control device

Abstract

A control device acquires data related to a mechanical structure of an industrial machine, data related to a workpiece, and data related to an operation condition, generates machine learning data for a process of machine learning based on the acquired data. An instruction causes a process of machine learning for estimating data related to tension in a conveying section of the industrial machine to be executed based on the generated machine learning data. Also, based on the instruction, the process of machine learning for estimating the data related to the tension in the conveying section is executed. In this way, the control device can adjust the tension of the conveying section without a tension sensor according to a specified condition at the time of actual operation.

Description

Technical Field

[0001] This invention relates to a control device for industrial machinery that controls a generally belt-shaped conveyor section equipped with drive rollers to transport workpieces. Background Technology

[0002] Factories and other manufacturing sites are equipped with various industrial machines, such as machine tools, robots, and conveyor systems. Conveyor systems are used to transport workpieces on the factory production line. Figure 6 This is a diagram illustrating the schematic structure of a conveyor that uses drive rollers to transport workpieces.

[0003] The conveyor 300 includes multiple rollers 301 and a conveyor belt 302 mounted on the rollers 301. The conveyor belt 302, for example, a portion of the rollers 301, is driven to rotate by an electric motor (not shown) and conveyed in the conveying direction. A tension sensor 303 detects the tension generated in the conveyor belt 302 during conveying. During the operation of the conveyor 300, the tension of the conveyor belt 302 is adjusted by a tension adjustment mechanism 304, such as a tension roller, to prevent slippage between the rollers 301 and the conveyor belt 302, and to prevent excessive tension on the conveyor belt 302. Figure 6 In the illustrated conveyor 300, workpiece 305 is placed and conveyed on conveyor belt 302.

[0004] When a target tension Taim is set in the control device of the conveyor 300, the tension deviation ΔT is calculated as the difference between the target tension Taim and the measured tension TFBK detected by the tension sensor 303. Then, based on the target tension Taim and the tension deviation ΔT, the tension / torque conversion circuit 310 calculates the target torque Qaim. On the other hand, the mechanical loss torque QL, which is used to compensate for mechanical losses caused by mechanical changes over time (wear, etc.), and the acceleration / deceleration torque QF, which is used to compensate for the motor output voltage required for acceleration and deceleration, are calculated by the mechanical loss torque compensation circuit 320 and the acceleration / deceleration torque compensation circuit 330, respectively.

[0005] Thus, in conveying devices such as conveyors that transport workpieces, tension sensors or the like are used to detect the state of the conveying process, and the tension of the conveyor belt or the like is adjusted based on the detected state (e.g., Patent Document 1, etc.).

[0006] Existing technical documents

[0007] Patent documents

[0008] Patent Document 1: International Publication No. 2014 / 103886 Summary of the Invention

[0009] The problem that the invention aims to solve

[0010] Industrial machinery uses numerous sensors to monitor the driving status of various components. In such cases, there is a desire to reduce the overall cost of the industrial machinery by decreasing the number of sensors used. Furthermore, reducing the number of sensors also reduces malfunctions in industrial machinery caused by sensor failures.

[0011] Therefore, a technology is desired that can adjust the tension according to specified conditions without the need for a tension sensor during actual operation.

[0012] Methods for solving problems

[0013] The control device of the present invention uses a machine learning machine instead of a tension sensor to calculate the torque of each axis according to specified conditions, thereby solving the aforementioned problem. In this specification, in conveying devices such as belt conveyors, the generally belt-shaped component used to convey workpieces is referred to as the conveyor section. During the learning process of the machine learning machine, data related to the mechanical structure of the conveying machinery, data related to the workpiece to be conveyed, and data related to the operating conditions of the machinery are used as data representing the operating state of the machinery. The machine learning machine learns the correlation between these data and the tension value of the conveyor section detected by the tension sensor. Furthermore, during actual operation, the tension value of the conveyor section is estimated by the machine learning machine instead of the tension sensor, and the tension of the conveyor belt and other components in the conveying machinery is adjusted based on the estimated results.

