modules. That is, in this embodiment, each software module is realized by the control unit 21 (CPU).

The learning data acquisition unit 211 acquires learning data to be used for machine learning of the prediction model 5. In the present embodiment, since the prediction model 5 is composed of a determination tree, the training data acquisition unit 211 has a training data set 221 composed of a combination of training data (input data) and teacher data (correct answer data). get. In the present embodiment, the prediction model 5 is configured to output a correction value 63 with respect to the reference value 60 of the command value in response to the input of the feature amount 61 of the work 40 and the attribute value 62 of the environment in which the product 41 is produced. NS. Therefore, the learning data set 221 is composed of a combination of the feature amount 2211 of the work, the attribute value 2212 of the environment in which the product is produced, and the correction value 2213 adapted to the attribute value 2212.

The learning processing unit 212 builds the trained prediction model 5 by performing machine learning using the acquired learning data set 221. When the feature amount 2211 and the attribute value 2212 are input, the learning processing unit 212 constructs a decision tree so as to reach the leaf node of the class corresponding to the correction value 2213 associated with the input feature amount 2211 and the attribute value 2212. .. As a result, the learning processing unit 212 can build the trained prediction model 5. Then, the learning processing unit 212 stores the information indicating the configuration of the constructed learned prediction model 5 and each branching condition in the storage unit 22 as the learning result data 225.

<Others>
Each software module of the control device 1 and the learning device 2 will be described in detail in an operation example described later. In this embodiment, an example in which each software module of the control device 1 and the learning device 2 is realized by a general-purpose CPU is described. However, some or all of the above software modules may be implemented by one or more dedicated processors. Further, with respect to the software configurations of the control device 1 and the learning device 2, software modules may be omitted, replaced, or added as appropriate according to the embodiment.

§3 Operation example [Control device]
Next, an operation example of the control device 1 will be described with reference to FIG. FIG. 9 is a flowchart illustrating an example of the processing procedure of the control device 1. The processing procedure described below is an example of the "control method" of the present invention. However, the processing procedure described below is only an example, and each processing may be changed as much as possible. Further, with respect to the processing procedure described below, steps can be omitted, replaced, and added as appropriate according to the embodiment.

(Step S100)
In step S100, the control unit 11 operates as the data acquisition unit 111 to acquire input data for the prediction model 5.

In the present embodiment, as described above, the prediction model 5 predicts the command value adapted to the production of the product 41 with respect to the input of the feature amount 61 of the work 40 and the attribute value 62 of the environment in which the product 41 is produced. It is built on. Therefore, the control unit 11 acquires the feature amount 61 of the work 40 and the attribute value 62 of the environment for producing the product 41 as input data for the prediction model 5.

The feature amount 61 of the work 40 is not particularly limited as long as it can exhibit some features of the work, and may be appropriately selected depending on the embodiment. Further, the attribute value 62 of the environment in which the product 41 is produced does not have to be particularly limited as long as it can indicate some attribute related to the environment in which the production apparatus 3 operates, and may be appropriately selected depending on the embodiment. ..

In the present embodiment, the production apparatus 3 is a press machine. As described above, in the production apparatus 3, if the pressing time is insufficient or the servomotor is not driven until the upper die 32 reaches the bottom dead center, the quality of the obtained product 41 deteriorates. It ends up. Therefore, it is preferable that the feature amount 61 of the work 40 and the attribute value 62 of the environment in which the product 41 is produced relate to the press molding process in the production apparatus 3, respectively.

Therefore, for the feature amount 61 of the work 40, for example, the hardness, dimensions, material, weight, heat, etc. of the work 40 may be selected. Further, for the attribute value 62 of the environment in which the product 41 is produced, for example, the temperature and humidity around the production apparatus 3, the degree of deterioration of the apparatus (for example, the number of years of processing, the number of times of processing, etc.), vibration, and the like may be selected. At this time, the feature amount 61 of the work 40 may directly indicate the features of the work 40 or may indirectly indicate the features of the work 40. That is, when the feature amount 61 of the work 40 relates to the hardness of the work 40, the feature amount 61 of the work 40 may indicate the hardness (hardness) of the work 40 itself by a numerical value, a class, or the like. Further, the feature amount 61 of the work 40 is a numerical value of a secondary index (for example, a load applied to the work, a torque applied at the time of measurement, etc.) obtained when measuring the hardness (hardness) of the work 40. , Class, etc. may be indicated. The same applies to the attribute value 62.

Further, the method of acquiring the feature amount 61 of the work 40 and the attribute value 62 of the environment in which the product 41 is produced may be appropriately selected according to the embodiment. For example, various sensors configured to be able to measure the feature amount 61 (for example, hardness, etc.) of the work 40 and the environmental attribute value 62 (for example, temperature, etc.) may be arranged in the production apparatus 3. As various sensors, known sensors may be appropriately used depending on the types of the feature amount 61 and the attribute value 62 to be measured. In this case, the control unit 11 can acquire the feature amount 61 of the work 40 and the attribute value 62 of the environment for producing the product 41 from various sensors arranged in the production device 3. When the feature amount 61 of the work 40 and the attribute value 62 of the environment for producing the product 41 are acquired, the control unit 11 proceeds to the next step S101.

(Step S101)
In step S101, the control unit 11 operates as the prediction calculation unit 112, and is a command value from the work 40 to the production device 3 that produces the product 41, and is a command value adapted to the production of the product 41 by the production device 3. The arithmetic processing of the prediction model 5 constructed to predict is started.

In the present embodiment, the prediction model 5 is composed of a decision tree, and information indicating the structure of the prediction model 5 and the branching conditions of each route is included in the learning result data 225. Therefore, the control unit 11 sets the prediction model 5 by referring to the learning result data 225. By this setting process, the control unit 11 is in a state where the search process of the decision tree (prediction model 5) can be started. As a result, when the arithmetic processing of the prediction model 5 is started, the control unit 11 proceeds to the next step S102.

(Step S102)
In step S102, the control unit 11 operates as the prediction calculation unit 112 to advance the calculation processing of the prediction model 5. In the present embodiment, since the prediction model 5 is composed of a decision tree, the arithmetic processing of the prediction model 5 is a search process that follows a link from the root node of the decision tree to the leaf node. Therefore, the control unit 11 advances the search process of the decision tree as the arithmetic process of the prediction model 5.

Specifically, when the search process has never been executed, the control unit 11 inputs the branch condition set in the root node as input data (feature amount 61 and attribute value 62) as the search process of the decision tree. ) Satisfies or not. Then, based on this determination result, the control unit 11 proceeds with the search to the corresponding node in the second stage (intermediate node N1 or intermediate node N2 in the example of FIG. 7A).

Similarly, when the search process is executed n times (n is a natural number of 1 or more), the search proceeds to the intermediate node in the n + 1th stage. In this case, the control unit 11 determines whether or not the input data satisfies the branch condition set in the corresponding intermediate node in the n + 1th stage. Then, based on this determination result, the control unit 11 proceeds with the search to the corresponding node in the n + 2nd stage.

When the decision tree search process is advanced in this way, the control unit 11 proceeds to the next step S103. The degree to which the search process proceeds in one process of this step S102 may be appropriately determined according to the embodiment. When the degree to which the search process is advanced is the depth k (k is a natural number of 1 or more), the control unit 11 executes the determination process of the branch condition k times, and then performs the process in the next step S103. Proceed.

(Step S103)
In step S103, the control unit 11 operates as the remaining processing time calculation unit 113, and calculates the remaining processing time until the calculation processing of the prediction model 5 is completed. The method of calculating the remaining processing time may be appropriately determined according to the embodiment. In the present embodiment, the control unit 11 calculates the remaining processing amount of the calculation processing by checking the progress state of the calculation processing, and calculates the remaining processing time based on the calculated remaining processing amount. Specifically, the control unit 11 calculates the remaining processing time by the following calculation (Equation 1).

「現在までの処理時間」は、ステップＳ１０１により予測モデル５の演算処理を開始してから本ステップＳ１０３を実行するまでの時間である。制御部１１は、例えば、タイマ（不図示）等により、ステップＳ１０１により予測モデル５の演算処理を開始してからの時間を計測することで、現在までの処理時間の値を得ることができる。

The “processing time up to now” is the time from the start of the arithmetic processing of the prediction model 5 in step S101 to the execution of this step S103. The control unit 11 can obtain the value of the processing time up to the present by measuring the time from the start of the arithmetic processing of the prediction model 5 in step S101 by, for example, a timer (not shown) or the like.

The “processing amount up to now” is the processing amount from the start of the arithmetic processing of the prediction model 5 in step S101 to the execution of this step S103. For example, the control unit 11 can obtain the value of the processing amount up to the present by measuring the processing amount of the CPU after starting the arithmetic processing of the prediction model 5 in step S101. Further, in the present embodiment, since the prediction model 5 is composed of a decision tree, the control unit 11 can also measure the number of times the branch condition determination process in the decision tree search process is executed, for example. The value of the processing amount up to can be obtained.

The “residual processing amount” is the difference between the total processing amount of the arithmetic processing of the prediction model 5 and the processing amount up to the present. The control unit 11 can obtain a value of the remaining processing amount by subtracting the processing amount up to the present from the total processing amount of the arithmetic processing of the prediction model 5.

The total amount of arithmetic processing of the prediction model 5 may be estimated as appropriate according to the embodiment. The total amount of processing may be estimated based on, for example, the result of simulating the arithmetic processing of the prediction model 5, the result of executing the arithmetic processing of the prediction model 5 in the past (actual value), and the like. In this case, the total processing amount may be estimated according to the execution time of the simulation or the past arithmetic processing.

Further, in the present embodiment, the prediction model 5 is composed of a decision tree. In this case, the arithmetic processing of the prediction model 5, that is, the time required for the search processing of the decision tree depends on the depth of the decision tree (particularly, the number of intermediate nodes that pass from the root node to the leaf node). do. Therefore, if the depth of the decision tree is fixed, the total processing amount of the decision tree search process is unlikely to fluctuate from the theoretical value derived from the decision tree search algorithm. Therefore, the total amount of processing may be estimated based on, for example, a theoretical value derived from the algorithm of the prediction model 5.

The format for expressing each of the above-mentioned processing amounts may not be particularly limited, and may be appropriately selected depending on the embodiment. Each processing amount may be expressed by, for example, the arithmetic amount of the CPU, the theoretical amount of the arithmetic processing, or the like. In this embodiment, the theoretical amount of arithmetic processing can be indicated by, for example, the depth of the decision tree.

As a specific example, the depth of the decision tree constituting the prediction model 5 is 11, and the determination process of the branch condition is completed up to the intermediate node of the second layer (that is, the search reaches the intermediate node of the third layer). It is assumed). When expressing each processing amount by the depth of the decision tree, if the depth of the decision tree is 11, the number of search processes reaching the leaf node is 10, so the value of the total processing amount is "10". Can be expressed as. Further, the value of the processing amount up to now can be expressed as "2". Therefore, the value of the residual processing amount can be expressed as "10-2 = 8". At this time, assuming that the processing time up to now is 0.01 seconds, the control unit 11 calculates the remaining processing time to be 0 by the calculation of "0.01 (seconds) x ((10-2) / 2)". It can be calculated to be .04 (seconds).

The expression format of the remaining processing time is not particularly limited as long as the timing at which the arithmetic processing of the prediction model 5 is completed can be expressed, and may be appropriately selected according to the embodiment. The remaining processing time may be expressed by a time system (seconds) according to the International System of Units, for example, as in the above example. Further, the remaining processing time may be expressed by, for example, the number of clocks of the CPU, the tact time indicating the operation cycle of the production apparatus 3, and the like. After calculating the remaining processing time, the control unit 11 proceeds to the next step S104.

(Step S104)
In step S104, the control unit 11 operates as the delay determination unit 114, and the determination of the command value based on the output obtained from the prediction model 5 based on the calculated remaining processing time controls the production operation by the production device 3. It is determined whether or not it is in time for the control timing to be performed.

In the present embodiment, the control unit 11 first calculates the remaining time until the control timing by the following equation 2.

「制御予定時刻」は、制御タイミングの時刻であり、予め決定されている。制御部１１は、例えば、制御プログラム８１等の生産装置３の動作を制御する予定を示す情報から、制御予定時刻の値を取得することができる。また、「現在時刻」は、本ステップＳ１０４を実行する時刻である。制御部１１は、例えば、タイマ（不図示）等により、現在時刻の値を取得することができる。ただし、残存時間の計算方法は、このような例に限定されなくてもよく、実施の形態に応じて適宜変更されてよい。

The "scheduled control time" is the time of the control timing and is determined in advance. The control unit 11 can acquire the value of the scheduled control time from, for example, the information indicating the schedule for controlling the operation of the production device 3 such as the control program 81. The "current time" is the time when the step S104 is executed. The control unit 11 can acquire the value of the current time by, for example, a timer (not shown) or the like. However, the method of calculating the remaining time does not have to be limited to such an example, and may be appropriately changed depending on the embodiment.

Next, the control unit 11 calculates the difference by subtracting the remaining processing time from the calculated remaining time. This difference indicates the estimated free time from the completion of the arithmetic processing of the prediction model 5 to the control timing. That is, when this difference is a negative value, it is estimated that the arithmetic processing of the prediction model 5 is completed at a time later than the control timing.

Therefore, the control unit 11 determines whether or not the difference obtained by subtracting the remaining processing time from the remaining time exceeds a predetermined threshold value. Then, when the obtained difference exceeds a predetermined threshold value, the control unit 11 determines that the determination of the command value is in time for the control timing. On the other hand, when the obtained difference is equal to or less than a predetermined threshold value, the control unit 11 determines that the command value is not determined in time for the control timing.

This predetermined threshold value may be set as appropriate. The predetermined threshold value serves as a reference for determining that the determination of the command value is in time for the control timing. Further, the predetermined threshold value can also be a reference for determining whether or not the arithmetic processing of the prediction model 5 is completed in a state where there is a time margin with respect to the control timing. That is, the predetermined threshold value may correspond to the grace (margin) time from the completion of the arithmetic processing of the prediction model 5 to the control of the operation of the production apparatus 3. Therefore, the control unit 11 determines whether or not the arithmetic processing of the prediction model 5 is completed in a state where there is a time margin with respect to the control timing as a determination as to whether or not the determination of the command value is in time for the control timing. You may.