[0014] Furthermore, one aspect of the present invention is a control device for industrial machinery that controls a generally belt-shaped conveyor section equipped with drive rollers to convey workpieces. This control device comprises: a data acquisition unit that acquires data related to the mechanical structure, workpiece, and operating conditions of the operating state of the industrial machinery; an acquisition data storage unit that stores the data acquired by the data acquisition unit related to the mechanical structure, workpiece, and operating conditions of the operating state of the industrial machinery; a machine learning data generation unit that generates machine learning data for machine learning processing based on the data stored in the acquisition data storage unit; a machine learning processing instruction unit that instructs, based on the data generated by the machine learning data generation unit, to perform machine learning processing for estimating data related to tension in the conveyor section; and a machine learning unit that, based on instructions from the machine learning processing instruction unit, performs machine learning processing for estimating data related to tension in the conveyor section.

[0015] Invention Effects

[0016] According to one aspect of the invention, a more appropriate inter-shaft tension can be output even without a tension sensor, thereby reducing the cost (including maintenance costs) of the tension sensor required in conventional tension control. Attached Figure Description

[0017] Figure 1 This is a schematic hardware structure diagram of a control device according to one embodiment.

[0018] Figure 2 This is a schematic block diagram illustrating the function of the control device in the first embodiment.

[0019] Figure 3 This is a schematic block diagram illustrating the function of the control device in the second embodiment.

[0020] Figure 4 This is a schematic block diagram illustrating the function of the control device in the third embodiment.

[0021] Figure 5 This is a diagram illustrating an example of industrial machinery used to transport workpieces on a thin film.

[0022] Figure 6 This is a diagram illustrating an example of existing conveying machinery. Detailed Implementation

[0023] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0024] Figure 1 This is a schematic hardware structure diagram showing the main parts of a control device according to one embodiment of the present invention. The control device 1 of the present invention has the function of controlling industrial machinery 3, such as conveyor machinery. In this embodiment, the structure of the control device 1 and the industrial machinery 3 during the machine learning stage of the machine learning machine 2 is shown.

[0025] The CPU 11 of the control device 1 in this embodiment is a processor that controls the entire control device 1. The CPU 11 reads the system program stored in the ROM 12 via the bus 22 and controls the control device 1 as a whole according to the system program. The RAM 13 temporarily stores temporary calculation data, display data, and various data input from the outside.

[0026] The non-volatile memory 14 is composed of, for example, a memory backed up by a battery (not shown), an SSD (Solid State Drive), etc., and maintains its storage state even when the power supply to the control device 1 is disconnected. The non-volatile memory 14 stores control programs and data read from the external device 72 via the interface 15, control programs and data input from the input device 71 via the interface 18, and control programs and data obtained from other devices such as the cloud computer 6 and cloud server 7 via the network 5. The data stored in the non-volatile memory 14 may include, for example, data related to the mechanical structure of the industrial machinery 3, data related to the conveyed workpiece, data related to the operating conditions of the machinery, tension value data of the conveyor belt, etc., detected by the tension sensor 4, and data related to various physical quantities detected by sensors (not shown) installed on other industrial machinery 3. The control programs and data stored in the non-volatile memory 14 can also be expanded in the RAM 13 during execution / use. Furthermore, various system programs, such as known analysis programs, are pre-written in the ROM 12.

[0027] One or more tension sensors 4 installed on the industrial machinery 3 detect the tension of the conveyor belt, etc., of the industrial machinery 3. The tension sensor 4 can be a force sensor that detects the reaction force when force is applied to the conveyor belt, or it can detect the tension of the conveyor belt in a non-contact manner by means of sound waves. The tension sensor 4 is necessary when the control device 1 operates in learning mode, but is not required when the control device 1 operates in actual operation mode.

[0028] Interface 15 is an interface for connecting the CPU 11 of the control device 1 to external devices 72 such as external storage media. Control programs and setting data used in the control of industrial machinery 3, for example, are read from the external device 72. Furthermore, control programs and setting data edited within the control device 1 can be stored via external device 72 on external storage media (not shown) such as CF cards and USB memories. The programmable logic controller (PLC) 16 executes a ladder logic program and outputs signals to and controls industrial machinery 3 and its peripheral devices (e.g., tool changing devices, actuators such as robots, tension sensors 4, temperature sensors, humidity sensors, etc., mounted on industrial machinery 3) via I / O unit 19. Additionally, it receives signals from various switches and peripheral devices on the control panel mounted on the main body of industrial machinery 3, performs necessary signal processing, and then transmits the signals to the CPU 11.

[0029] Interface 20 is used to connect the CPU 11 of control device 1 to a wired or wireless network 5. Network 5 can communicate using technologies such as RS-485 serial communication, Ethernet communication, optical communication, wireless LAN, Wi-Fi, and Bluetooth. Control devices for controlling other industrial machinery, a fog computer 6, a cloud server 7, and other higher-level management devices are connected to network 5, and data is exchanged between them and control device 1.

[0030] In the display device 70, data read into the memory and data obtained as a result of executing a program are output and displayed via the interface 17. In addition, the input device 71, which consists of a keyboard and an indicator device, transmits instructions and data based on the operator's operation to the CPU 11 via the interface 18.

[0031] The axis control circuit 30, used to drive the drive unit of the industrial machinery 3, receives movement commands from the CPU 11 and outputs these commands to the servo amplifier 40. The servo amplifier 40 receives these commands and drives the servo motors 50 of the industrial machinery 3. Each servo motor 50 has a built-in position / speed detector, and the position / speed feedback signals from these detectors are fed back to the axis control circuit 30 for position / speed feedback control. Furthermore, in... Figure 1 In the hardware structure diagram, only one of the axis control circuit 30, servo amplifier 40 and servo motor 50 is shown, but in reality, it only represents the number of drive units possessed by the industrial machinery 3 that is intended to be the object of control.

[0032] Interface 21 is used to connect CPU 11 to machine learning device 2. Machine learning device 2 includes a processor 201 for controlling the entire machine learning device, a ROM 202 storing system programs, RAM 203 for temporary storage during various machine learning processes, and non-volatile memory 204 for storing learning models, etc. Machine learning device 2 can observe data that can be obtained from control device 1 via interface 21 (e.g., data related to mechanical structure, data related to conveyed workpieces, data related to machine operating conditions, tension values ​​of conveyor belts, etc., detected by tension sensor 4, etc.). Furthermore, control device 1 obtains the processing results output from machine learning device 2 via interface 21, stores or displays the results, or sends them to other devices via network 5, etc. In addition, in Figure 1 In this system, the machine learning device 2 is built into the control device 1, but it can also be externally connected to the control device 1 via a predetermined interface.

[0033] Figure 2This is a schematic block diagram illustrating the functions of the control device 1 according to the first embodiment of the present invention. The functions of the control device 1 in this embodiment are described by... Figure 1 The CPU 11 of the control device 1 and the processor 201 of the machine learning device 2 shown execute the system program and control the actions of each part of the control device 1 and the machine learning device 2 to achieve this.

[0034] The control device 1 of this embodiment includes a control unit 100, a data acquisition unit 110, a machine learning data generation unit 120, a machine learning processing instruction unit 130, and a tension adjustment unit 140. Furthermore, the machine learning unit 205 constituting the machine learning machine 2 includes a learning unit 206 and an estimation unit 208. Additionally, the RAM 13 or non-volatile memory 14 of the control device 1 includes an acquisition data storage unit 190 for storing data acquired from the industrial machinery 3, and the RAM 203 or non-volatile memory 204 of the machine learning machine 2 includes a model storage unit 209 for storing the learning model generated by the learning unit 206.

[0035] The control unit 100 controls the operation of the industrial machinery 3 based on preset operating conditions and control programs. The control unit 100 includes, for example... Figure 6 The control unit 100 provides general functions required for controlling the operation of the industrial machine 3, as illustrated in the example. Based on the feed rate, acceleration, target tension, etc., specified by the operating conditions and control program, the control unit 100 controls the operation of the industrial machine 3 while referring to the measured speed and measured tension fed back from the industrial machine 3.