From the above, the control unit 11 can determine whether or not the determination of the command value is in time for the control timing. When it is determined that the command value is determined in time for the control timing, the control unit 11 proceeds to step S107. On the other hand, when it is determined that the determination of the command value is not in time for the control timing, the control unit 11 proceeds to the process in step S105. However, the method of determining whether or not the determination of the command value is in time for the control timing is not limited to such an example, and may be appropriately changed depending on the embodiment.

The expression format of the remaining time may not be particularly limited as long as the control timing can be expressed, as in the case of the remaining processing time, and may be appropriately selected according to the embodiment. The remaining time may be expressed by, for example, a time system (seconds) according to the International System of Units, the number of clocks of the CPU, a tact time indicating an operation cycle of the production apparatus 3, and the like.

(Step S105)
In step S105, the control unit 11 operates as the acceleration processing unit 115, and determines whether or not the arithmetic processing of the prediction model 5 can be accelerated.

In step S106, which will be described later, for example, the computational processing of the prediction model 5 is accelerated by preferentially allocating the computational resources of the hardware to the arithmetic processing of the prediction model 5. Therefore, in the present embodiment, the control unit 11 can accelerate the arithmetic processing of the prediction model 5 by determining whether or not there is room for the computational resources to be preferentially allocated to the arithmetic processing of the prediction model 5. It may be determined whether or not.

When it is determined that there is room for computational resources to be preferentially allocated to the arithmetic processing of the prediction model 5 and the arithmetic processing of the prediction model 5 can be accelerated, the control unit 11 performs the processing in step S106. Proceed. On the other hand, when it is determined that there is no room for preferentially allocating computational resources to the arithmetic processing of the prediction model 5 and it is not possible to accelerate the arithmetic processing of the prediction model 5, the control unit 11 performs the operation unit 11 in step S111. Proceed with processing. However, the method for determining whether or not the arithmetic processing of the prediction model 5 is possible is not limited to such an example, and may be appropriately changed depending on the embodiment.

(Step S106)
In step S106, the control unit 11 operates as the acceleration processing unit 115 to accelerate the arithmetic processing of the prediction model 5. The control unit 11 can accelerate the arithmetic processing of the prediction model 5, for example, by preferentially allocating the computational resources of the hardware to the arithmetic processing of the prediction model 5. Computational resources are, for example, a CPU, a memory, and the like.

Here, an example of a method in which the control unit 11 (CPU) executes a plurality of tasks including the arithmetic processing of the prediction model 5 will be described with reference to FIGS. 10 and 11. FIG. 10 schematically illustrates an example of allocation data 121 showing the status of task allocation to the CPU. FIG. 11 schematically illustrates an example of scheduling in which the CPU executes a task with respect to the assignment of the task shown in FIG.

A CPU, which is a hardware processor, is composed of one or a plurality of cores. The core is a circuit that executes arithmetic processing, and as the number of cores increases, the CPU can execute many arithmetic processing in parallel. Within one cycle (cycle), each core of the CPU executes each task in order from the task with the highest priority by the assigned utilization rate.

For example, in the example of FIG. 10, the CPU has at least two cores, and three tasks A to C are assigned to the cores 01. The priority of task A is set to the highest, the priority of task B is set to the second highest, and the priority of task C is set to the lowest. Further, in the example of FIG. 10, the lengths of the tasks A to C in the vertical direction indicate the utilization rate of the core. Therefore, in the core 01, the utilization rate of task B is set to be the highest, the utilization rate of task A is set to be the second highest, and the utilization rate of task C is set to be the lowest. The unused part is free space. Tasks A to C can be assigned (increasing the utilization rate), and other tasks can be assigned to this free area.

In this case, as shown in FIG. 11, the core 01 first executes the task A having the highest priority in each cycle by the amount corresponding to the utilization rate. Next, the core 01 executes the second highest priority task B by the amount according to the utilization rate. Then, the core 01 executes the task C having the lowest priority by the amount corresponding to the utilization rate.

At the end time of each cycle, each core does not terminate or execute the execution of the low-priority task, but starts the processing of the next cycle, that is, the execution of the highest-priority task. Further, when the execution of each task is completed, the task disappears, and the execution of the task having the next highest priority is given priority.

In the example of FIG. 11, in the first cycle (s) and the second cycle (s + 1), the execution of the task C is terminated, and the execution of the task B is completed in the middle of the third cycle (s + 2) at time t. It is supposed to be done. Therefore, in the third cycle, after the execution of the task B is completed, the core 01 executes the task C by the amount corresponding to the utilization rate. However, the method in which each core executes the task is not limited to such an example, and may be appropriately determined according to the embodiment.

As described above, when the degree of task execution is determined based on the priority and utilization rate, the computational resources for the task are prioritized by increasing at least one of the task priority and utilization rate. Can be assigned to to accelerate task execution. In the present embodiment, the control unit 11 can accelerate the arithmetic processing of the prediction model 5 by, for example, the following four methods.

(1) First Method First, a first method for accelerating the arithmetic processing of the prediction model 5 will be described with reference to FIG. 12A. FIG. 12A schematically illustrates a first method of accelerating the arithmetic processing of the prediction model 5.

In the first method, it is assumed that the CPU, which is a hardware processor, is composed of one or a plurality of cores. In this first method, for example, in step S101 or the like, at least one core is assigned to the arithmetic processing of the prediction model 5. In this state, the control unit 11 accelerates the arithmetic processing of the prediction model 5 by increasing the utilization rate of the assigned core with respect to the arithmetic processing of the prediction model 5.

In the example of FIG. 12A, the arithmetic processing of the prediction model 5 (hereinafter, also referred to as “task Z”) is assigned to the core 01. The control unit 11 refers to the allocation data 121 and confirms whether or not there is a free area in the core 01 allocated to the task Z. If there is no free area in the core 01, the control unit 11 determines that it is impossible to accelerate the arithmetic processing by this first method. On the other hand, when there is a free area in the core 01, the control unit 11 determines that the arithmetic processing by the first method can be accelerated, and at least a part of the free area is assigned to the task Z. Further allocation increases the utilization rate of core 01 for task Z. The amount of free space allocated to task Z may be appropriately set according to the embodiment. As a result, the computational resources allocated to the arithmetic processing of the prediction model 5 can be increased, so that the control unit 11 can accelerate the arithmetic processing of the prediction model 5 (that is, increase the execution amount of the task Z). can.

(2) Second Method Next, a second method for accelerating the arithmetic processing of the prediction model 5 will be described with reference to FIG. 12B. FIG. 12B schematically illustrates a second method of accelerating the arithmetic processing of the prediction model 5.

In the second method, it is assumed that the CPU, which is a hardware processor, is composed of one or a plurality of cores, as in the first method. In this second method, for example, in step S101 or the like, at least one core is assigned to the arithmetic processing of the prediction model 5. In this state, the control unit 11 accelerates the arithmetic processing of the prediction model 5 by increasing the priority of the arithmetic processing of the prediction model 5 in the assigned core.

In the example of FIG. 12B, the arithmetic processing (task Z) of the prediction model 5 is assigned to the core 01. In addition to task Z, other tasks are assigned to core 01. In this case, the control unit 11 refers to the allocation data 121 and confirms the priority of each task. Then, the control unit 11 confirms whether or not the priority of the task Z can be higher than that of the other tasks.

When the priority of task Z is the highest, the task has a higher priority than task Z, and there is no task whose priority can be lower than that of task Z, the task Z has a higher priority. It is difficult to raise the priority over other tasks. Therefore, in such a case, the control unit 11 determines that it is impossible to accelerate the arithmetic processing by this second method.