[0036] The data acquisition unit 110 acquires, for example, data Is related to the mechanical structure of the industrial machine 3, data Iw related to the conveyed workpiece, data Ic related to the machine's operating conditions, and tension value data Ot of the conveyor belt, etc., detected by the tension sensor 4 installed on the industrial machine 3, from the control unit 100. The data acquisition unit 110 may also acquire data related to warnings generated in the industrial machine 3 and input value data input by the operator via the input device 71. Furthermore, the data acquisition unit 110 may also acquire data related to the operation of the industrial machine 3 obtained and stored by external devices 72, fog computers 6, cloud servers 7, etc. The data acquired by the data acquisition unit 110 is at least data related to the operating state of the industrial machine 3 during operation.

[0037] The data Is related to the mechanical structure acquired by the data acquisition unit 110 includes, for example, the number of shafts of the industrial machine 3 (the number of driven rotors, etc.), the distance between each shaft, whether it is a gravity shaft, the direction of movement, the diameter of the rollers, and the type and material of the conveyor belt. Additionally, the workpiece-related data Iw includes the workpiece material, thickness, weight, and shape. Furthermore, the data Ic related to operating conditions includes feed speed, acceleration, total operating time, and total operating distance. The tension value data Ot is the tension value detected by the tension sensor 4 installed on the industrial machine 3 when acquiring data associated with each operating state. This tension value data Ot may include multiple tension values ​​measured by multiple tension sensors 4. These data are summarized at each moment of acquisition or detection and stored in the acquisition data storage unit 190.

[0038] The machine learning data generation unit 120 generates learning data for machine learning processing related to tension value adjustment performed by the machine learning unit 205, based on data stored in the data acquisition and storage unit 190. More specifically, when the control device 1 operates in learning mode, the machine learning data generation unit 120 generates learning data for learning processing performed by the machine learning unit 205, based on data stored in the data acquisition and storage unit 190. The learning data generated by the machine learning data generation unit 120 associates at least the tension value data Ot of the conveyor belt, etc., detected by the tension sensor 4, with the data Is related to the mechanical structure, the data Iw related to the workpiece, and the data Ic related to the operating conditions. The learning data generated by the machine learning data generation unit 120 may also include additional data, depending on the machine learning method of the machine learning unit 205.

[0039] On the other hand, when the control device 1 operates in the actual operating mode, the machine learning data generation unit 120 generates estimation data for estimation processing by the estimation unit 208 based on the data stored in the data acquisition and storage unit 190. The estimation data generated by the machine learning data generation unit 120 includes at least data Is related to the mechanical structure, data Iw related to the workpiece, and data Ic related to the operating conditions. The estimation data generated by the machine learning data generation unit 120 may also include additional data, depending on the machine learning method of the machine learning unit 205.

[0040] The machine learning processing instruction unit 130 instructs the machine learning unit 205 to perform machine learning processing related to the estimation of tension-related data based on the data generated by the machine learning data generation unit 120. More specifically, when the control device 1 operates in learning mode, the machine learning processing instruction unit 130 instructs the machine learning unit 205 to learn the correlation between data such as Is related to the mechanical structure, Iw related to the workpiece, and Ic related to the operating conditions, and the tension value data Ot of the conveyor belt, etc., detected by the tension sensor 4, and the tension value adjustment behavior data, based on the learning data generated by the machine learning data generation unit 120.

[0041] On the other hand, when the control device 1 operates in the actual operating mode, the machine learning processing instruction unit 130 instructs the machine learning unit 205 to estimate the tension value and tension adjustment behavior based on the estimation data generated by the machine learning data generation unit 120, such as the data Is related to the mechanical structure, the data Iw related to the workpiece, and the data Ic related to the operating conditions.

[0042] The tension adjustment unit 140 instructs the control unit 100 to perform actions to adjust the tension value based on tension-related data (tension value itself, tension adjustment behavior, etc.) estimated by the machine learning unit 205. For example, when the tension value is estimated by the machine learning unit 205, the tension adjustment unit 140 instructs the control unit 100 to adjust from the estimated tension value to the target tension value. Furthermore, when the tension adjustment unit 140 estimates the tension value adjustment behavior by the machine learning unit 205, it instructs the control unit 100 to perform the estimated adjustment behavior. However, the tension adjustment unit 140 may not be necessary when the control device 1 operates in learning mode.

[0043] The learning unit 206 of the machine learning unit 205 generates a model based on learning data included in the instructions received from the machine learning processing instruction unit 130. This model learns data related to the tension of the conveyor belt, etc., corresponding to the operating state of the industrial machinery 3. The model is then stored in the model storage unit 209 based on the learning data received from the machine learning processing instruction unit 130. More specifically, for example, during supervised learning, the learning unit 206 generates a model that learns the correlation between tension value data Ot of the conveyor belt, etc., detected by the tension sensor 4, and data such as Is (related to the mechanical structure), Iw (related to the workpiece), and Ic (related to the operating conditions), representing the operating state of the industrial machinery 3. Alternatively, for example, during reinforcement learning, the learning unit 206 generates a model that learns the correlation between the adjustment behavior of the tension value of the conveyor belt, etc., detected by the tension sensor 4, and data such as Is (related to the mechanical structure), Iw (related to the workpiece), and Ic (related to the operating conditions), representing the operating state of the industrial machinery 3.

[0044] The machine learning performed by Learning Department 206 can be supervised learning or reinforcement learning, both of which are well-known methods. The model generated by Learning Department 206, through machine learning, can infer data related to the tension of conveyor belts, etc. (the tension value itself or the adjustment behavior of the tension value) based on data Is related to the mechanical structure, Iw related to the workpiece, and Ic related to operating conditions. Examples of models generated by Learning Department 206 include regression learners and multilayer neural networks.

[0045] The estimation unit 208 of the machine learning unit 205 performs estimation processing based on estimation data received from the machine learning processing instruction unit 130, using tension-related data (tension value itself or tension value adjustment behavior) of the model stored in the model storage unit 209. The estimation processing performed by the estimation unit 208 can be based on known methods such as supervised learning or reinforcement learning. Furthermore, the estimation unit 208 may not be necessary when the control device 1 operates in learning mode.

[0046] In this embodiment, the control device 1, equipped with the aforementioned structure, learns a machine learning model in learning mode that can estimate the tension value data Ot of the conveyor belt, etc., based on data Is related to the mechanical structure, data Iw related to the workpiece, and data Ic related to operating conditions. The model generated by the learning unit 206 can replace the tension sensor 4 in estimating the tension value of the conveyor belt, etc., relative to the operating state of the industrial machinery 3 during actual operation. Furthermore, in actual operation mode, the control device 1 estimates the tension value of the conveyor belt, etc., of the industrial machinery 3 based on the data Is related to the mechanical structure, data Iw related to the workpiece, and data Ic related to operating conditions. Therefore, the tension sensor 4 is not required in the industrial machinery 3, reducing the manufacturing and operation costs of the industrial machinery 3.

[0047] Figure 3 The functions of the control device 1 according to the second embodiment of the present invention are represented by a schematic block diagram. In this embodiment, the machine learning unit 205 performs supervised learning.

[0048] The machine learning data generation unit 120 of this embodiment includes a state observation unit 122 and a label generation unit 124.

[0049] The state observation unit 122 generates state data S representing the operating state of the industrial machinery 3 when it is in operation, based on the data stored in the data acquisition and storage unit 190. The state data S representing the operating state of the industrial machinery 3 when it is in operation includes at least data Is related to the mechanical structure, data Iw related to the workpiece, and data Ic related to the operating conditions.

[0050] The tag generation unit 124 generates tag data L based on the data stored in the data storage unit 190, targeting the data generated by the state observation unit 122. The tag data L includes at least the constant tension value data Ot of the conveyor belt when the industrial machinery 3 is in operation, which is observed as state data S representing the operation state of the industrial machinery 3.

[0051] When the control device 1 operates in learning mode, the machine learning data generation unit 120 generates training data T for machine learning processing based on the state data S representing the operating state of the industrial machinery 3 generated by the state observation unit 122 and the label data L generated by the label generation unit 124. The training data T is output to the machine learning processing instruction unit 130. On the other hand, when the control device 1 operates in actual operation mode, the machine learning data generation unit 120 outputs the state data S representing the operating state of the industrial machinery 3 generated by the state observation unit 122 as estimation data to the machine learning processing instruction unit 130.