On the other hand, when there is room to raise the priority of task Z higher than that of other tasks, the control unit 11 determines that it is possible to accelerate the arithmetic processing by this second method, and determines that the priority of task Z is high. Is higher than other tasks. In the example of FIG. 12B, the priority of task Z is "low" and the priority of other tasks is "medium" in the state before acceleration. That is, the priority of task Z is set lower than the priority of other tasks. Therefore, the control unit 11 may change the priority of the task Z so as to be higher than the priority of the other tasks.

In the example of FIG. 12B, the control unit 11 does not change the priority of the other task, but changes the priority of the task Z from "low" to "high", thereby changing the priority of the task Z to the other task. It is higher than the priority of. However, the method of increasing the priority of task Z does not have to be limited to such an example. The control unit 11 may change the priority of the task Z and may change the priority of the other tasks to be lower without changing the priority of the task Z. By these methods, the control unit 11 can give the priority of the task Z relatively higher than that of other tasks. As a result, the arithmetic processing of the prediction model 5 can be prioritized, so that the control unit 11 can accelerate the arithmetic processing of the prediction model 5.

(3) Third Method Next, a third method for accelerating the arithmetic processing of the prediction model 5 will be described with reference to FIG. 12C. FIG. 12C schematically illustrates a third method of accelerating the arithmetic processing of the prediction model 5.

In the third method, it is assumed that the CPU, which is a hardware processor, is composed of a plurality of cores. In this third method, for example, in step S101 or the like, the first core among the plurality of cores is assigned to the arithmetic processing of the prediction model 5. On the other hand, before the acceleration processing in step S106, it is assumed that there is a second core among the plurality of cores, which is not assigned to the arithmetic processing of the prediction model 5. There is. In the example of FIG. 12C, the core 01 is an example of the first core, and the core 02 is an example of the second core.

In this state, the control unit 11 accelerates the arithmetic processing of the prediction model 5 by further allocating the second core that was not assigned to the arithmetic processing of the prediction model 5 to the arithmetic processing of the prediction model 5. That is, the control unit 11 accelerates the arithmetic processing of the prediction model 5 by increasing the number of cores assigned to the arithmetic processing of the prediction model 5. The number of the first core and the number of the second core may be one or plural.

In the example of FIG. 12C, the arithmetic processing (task Z) of the prediction model 5 is assigned to the core 01. On the other hand, the arithmetic processing of the prediction model 5 is not assigned to the core 02, and other tasks are assigned. There is a free area in the core 02. The control unit 11 refers to the allocation data 121 and confirms whether or not there is a second core in which the task Z is not allocated and the free area exists.

If such a second core does not exist, the control unit 11 determines that it is impossible to accelerate the arithmetic processing by this third method. On the other hand, when the second core exists, the control unit 11 determines that the arithmetic processing by the third method can be accelerated, and at least a part of the free area of the second core. Allocate space to task Z. The amount of free space allocated to task Z may be appropriately determined according to the embodiment. As a result, the computational resources allocated to the arithmetic processing of the prediction model 5 can be increased, so that the control unit 11 can accelerate the arithmetic processing of the prediction model 5.

(4) Fourth Method Next, a fourth method for accelerating the arithmetic processing of the prediction model 5 will be described with reference to FIG. 12D. FIG. 12D schematically illustrates a fourth method of accelerating the arithmetic processing of the prediction model 5.

In the fourth method, it is assumed that the CPU, which is a hardware processor, is composed of a plurality of cores. In this fourth method, for example, in step S101 or the like, a third core out of a plurality of cores is assigned to the arithmetic processing of the prediction model 5. On the other hand, before the acceleration processing of this step S106, the plurality of cores are the fourth cores that are not assigned to the arithmetic processing of the prediction model 5, and are allocated to the arithmetic processing in the third core. It is assumed to contain a fourth core with more free space than the rate. In the example of FIG. 12D, the core 01 is an example of the third core, and the core 02 is an example of the fourth core.

In this state, the control unit 11 replaces the allocation for the arithmetic processing of the prediction model 5 from the third core to the fourth core, and the utilization rate is larger than the utilization rate allocated for the arithmetic processing in the third core. Is allocated to the arithmetic processing in the fourth core, thereby accelerating the arithmetic processing of the prediction model 5. That is, the control unit 11 calculates the prediction model 5 by changing the core assigned to the task Z from the original core to which the task Z is assigned to a new core capable of allocating a larger utilization rate to the task Z. Accelerate processing. The number of the third core and the number of the fourth core may be one or a plurality, respectively.

In the example of FIG. 12D, the arithmetic processing (task Z) of the prediction model 5 is assigned to the core 01. On the other hand, the arithmetic processing of the prediction model 5 is not assigned to the core 02, and other tasks are assigned. The core 02 has a free area (“unused” area in the figure) that is larger than the utilization rate allocated to the task Z in the core 01. The control unit 11 refers to the allocation data 121 and confirms whether or not there is a fourth core in which the task Z is not allocated and there is a free area larger than the utilization rate of the task Z in the core 01. do.

If such a fourth core does not exist, the control unit 11 determines that it is impossible to accelerate the arithmetic processing by this fourth method. On the other hand, when the fourth core is present, the control unit 11 determines that it is possible to accelerate the arithmetic processing by this fourth method. Then, the control unit 11 allocates at least a part of the free area of the fourth core to the task Z, which is larger than the utilization rate allocated to the task Z in the core 01. At this time, as shown in FIG. 12D, the control unit 11 may delete the assignment of task Z in the core 01 (third core). As a result, the computational resources allocated to the arithmetic processing of the prediction model 5 can be increased, so that the control unit 11 can accelerate the arithmetic processing of the prediction model 5.

In the example of FIG. 12D, tasks other than task Z are assigned to the other areas of the core 01. Therefore, in the core 01, there is no room for increasing the utilization rate of the task Z by the first method. The control unit 11 may adopt the fourth method when there is no room for adopting the first method in the third core as described above. On the other hand, if there is room for increasing the utilization rate of the task Z in the third core, the control unit 11 may adopt the first method.

(5) Others By adopting at least one of the first to fourth methods, the control unit 11 can accelerate the arithmetic processing of the prediction model 5. When the arithmetic processing of the prediction model 5 is accelerated by at least one of the first to fourth methods, the control unit 11 proceeds to the next step S107.

However, the method of accelerating the arithmetic processing of the prediction model 5 does not have to be limited to these examples. For example, it is assumed that the virtual memory of the control device 1 is composed of a cache memory (not shown), a RAM of the control unit 11, and a storage unit 12. In addition, it is assumed that the cache memory, the RAM of the control unit 11, and the storage unit 12 are closer to the CPU in this order. In this case, the control unit 11 may control the allocation of the memory so that the specific weight of the memory closer to the CPU is higher in the virtual memory allocated to the arithmetic processing of the prediction model 5. As a result, the control unit 11 can increase the access speed to the data used for the arithmetic processing, so that the arithmetic processing of the prediction model 5 can be accelerated.

In step S105, if none of the first to fourth methods can be adopted, the control unit 11 may determine that the arithmetic processing of the prediction model 5 cannot be accelerated. On the other hand, if at least one of the first to fourth methods can be adopted, the control unit 11 may determine that the arithmetic processing of the prediction model 5 can be accelerated.

(Step S107)
Returning to FIG. 9, in step S107, the control unit 11 operates as the prediction calculation unit 112, and determines whether or not the calculation processing of the prediction model 5 is completed. The method of determining whether or not the arithmetic processing of the prediction model 5 is completed may be appropriately determined according to the type of the model adopted in the prediction model 5.

In the present embodiment, since the prediction model 5 is composed of a decision tree, the control unit 11 determines whether or not the search process in step S102 has reached any leaf node of the decision tree. When the search process has reached any of the leaf nodes, the control unit 11 determines that the arithmetic process of the prediction model 5 has been completed, and proceeds to the next step S108. On the other hand, when the search process has not reached any leaf node, that is, when the search process has progressed only to the intermediate node, the control unit 11 states that the arithmetic processing of the prediction model 5 has not been completed. The determination is made, and the process is repeated from step S102.

When it is determined in step S104 that the determination of the command value is in time for the control timing, the control unit 11 completes the arithmetic processing of the prediction model 5 by repeating the series of processes of steps S102 to S104 and S107. Continue until you do. As a result, the control unit 11 can acquire the output value from the prediction model 5 as the final result of the arithmetic processing. In the present embodiment, when the search process of the decision tree is completed, the control unit 11 can acquire the value associated with the leaf node reached in the search process as the final result.

Further, after accelerating the arithmetic processing of the prediction model 5 in step S106, if it is determined that the arithmetic processing of the prediction model 5 is not completed, the control unit 11 executes steps S102 to S104 again. When the acceleration in step S106 is sufficient, it is determined in step S104 that the command value is determined in time for the control timing. As a result, the control unit 11 can continue until the arithmetic processing of the prediction model 5 is completed.

On the other hand, if the acceleration in step S106 is insufficient, in step S104 that is executed again, it is determined again that the determination of the command value is not in time for the control timing. In this case, as long as it is determined in step S105 that acceleration is possible, the control unit 11 executes step S106 until it is determined that the command value is determined in time for the control timing. It is possible to accelerate the arithmetic processing of. Then, by sufficiently accelerating the arithmetic processing of the prediction model 5, the control unit 11 can complete the arithmetic processing of the prediction model 5 so that the determination of the command value is in time for the control timing.

On the other hand, in step S104, when it is determined that the determination of the command value is not in time for the control timing, and in step S105, it is determined that the arithmetic processing of the prediction model 5 cannot be accelerated, the control unit. 11 proceeds to step S111 as described above. When it is determined again in step S104 executed after step S106 that the determination of the command value is not in time for the control timing, and in the subsequent step S105, it is determined that the arithmetic processing of the prediction model 5 cannot be accelerated. Is the same. As a result, the control unit 11 omits a series of processes from step S107 to step S109 described later, stops the arithmetic processing of the prediction model 5 as described later, and controls the operation of the production apparatus 3 by a method other than the prediction control. do.

(Step S108)
In step S108, the control unit 11 operates as the operation control unit 116, and determines the command value to the production device 3 based on the output acquired from the prediction model 5. The method of determining the command value may be appropriately determined according to the output format of the prediction model 5.

In the present embodiment, as described above, each leaf node of the decision tree constituting the prediction model 5 has a reference value 60 of the command value as an output value corresponding to the prediction result of the command value adapted to the production of the product 41. The correction value 63 for is associated with. Therefore, when the search process of the decision tree is completed, the control unit 11 can acquire the correction value 63 associated with the reached leaf node as the output from the prediction model 5. Therefore, the control unit 11 can determine the command value to the production device 3 by correcting the reference value 60 with the acquired correction value 63. When the command value is determined, the control unit 11 proceeds to the next step S109.

(Step S109)
In step S109, the control unit 11 operates as the operation control unit 116, and controls the operation of the production device 3 based on the determined command value. The method of controlling the operation of the production apparatus 3 based on the command value may be appropriately selected according to the format of the command value.

In the present embodiment, the production apparatus 3 is a press machine and includes a servo driver 31 that drives the upper die 32. Therefore, the command value may indicate the number of pulses that defines the driving amount of the servomotor. In this case, the control unit 11 transmits a command value to the servo driver 31 of the production device 3 via the external interface 14. The servo driver 31 drives the servomotor based on the command value received from the control device 1. As a result, the control unit 11 can control the operation of the production device 3 based on the determined command value. When the operation of the production device 3 is controlled, the control unit 11 ends the process related to this operation example.

The format of the command value does not have to be limited to such an example. The command value may be expressed by an intermediate index such as a driving amount of the servomotor and a moving amount of the upper mold 32. In this case, the control unit 11 may transmit the command value expressed by the intermediate index as it is to the production apparatus 3, or convert the command value expressed by the intermediate index into a directly usable format such as the number of pulses. Then, the changed command value may be transmitted to the production apparatus 3.

(Step S111)
As described above, step S111 is executed when it is determined that the determination of the command value is not in time for the control timing and it is determined that the arithmetic processing of the prediction model 5 cannot be accelerated. That is, step S111 is executed when the arithmetic processing of the prediction model 5 cannot be accelerated so that the determination of the command value is in time for the control timing. In this step S111, the control unit 11 operates as the prediction calculation unit 112 and stops the calculation processing of the prediction model 5. When the arithmetic processing of the prediction model 5 is stopped, the control unit 11 proceeds to the next step S112.

(Step S112)
In step S112, the control unit 11 operates as the operation control unit 116, and designates a set value given in advance as a command value.

The set value given in advance may be appropriately determined according to the embodiment. The set value given in advance may be, for example, a reference value 60 of the command value or a past command value. When a series of processes for controlling the operation of the production device 3 according to this operation example are repeatedly executed, the past command value may be, for example, the previous value when the operation of the production device 3 is controlled immediately before. good.

This previous value may be a command value when the operation of the production apparatus 3 is controlled immediately before, regardless of the input data acquired in step S100. In this case, the control unit 11 specifies the same command value twice in succession, immediately before and this time.

Further, the previous value may be a command value when the operation of the production apparatus 3 is controlled immediately before in the case where the input data acquired in step S100 is the same or similar. The control unit 11 determines whether or not the difference between the input data at the time of past control and the input data acquired this time is equal to or less than the threshold value, so that the input data in this case and the past case Can be determined whether or not they are similar.

The preset value given in advance may be held in the RAM of the storage unit 12, the control unit 11, or the like. The control unit 11 can specify the command value to the production device 3 by referring to the set value held in the storage unit 12, the RAM, or the like. When the command value is specified, the control unit 11 proceeds to the next step S113.

(Step S113)
In step S113, the control unit 11 operates as the operation control unit 116 and controls the operation of the production device 3 based on the designated command value. This step S113 is the same as the above step S109. When the operation of the production device 3 is controlled, the control unit 11 ends the process related to this operation example.

(After the end)
As described above, the control unit 11 ends a series of processes for controlling the operation of the production apparatus 3 according to this operation example. The control unit 11 can continuously control the operation of the production device 3 by repeatedly executing this series of processes.

[Learning device]
Next, an operation example of the learning device 2 will be described with reference to FIG. FIG. 13 is a flowchart illustrating an example of the processing procedure of the learning device 2. The processing procedure related to machine learning of the prediction model 5 described below is an example of a learning method by a computer. However, the processing procedure described below is only an example, and each processing may be changed as much as possible. Further, with respect to the processing procedure described below, steps can be omitted, replaced, and added as appropriate according to the embodiment.