[0052] Then, when the control device 1 operates in learning mode, the machine learning processing instruction unit 130 instructs the machine learning unit 205 to perform supervised learning processing based on the training data T. On the other hand, when the control device 1 operates in actual operation mode, the machine learning processing instruction unit 130 instructs the machine learning unit 205 to estimate the tension value based on the estimated data.

[0053] Figure 4 This is a schematic block diagram illustrating the functions of the control device 1 according to the third embodiment of the present invention. In this embodiment, the machine learning unit 205 performs reinforcement learning.

[0054] The machine learning data generation unit 120 of this embodiment includes a state observation unit 122 and a decision data generation unit 126.

[0055] The state observation unit 122 generates state data S representing the operating state of the industrial machinery 3 when it is in operation, based on the data stored in the data acquisition and storage unit 190. The state data S representing the operating state of the industrial machinery 3 when it is in operation includes at least data Is related to the mechanical structure, data Iw related to the workpiece, and data Ic related to the operating conditions.

[0056] The determination data generation unit 126 generates determination data D for the data generated by the state observation unit 122 based on the data stored in the data storage unit 190. The determination data D includes at least the difference between the tension value data Ot of the conveyor belt, etc., when the state data S representing the operating state of the industrial machinery 3 is observed and the predetermined tension value adjustment action a is taken, and the target tension value.

[0057] When the control device 1 operates in learning mode, the machine learning data generation unit 120 generates learning data for machine learning processing based on the state data S representing the operating state of the industrial machinery 3 generated by the state observation unit 122, the predetermined tension value adjustment behavior a, and the judgment data D generated by the judgment data generation unit 126. The generated learning data is output to the machine learning processing instruction unit 130. On the other hand, when the control device 1 operates in actual operation mode, the machine learning data generation unit 120 outputs the state data S representing the operating state of the industrial machinery 3 generated by the state observation unit 122 as estimation data to the machine learning processing instruction unit 130.

[0058] Furthermore, when the control device 1 operates in learning mode, the machine learning processing instruction unit 130 instructs the machine learning unit 205 to perform reinforcement learning processing based on the learning data. On the other hand, when the control device 1 operates in actual operation mode, the machine learning unit 205 instructs the machine learning unit 205 to estimate the adjustment behavior of the tension value based on the estimated data.

[0059] In addition to the learning unit 206 and the estimation unit 208, the machine learning unit 205 of this embodiment also includes a reward calculation unit 207.

[0060] The learning unit 206, based on the learning data received from the machine learning processing instruction unit 130, performs reinforcement learning to assess the merits of an adjustment action 'a' with a predetermined tension value for the operational state of the industrial machine 3. Reinforcement learning learns the merits of the outcome when the predetermined action is performed, in the form of a reward representing the value of that action, based on the current state of the learning object. Through a trial-and-error learning loop, the strategy that maximizes the total reward (the adjustment action with the tension value for the operational state of the industrial machine 3) is learned as the optimal solution. Examples of reinforcement learning methods include Q-learning. In this case, the learning unit 206 can generate a value function Q (model) for reinforcement learning, for example, using a regression learner or a neural network.

[0061] Return Calculation Department 207 passed Figure 1 The processor 201 of the machine learning machine 2 shown executes the system program read from the ROM 202, mainly through computational processing using RAM 203 and non-volatile memory 204. The reward calculation unit 207 calculates a predetermined reward R representing the value of a predetermined action based on the judgment data D included in the learning data received from the machine learning processing instruction unit 130, and outputs the calculation result to the learning unit 206. In, for example, Q-learning in the learning unit 206, the closer the judgment data D is to 0 (e.g., the tension value data Ot of the conveyor belt, etc., during the operation of industrial machinery 3 matches the target tension value), the more positive (plus) the reward R is set; the further away from 0, the more negative (minus) the reward R is set.