(Step S201)
In step S201, the control unit 21 operates as the learning data acquisition unit 211 to acquire the learning data used for machine learning of the prediction model 5. The format of the training data may be appropriately determined according to the type of the prediction model 5.

In this embodiment, the prediction model 5 is composed of a decision tree. Further, the decision tree is constructed so as to output a correction value 63 with respect to the reference value 60 of the command value in response to the input of the feature amount 61 of the work 40 and the attribute value 62 of the environment in which the product 41 is produced. Therefore, the control unit 21 acquires a learning data set 221 composed of a combination of the feature amount 2211 of the work, the attribute value 2212 of the environment in which the product is produced, and the correction value 2213 adapted to the attribute value 2212.

The learning data set 221 may be appropriately generated according to the embodiment. For example, the feature amount 2211 of the work and the attribute value 2212 of the environment in which the product is produced are acquired under various conditions. A known sensor may be used to acquire the feature amount 2211 and the attribute value 2212. For example, when acquiring the temperature as the attribute value 2212, a temperature sensor may be used. Then, the obtained feature amount 2211 and the attribute value 2212 are combined with the correction value 2213 for obtaining an appropriate command value under this condition. As a result, the training data set 221 can be generated.

The generation of the learning data set 221 may be performed by the learning device 2. In this case, the control unit 21 may generate the learning data set 221 according to the operation of the input device 24 by the operator. Further, the control unit 21 may automatically generate the learning data set 221 by the processing of the learning program 82. By executing this generation process, in this step S201, the control unit 21 can acquire the learning data set 221.

Further, the learning data set 221 may be generated by an information processing device other than the learning device 2. In other information processing devices, the learning data set 221 may be manually generated by the operator or automatically generated by the processing of the program. In this case, in this step S201, the control unit 21 can acquire the learning data set 221 generated by another information processing device via the network, the storage medium 92, or the like.

The number of learning data sets 221 acquired in this step S201 may be appropriately determined according to the embodiment, and may be appropriately determined to the extent that machine learning of the decision tree can be performed, for example. As a result, when the learning data set 221 is acquired, the control unit 21 proceeds to the next step S202.

(Step S202)
In step S202, the control unit 21 operates as the learning processing unit 212 and constructs the trained prediction model 5 by performing machine learning using the acquired learning data set 221.

In the present embodiment, when the feature amount 2211 and the attribute value 2212 are input, the control unit 21 can search to reach the leaf node of the class corresponding to the correction value 2213 associated with the input feature amount 2211 and the attribute value 2212. Build a decision tree. As a learning method of this decision tree, CLS (Concept Learning System), ID3 (Iterative Dichotomiser 3), C4.5 or the like may be used. As a result, the control unit 21 can build the trained prediction model 5.

(Step S203)
In step S203, the control unit 21 operates as a learning processing unit 212 and stores information indicating the configuration of the decision tree (learned prediction model 5) constructed by machine learning and each branching condition as learning result data 225. Store in 22. As a result, the control unit 21 ends the learning process of the prediction model 5 according to this operation example.

The control unit 21 may transfer the created learning result data 225 to the control device 1 after the process of step S203 is completed. Further, the control unit 21 may periodically update the learning result data 225 by periodically executing the learning processes of steps S201 to S203. Then, the control unit 21 may periodically update the learning result data 225 held by the control device 1 by transferring the created learning result data 225 to the control device 1 each time the learning process is executed. Further, for example, the control unit 21 may store the created learning result data 225 in a data server such as NAS (Network Attached Storage). In this case, the control device 1 may acquire the learning result data 225 from this data server. Further, the learning result data 225 created by the learning device 2 may be incorporated in the control device 1 in advance.

[feature]
As described above, in the present embodiment, when the command value to the production apparatus 3 is predicted by using the prediction model 5, the remaining processing time is calculated in steps S103 and S104, and the remaining processing time is calculated based on the calculated remaining processing time. Therefore, it is determined whether or not the determination of the command value by the prediction model 5 is in time for the control timing. Thereby, it is possible to specify whether or not the arithmetic processing related to the prediction of the command value using the prediction model 5 is in time for the timing of controlling the operation of the production apparatus 3 by using the predicted command value. That is, it is possible to specify whether or not the determination of the command value in step S108 is completed by the scheduled control time for controlling the operation of the production apparatus 3. Therefore, according to the present embodiment, it is possible to appropriately manage the calculation time of the prediction control of the production apparatus 3.

In particular, the control device 1 can process a plurality of tasks including tasks other than the arithmetic processing of the prediction model 5 (for example, FIGS. 12B to 12D). In this case, for example, due to the occurrence of an interrupt task, the computational resources (hardware processor, memory, etc.) in the control device 1 may not always be available under the same conditions. That is, the state of the computational resources in the control device 1 changes from moment to moment, and the computational resources that can be used for the computational processing of the prediction model 5 may change accordingly. Therefore, the processing time required for the arithmetic processing of the prediction model 5 is not always constant. Therefore, even if it is determined that the arithmetic processing of the prediction model 5 can be processed in time for the control timing based on the processing time specified in advance by simulation or the like, the control device 1 may perform the operation. It is not always possible to complete the arithmetic processing of the prediction model 5 in time for the control timing of the production apparatus 3. On the other hand, in the present embodiment, the control device 1 confirms the status of the computational resources (specifically, the progress of the arithmetic processing) at the time when the arithmetic processing of the prediction model 5 is executed in the step S103. Based on this, the remaining processing time required for the arithmetic processing is calculated. As a result, even if the state of the computational resources changes from moment to moment, the control device 1 can appropriately determine in step S104 whether or not the arithmetic processing of the prediction model 5 is in time for the control timing. Therefore, according to the present embodiment, it is possible to appropriately manage the calculation time of the prediction control of the production apparatus 3.

In addition, in the present embodiment, when it is determined in step S104 that the determination of the command value is in time for the control timing, the control device 1 repeats the series of processes of steps S102 to S104 and S107, and then steps S108 and S108. The process of S109 is executed. As a result, the control device 1 can appropriately execute the predictive control as it is. On the other hand, if it is determined in step S104 that the determination of the command value is not in time for the control timing, and in step S105 it is determined that the arithmetic processing of the prediction model 5 cannot be accelerated, the control device 1 determines. The arithmetic processing of the prediction model 5 is stopped. As a result, the control device 1 can stop the predictive control and appropriately control the operation of the production device 3 by another method (in the present embodiment, the set value given in advance is used). Therefore, according to the present embodiment, it is possible to prevent the process of controlling the operation of the production apparatus 3 from being delayed because the determination of the command value by the prediction model 5 is not in time for the control timing.

Further, in the present embodiment, when it is determined in step S104 that the determination of the command value is in time for the control timing, the arithmetic processing of the prediction model 5 can be accelerated by executing the processing of step S106. As a result, the control device 1 can make the determination of the command value by the prediction model 5 in time for the control timing. Therefore, according to the present embodiment, the possibility of terminating the prediction control by the prediction model 5 can be reduced, and the prediction control can be appropriately implemented.

§4 Modifications Although the embodiments of the present invention have been described in detail above, the above description is merely an example of the present invention in all respects. Needless to say, various improvements and modifications can be made without departing from the scope of the present invention. For example, the following changes can be made. In the following, the same reference numerals will be used for the same components as those in the above embodiment, and the same points as in the above embodiment will be omitted as appropriate. The following modifications can be combined as appropriate.

<4.1>
In the above embodiment, when it is determined in step S104 that the determination of the command value by the prediction model 5 is not in time for the control timing, the control unit 11 accelerates the arithmetic processing of the prediction model 5 in steps S105 and S106. The process of accelerating the arithmetic process of the prediction model 5 may be omitted. When the acceleration processing is omitted, the steps S105 and S106 may be omitted, and the acceleration processing unit 115 may be omitted from the software configuration of the control device 1.

<4.2>
In the above embodiment, the prediction model 5 is configured to output a correction value 63 with respect to the reference value 60 of the command value as a result of prediction of the command value adapted to the production of the product 41. However, the output format of the prediction model 5 does not have to be limited to such an example, and may be appropriately determined according to the embodiment. For example, the prediction model 5 may be configured to output the command value itself.

<4.3>
In the above embodiment, both the feature amount 61 of the work 40 and the attribute value 62 of the environment for producing the product 41 are used for the input of the prediction model 5. However, the input of the prediction model 5 does not have to be limited to such an example. One of the feature amount 61 of the work 40 and the attribute value 62 of the environment in which the product 41 is produced may be omitted. That is, the prediction model 5 may be constructed so as to predict a command value adapted to the production of the product 41 for at least one of the feature amount 61 of the work 40 and the attribute value 62 of the environment in which the product 41 is produced. .. Further, the input of the prediction model 5 is data on factors that may affect the production process of the product 41 by the production apparatus 3, other than the feature amount 61 of the work 40 and the attribute value 62 of the environment in which the product 41 is produced. Data may be used.

<4.4>
In the above embodiment, the prediction model 5 is composed of a decision tree. However, the configuration of the prediction model 5 is not limited to such an example as long as the command value to the production apparatus 3 at a time point (future time point) before the time point at which the prediction process is executed can be predicted. Often, it may be appropriately selected according to the embodiment. For the prediction model 5, for example, a learning model other than a decision tree such as a neural network or a support vector machine may be used. Further, as the prediction model 5, a model other than the learning model (for example, a predetermined function) may be used.

1 ... Control device,
11 ... Control unit, 12 ... Storage unit, 13 ... Communication interface,
14 ... External interface,
15 ... input device, 16 ... output device, 17 ... drive,
111 ... Data acquisition unit, 112 ... Prediction calculation unit,
113 ... Residual processing time calculation unit, 114 ... Delay determination unit,
115 ... Acceleration processing unit, 116 ... Motion control unit,
81 ... Control program, 121 ... Allocation data,
91 ... Storage medium,
2 ... Learning device,
21 ... Control unit, 22 ... Storage unit, 23 ... Communication interface,
24 ... Input device, 25 ... Output device, 26 ... Drive,
211 ... Learning data acquisition unit, 212 ... Learning processing unit,
82 ... Learning program, 221 ... Learning dataset,
225 ... Learning result data,
92 ... Storage medium,
3 ... Production equipment (press machine),
31 ... Servo driver, 32 ... Upper mold, 33 ... Lower mold,
40 ... work, 41 ... product,
5 ... Prediction model (decision tree),
60 ... Reference value (of command value),
61 ... feature amount, 62 ... attribute value, 63 ... correction value

Claims (11)