[0062] The estimation unit 208, based on the estimation data received from the machine learning processing instruction unit 130, executes an estimation of the tension value adjustment behavior instruction using the model stored in the model storage unit 209. Using the model generated through reinforcement learning by the learning unit 206, the estimation unit 208 calculates the reward for each of the multiple tension value adjustment behaviors ai (i = 1 to n) that can be taken under the current conditions, given the observed state data S related to the operating state of the industrial machinery 3 included in the estimation data. The adjustment behavior a with the highest calculated reward is estimated as the optimal solution. The tension value adjustment behavior instruction estimated by the estimation unit 208 is output to the tension adjustment unit 140.

[0063] The present invention has been described above as one embodiment, but the present invention is not limited to the described embodiment and can be implemented in various ways by applying appropriate modifications.

[0064] For example, in the described embodiment, examples of using data related to the operating state of the industrial machinery 3, namely data Is related to the mechanical structure, data Iw related to the workpiece, and data Ic related to the operating conditions, are shown. However, data obtained by adding environmental data Ie related to the environment in which the industrial machinery 3 operates can also be used as data related to the operating state of the industrial machinery 3. Examples of environmental data Ie include ambient temperature and ambient humidity. When the ambient temperature changes, it affects the hardness, viscous strength, rigidity, etc., of the components used in the conveying section. In addition, when the ambient humidity changes, it affects the friction between the conveying section and other parts. Therefore, by processing this environmental data Ie, tension can be adjusted with higher precision.

[0065] Furthermore, in the described embodiment, conveying machinery such as belt conveyors was exemplified as industrial machinery 3, but for example, Figure 5 As illustrated, it can also be applied to the machine 400 that delivers the thin-film workpiece 402. Figure 5 In the illustrated machine 400, the workpiece 402 itself functions as a conveyor. Typically, in thin-film forms such as films, the workpiece 402 is fed from the work roller 401 via multiple rollers 403. Furthermore, a tension sensor 404 detects the tension associated with the workpiece 402, preventing slippage between the workpiece 402 and the rollers 403. A tension adjustment mechanism 405 adjusts the tension to avoid applying excessive load to the workpiece 402. With the application of the control device 1 of this application, tension adjustment can be performed without the need for the tension sensor 404.

[0066] Explanation of reference numerals in the attached figures

[0067] 1 Control device

[0068] 3 Industrial machinery

[0069] 4 tension sensors

[0070] 5 Networks

[0071] 6 Fog Computer

[0072] 7 cloud servers

[0073] 11 CPU

[0074] 12 ROM

[0075] 13 RAM

[0076] 14 Non-volatile memory

[0077] Interfaces 15, 17, 18, 20, and 21

[0078] 16 PLC

[0079] 19 I / O Units

[0080] 22 bus

[0081] 30-axis control circuit

[0082] 40 servo amplifier

[0083] 50 servo motors

[0084] 70 display devices

[0085] 71 Input Device

[0086] 72 External Devices

[0087] 100 Control Department

[0088] 110 Data Acquisition Department

[0089] 120 Machine Learning Data Generation Department

[0090] 130 Machine Learning Processing Command Unit

[0091] 140 tension adjustment section

[0092] 2 Machine Learning Machines

[0093] 201 processor

[0094] 202 ROM

[0095] 203 RAM

[0096] 204 non-volatile memory

[0097] 205 Machine Learning Department

[0098] 206 Study Department

[0099] 207 Return Calculation Department

[0100] 208 Presumption Department

[0101] Model 209 Storage Department.

Claims

1. A control device for controlling an industrial machine having a generally belt-shaped conveyor section equipped with drive rollers to convey workpieces, characterized in that, have: The data acquisition unit acquires data related to the mechanical structure, workpiece, and operating conditions of the industrial machinery's operating status. The data acquisition unit stores data related to the mechanical structure, workpiece, and operating conditions of the industrial machinery, which are acquired by the data acquisition unit. The machine learning data generation unit generates machine learning data for machine learning processing based on the data stored in the acquired data storage unit. The machine learning processing instruction unit executes machine learning processing to estimate data related to the tension of the conveying unit based on data generated by the machine learning data generation unit. as well as The machine learning unit performs machine learning processing to estimate data related to tension in the conveying unit, based on instructions from the machine learning processing instruction unit.