With a hardware processor
A memory that holds a program executed by the hardware processor, and
It is a control device equipped with
The hardware processor is based on a series of instructions contained in the program.
A step of starting arithmetic processing of a prediction model that is a command value from a work to a production device that produces a product and is constructed so as to predict a command value adapted to the production of the product by the production device.
A step of calculating the remaining processing time until the calculation processing of the prediction model is completed after starting the calculation processing of the prediction model, and
A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the calculated remaining processing time is in time for the control timing for controlling the operation of the production by the production apparatus.
When it is determined that the determination of the command value is in time for the control timing, the output is acquired from the prediction model by continuing until the calculation processing of the prediction model is completed, and the output is based on the output acquired from the prediction model. Then, the step of determining the command value to the production device and
A step of controlling the operation of the production apparatus based on the determined command value, and
When it is determined that the determination of the command value is not in time for the control timing, the step of accelerating the arithmetic processing of the prediction model and
The execution,
The hardware processor is composed of a plurality of cores.
One of the plurality of cores is assigned to the arithmetic processing of the prediction model.
The plurality of cores are other cores that are not assigned to the arithmetic processing of the prediction model, and other cores having a free capacity larger than the utilization rate allocated to the arithmetic processing in the one core. Includes
The hardware processor replaces the allocation for the arithmetic processing of the prediction model from the one core to the other core, and the utilization rate higher than the utilization rate allocated to the arithmetic processing in the one core is said. By allocating to the arithmetic processing in another core, the arithmetic processing of the prediction model is accelerated.
Control device. With a hardware processor
A memory that holds a program executed by the hardware processor, and
It is a control device equipped with
The hardware processor is based on a series of instructions contained in the program.
A step of starting arithmetic processing of a prediction model that is a command value from a work to a production device that produces a product and is constructed so as to predict a command value adapted to the production of the product by the production device.
A step of calculating the remaining processing time until the calculation processing of the prediction model is completed after starting the calculation processing of the prediction model, and
A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the calculated remaining processing time is in time for the control timing for controlling the operation of the production by the production apparatus.
When it is determined that the determination of the command value is in time for the control timing, the output is acquired from the prediction model by continuing until the calculation processing of the prediction model is completed, and the output is based on the output acquired from the prediction model. Then, the step of determining the command value to the production device and
A step of controlling the operation of the production apparatus based on the determined command value, and
When it is determined that the determination of the command value is not in time for the control timing, the step of accelerating the arithmetic processing of the prediction model and
And
After executing the step of accelerating the arithmetic processing of the prediction model, the hardware processor performs a step of calculating the remaining processing time and a step of determining whether or not the determination of the command value is in time for the control timing. Run again,
Control apparatus. With a hardware processor
A memory that holds a program executed by the hardware processor, and
It is a control device equipped with
The hardware processor is based on a series of instructions contained in the program.
A step of starting arithmetic processing of a prediction model that is a command value from a work to a production device that produces a product and is constructed so as to predict a command value adapted to the production of the product by the production device.
A step of calculating the remaining processing time until the calculation processing of the prediction model is completed after starting the calculation processing of the prediction model, and
A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the calculated remaining processing time is in time for the control timing for controlling the operation of the production by the production apparatus.
When it is determined that the determination of the command value is in time for the control timing, the output is acquired from the prediction model by continuing until the calculation processing of the prediction model is completed, and the output is based on the output acquired from the prediction model. Then, the step of determining the command value to the production device and
A step of controlling the operation of the production apparatus based on the determined command value, and
When it is determined that the determination of the command value is not in time for the control timing, the step of accelerating the arithmetic processing of the prediction model and
And
The hardware processor
It is determined whether or not the arithmetic processing of the prediction model can be accelerated.
When it is determined that the calculation processing of the prediction model cannot be accelerated, the calculation processing of the prediction model is stopped, a set value given in advance is specified as a command value, and the operation is described based on the specified command value. Control the operation of production equipment,
Control apparatus. The prediction model is configured to output a correction value with respect to a reference value of the command value as a result of prediction of the command value adapted to the production of the product.
The control device according to any one of claims 1 to 3. The prediction model is constructed so as to predict the command value adapted to the production of the product for at least one input of the feature amount of the work and the attribute value of the environment in which the product is produced.
The control device according to any one of claims 1 to 4. The computer 's hardware processor
A step of starting arithmetic processing of a prediction model that is a command value from a work to a production device that produces a product and is constructed so as to predict a command value adapted to the production of the product by the production device.
A step of calculating the remaining processing time until the calculation processing of the prediction model is completed after starting the calculation processing of the prediction model, and
A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the calculated remaining processing time is in time for the control timing for controlling the operation of the production by the production apparatus.
When it is determined that the determination of the command value is in time for the control timing, the output is acquired from the prediction model by continuing until the calculation processing of the prediction model is completed, and the output is based on the output acquired from the prediction model. Then, the step of determining the command value to the production device and
A step of controlling the operation of the production apparatus based on the determined command value, and
When it is determined that the determination of the command value is not in time for the control timing, the step of accelerating the arithmetic processing of the prediction model and
The execution,
The hardware processor is composed of a plurality of cores.
One of the plurality of cores is assigned to the arithmetic processing of the prediction model.
The plurality of cores are other cores that are not assigned to the arithmetic processing of the prediction model, and other cores having a free capacity larger than the utilization rate allocated to the arithmetic processing in the one core. Includes
The hardware processor replaces the allocation for the arithmetic processing of the prediction model from the one core to the other core, and the utilization rate higher than the utilization rate allocated to the arithmetic processing in the one core is said. By allocating to the arithmetic processing in another core, the arithmetic processing of the prediction model is accelerated.
Control method. The computer's hardware processor
A step of starting arithmetic processing of a prediction model that is a command value from a work to a production device that produces a product and is constructed so as to predict a command value adapted to the production of the product by the production device.
A step of calculating the remaining processing time until the calculation processing of the prediction model is completed after starting the calculation processing of the prediction model, and
A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the calculated remaining processing time is in time for the control timing for controlling the operation of the production by the production apparatus.
When it is determined that the determination of the command value is in time for the control timing, the output is acquired from the prediction model by continuing until the calculation processing of the prediction model is completed, and the output is based on the output acquired from the prediction model. Then, the step of determining the command value to the production device and
A step of controlling the operation of the production apparatus based on the determined command value, and
When it is determined that the determination of the command value is not in time for the control timing, the step of accelerating the arithmetic processing of the prediction model and
And
After executing the step of accelerating the arithmetic processing of the prediction model, the hardware processor performs a step of calculating the remaining processing time and a step of determining whether or not the determination of the command value is in time for the control timing. Run again,
Control method. The computer's hardware processor
A step of starting arithmetic processing of a prediction model that is a command value from a work to a production device that produces a product and is constructed so as to predict a command value adapted to the production of the product by the production device.
A step of calculating the remaining processing time until the calculation processing of the prediction model is completed after starting the calculation processing of the prediction model, and
A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the calculated remaining processing time is in time for the control timing for controlling the operation of the production by the production apparatus.
When it is determined that the determination of the command value is in time for the control timing, the output is acquired from the prediction model by continuing until the calculation processing of the prediction model is completed, and the output is based on the output acquired from the prediction model. Then, the step of determining the command value to the production device and
A step of controlling the operation of the production apparatus based on the determined command value, and
When it is determined that the determination of the command value is not in time for the control timing, the step of accelerating the arithmetic processing of the prediction model and
And
The hardware processor
It is determined whether or not the arithmetic processing of the prediction model can be accelerated.
When it is determined that the calculation processing of the prediction model cannot be accelerated, the calculation processing of the prediction model is stopped, a set value given in advance is specified as a command value, and the operation is described based on the specified command value. Control the operation of production equipment,
Control method. For computer hardware processors
A step of starting arithmetic processing of a prediction model that is a command value from a work to a production device that produces a product and is constructed so as to predict a command value adapted to the production of the product by the production device.
A step of calculating the remaining processing time until the calculation processing of the prediction model is completed after starting the calculation processing of the prediction model, and
A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the calculated remaining processing time is in time for the control timing for controlling the operation of the production by the production apparatus.
When it is determined that the determination of the command value is in time for the control timing, the output is acquired from the prediction model by continuing until the calculation processing of the prediction model is completed, and the output is based on the output acquired from the prediction model. Then, the step of determining the command value to the production device and
A step of controlling the operation of the production apparatus based on the determined command value, and
When it is determined that the determination of the command value is not in time for the control timing, the step of accelerating the arithmetic processing of the prediction model and
Is a control program for executing
The hardware processor is composed of a plurality of cores.
One of the plurality of cores is assigned to the arithmetic processing of the prediction model.
The plurality of cores are other cores that are not assigned to the arithmetic processing of the prediction model, and other cores having a free capacity larger than the utilization rate allocated to the arithmetic processing in the one core. Includes
The hardware processor is made to replace the allocation for the arithmetic processing of the prediction model from the one core to the other core, and the utilization rate larger than the utilization rate allocated to the arithmetic processing in the one core is said. By allocating to the arithmetic processing in another core, the arithmetic processing of the prediction model is accelerated.
Control program. For computer hardware processors
A step of starting arithmetic processing of a prediction model that is a command value from a work to a production device that produces a product and is constructed so as to predict a command value adapted to the production of the product by the production device.
A step of calculating the remaining processing time until the calculation processing of the prediction model is completed after starting the calculation processing of the prediction model, and
A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the calculated remaining processing time is in time for the control timing for controlling the operation of the production by the production apparatus.
When it is determined that the determination of the command value is in time for the control timing, the output is acquired from the prediction model by continuing until the calculation processing of the prediction model is completed, and the output is based on the output acquired from the prediction model. Then, the step of determining the command value to the production device and
A step of controlling the operation of the production apparatus based on the determined command value, and
When it is determined that the determination of the command value is not in time for the control timing, the step of accelerating the arithmetic processing of the prediction model and
Is a control program for executing
After executing the step of accelerating the arithmetic processing of the prediction model, the hardware processor calculates the remaining processing time, and determines whether or not the determination of the command value is in time for the control timing. To run again,
Control program. For computer hardware processors
A step of starting arithmetic processing of a prediction model that is a command value from a work to a production device that produces a product and is constructed so as to predict a command value adapted to the production of the product by the production device.
A step of calculating the remaining processing time until the calculation processing of the prediction model is completed after starting the calculation processing of the prediction model, and
A step of determining whether or not the determination of the command value based on the output obtained from the prediction model based on the calculated remaining processing time is in time for the control timing for controlling the operation of the production by the production apparatus.
When it is determined that the determination of the command value is in time for the control timing, the output is acquired from the prediction model by continuing until the calculation processing of the prediction model is completed, and the output is based on the output acquired from the prediction model. Then, the step of determining the command value to the production device and
A step of controlling the operation of the production apparatus based on the determined command value, and
When it is determined that the determination of the command value is not in time for the control timing, the step of accelerating the arithmetic processing of the prediction model and
Is a control program for executing
To the hardware processor
It is determined whether or not the arithmetic processing of the prediction model can be accelerated.
When it is determined that the calculation processing of the prediction model cannot be accelerated, the calculation processing of the prediction model is stopped, a set value given in advance is specified as a command value, and the operation is described based on the specified command value. Control the operation of production equipment,
Control program.

Images