2. The control device according to claim 1, characterized in that, The data acquisition unit also acquires tension value data from the conveying unit. The machine learning data generation unit has the following features: The status observation unit generates status data based on the data stored in the data acquisition and data storage unit, which includes data related to the mechanical structure, data related to the workpiece, and data related to the operating conditions. as well as The label generation unit generates label data that includes the tension value data from the conveying unit, based on the data stored in the data acquisition and storage unit. The machine learning processing instruction unit instructs the machine learning unit to perform a learning process based on the state data observed by the state observation unit and the label data generated by the label generation unit to generate a model for estimating data related to the tension in the conveying unit. The machine learning unit has: The learning unit generates a supervised learning model based on the state data and the label data to learn the correlation between the tension value of the conveyor and the operating state of the industrial machinery. as well as A model storage unit stores the models generated by the learning unit.

3. The control device according to claim 1, characterized in that, The machine learning data generation unit includes a state observation unit, which generates state data based on data stored in the acquired data storage unit. This state observation unit includes data related to the mechanical structure, data related to the workpiece, and data related to operating conditions. The machine learning processing instruction unit instructs the machine learning unit to perform data processing that estimates the tension-related data in the conveying unit based on the state data observed by the state observation unit, as the machine learning process. The machine learning unit includes an estimation unit that, based on the state data, estimates the tension value of the conveyor using a supervised learning model. The supervised learning model learns the correlation between the tension value of the conveyor and data related to the operating state of the industrial machinery. The control device also includes a tension adjustment unit, which adjusts the tension of the conveying unit based on the tension value of the conveying unit estimated by the estimation unit.

4. The control device according to claim 1, characterized in that, The data acquisition unit also acquires tension value data from the conveying unit. The machine learning data generation unit has the following features: The status observation unit generates status data based on the data stored in the data acquisition and data storage unit, which includes data related to the mechanical structure, data related to the workpiece, and data related to the operating conditions. as well as The determination data generation unit generates determination data based on the data stored in the acquisition data storage unit. The determination data includes the difference between the adjusted tension value in the conveyor and the target tension value of the conveyor, assuming a predetermined tension adjustment action was taken when the state data was observed. The machine learning processing instruction unit instructs the machine learning unit to perform processing to generate a model for estimating data related to tension in the conveying unit and processing to estimating the tension adjustment behavior of the conveying unit, based on the state data observed by the state observation unit and the determination data generated by the determination data generation unit. The machine learning unit has: The return calculation unit calculates a return based on the determination data, representing the value of the predetermined tension adjustment behavior of the conveying unit; The learning unit, based on the state data and the reward, generates a reinforcement learning model that learns the value of the adjustment behavior for the action state of the industrial machinery. A model storage unit that stores the models generated by the learning unit; as well as The estimation unit, based on the state data, estimates the tension adjustment behavior of the conveying unit using a reinforcement learning model stored in the model storage unit. The control device further includes a tension adjustment unit that adjusts the tension of the conveying unit based on the tension adjustment behavior of the conveying unit estimated by the estimation unit.

5. The control device according to claim 1, characterized in that, The machine learning data generation unit includes a state observation unit, which generates state data based on data stored in the acquired data storage unit. This state observation unit includes data related to the mechanical structure, data related to the workpiece, and data related to operating conditions. The machine learning processing instruction unit instructs the machine learning unit to perform processing to estimate the tension adjustment behavior of the conveying unit based on the state data observed by the state observation unit, as the machine learning processing. The machine learning unit includes an estimation unit that, based on the state data, estimates the tension adjustment behavior of the conveyor using a reinforcement learning model. The reinforcement learning model learns the value of the tension adjustment behavior of the conveyor for the operational state of the industrial machinery. The control device further includes a tension adjustment unit that adjusts the tension of the conveying unit based on the tension adjustment behavior of the conveying unit estimated by the estimation unit.

Images