JP7226411B2 - An expansion module, an industrial device, and a method for estimating a parameter of an industrial device or an internal state of a device controlled by said industrial device - Google Patents

Description

The present invention relates to an expansion module, an industrial device and a method for estimating parameters of an industrial device or an internal state of a device controlled by said industrial device.

In recent years, the technology in the field of AI (artificial intelligence) has progressed, and its fields of use and applications have expanded greatly. The core technology is so-called deep learning, but the construction and execution of AI using deep learning requires large-scale parallel arithmetic processing, so the computer used must have high computing power. .

On the other hand, small computers such as single-board computers used in embedded devices do not have high computing power and may not be powerful enough to execute AI programs based on deep learning, which require a large amount of computation. For this reason, external devices that add computing power suitable for AI programs to such devices are commercially available. For example, by connecting to a USB (Universal Serial Bus) port of a small computer, an external VPU (Visual Processing Unit) can be used.

Since the external devices mentioned above are general-purpose products and only provide extended computing power, their use requires general AI programming knowledge and skills, as well as It is necessary to understand the SDK (software development kit) provided for it, and it is necessary to perform AI programming by yourself according to the intended use.

In industrial equipment such as PLCs (programmable logic controllers) and FA (factory automation) equipment such as servo controllers, the computing power is determined by the balance between cost and power consumption. Only the degree necessary for operation is secured, and in general, it is difficult to say that the spare capacity is very large.

Patent Document 1 discloses a machine learning device that obtains a feedforward coefficient in a servo control device, which is one of such industrial devices, by machine learning. In the device disclosed in this document, a servo control device and a machine learning device are connected by an information communication network such as a LAN (local area network) built in a factory, the Internet, or a public telephone network.

JP 2018-152012 A

When AI technology is applied to estimate the parameters suitable for the mode of use of industrial equipment and the internal state of equipment controlled by industrial equipment, the individual industrial equipment lacks computing power and memory capacity. In addition, since industrial equipment is not often connected to a general information communication network, it is not always possible to communicate with a computer having high computing power via the information communication network.

Also, even if industrial equipment is connected to an appropriate information communication network, engineers in the FA field who regularly use industrial equipment do not necessarily have specialized training in AI programming. Even if it is possible to secure personnel who have undergone specialized training, it is unrealistic to perform programming each time because the mode of use of industrial equipment varies depending on the situation of use.

The present invention has been made based on such circumstances, and its object is to perform machine learning such as deep learning, which has a high computational load and a large amount of memory, without increasing the cost of the main body of industrial equipment. is available only when needed.

An expansion module according to one aspect of the present invention is an expansion module that has at least a processor and a memory and is connected to an external terminal of a servo controller, wherein the processor is a machine suitable for the configuration assumed for the memory. To specify a machine learning model including a displacement of a motor controlled by the servo controller from the servo controller by being connected to the external terminal of the servo controller when the learning model is not stored Necessary information is collected, and when the processor is detached from the external terminal of the servo controller and connected to an external server so as to be able to communicate information, the processor needs to specify the machine learning model to the server. While transmitting information, an intermediate learning model is received from the server and stored in the memory, and the processor is again connected to the external terminal of the servo controller, so that the intermediate learning model is processed by the servo controller. Learning is performed using the information obtained from the data as teacher data.

Also, in the expansion module according to one aspect of the present invention, the intermediate learning model may be a learning model that has completed learning for an event with a low occurrence frequency.

In addition, in the expansion module according to one aspect of the present invention, the infrequently occurring event is a failure caused by at least one of wear and aging caused by long-term operation of the motor, It may include at least one event selected from a sign of failure caused by at least one of wear and aging that occurs during operation, and an accidental event including at least one of foreign matter getting stuck and breakage.

Further, in the expansion module according to one aspect of the present invention, the information necessary to identify the machine learning model includes information about the servo controller and information identifying the device controlled by the servo controller and its configuration. may contain.

Further, in the method for estimating a parameter of an industrial device or an internal state of a device controlled by the industrial device according to one aspect of the present invention, the processor further transmits the learned learning model to the industrial device and the industrial equipment inputs its own information as input data into the trained learning model, and outputs an estimated value of the parameters of the industrial equipment or the internal state of equipment controlled by the industrial equipment. can be obtained as

1 is an overview perspective view of an industrial device, an expansion module connected to the industrial device, and a control target device controlled by the industrial device; FIG. FIG. 3 is a schematic block diagram showing the hardware configuration of an industrial device, an expansion module, and a device to be controlled; FIG. 2 is a functional block diagram showing an internal configuration focusing on the function of each device in a state where the industrial device and the expansion module are connected; FIG. 4 is a functional block diagram showing the internal configuration of an industrial device equipped with a trained machine learning model; FIG. 4 is a flow chart showing a procedure from performing machine learning to estimating parameters of an industrial device or estimating an internal state of a device to be controlled; It is a functional block diagram explaining industrial equipment and an extension module concerning a 2nd embodiment of the present invention. It is a functional block diagram explaining industrial equipment and an extension module concerning a 3rd embodiment of the present invention. FIG. 10 is a flow diagram showing a procedure for performing learning on a machine learning model in the third embodiment of the present invention; FIG. 11 is a flowchart showing a procedure for estimating parameters and the internal state of a device using a learned machine learning model in the third embodiment of the present invention; FIG. 13 is a flow diagram showing a procedure for automatically selecting and acquiring a large number of machine learning models in the fourth embodiment of the present invention;

Hereinafter, the internal state of the industrial device 1 and the expansion module 2 according to the first embodiment of the present invention, and the internal state of the industrial device 1 using them or the internal state of the control target device 3 controlled by the industrial device 1 will be described. The estimation method will be explained with reference to FIGS.

FIG. 1 is an overview perspective view of an industrial device 1, an expansion module 2 connected to the industrial device 1, and a controlled device 3 controlled by the industrial device 1. FIG. Here, in the embodiment shown in this specification, the industrial equipment 1 is shown as a servo controller, and henceforth, the servo controller is illustrated and demonstrated as the industrial equipment 1. FIG. In this specification, the term "industrial equipment" refers to equipment that can be used for FA, and has the function of controlling some other equipment (control target equipment 3) or inputting/outputting information with other equipment. , and may be a PLC, a single board microcomputer, or the like, in addition to the servo controller. Note that the actual use of the industrial equipment 1 is not limited to FA, and may be used by being incorporated in various devices such as vehicles and amusement equipment as well as production equipment.

Also, the controlled device 3 is a device that is controlled by the industrial device 1 or that inputs/outputs information indicating the state of the device. Here, a servomotor is shown as the device 3 to be controlled, and a servomotor will be exemplified and described below as the device 3 to be controlled. Note that the control target device 3 may be other types of rotating electric motors (for example, stepping motors, etc.), various actuators, switches, sensors, and the like. Moreover, the number of controlled devices 3 connected to the industrial device 3 does not have to be one, and a plurality of controlled devices 3 may be connected.

The industrial equipment 1 is provided with an external terminal 101 for connecting with an appropriate external equipment. Although the external terminals 101 are provided on the front surface of the industrial equipment 1 in FIG. 1, they may be provided on another surface, for example, the rear surface, or a plurality of external terminals 101 may be provided. The external terminal 101 is a connection terminal for inputting/outputting information between the industrial device 1 and an appropriate external device, and may be of any standard. In this embodiment, the external terminal 101 is a USB terminal, but it may be a terminal conforming to other appropriate standards or a terminal conforming to its own standard.

As shown in FIG. 1, the expansion module 2 is small and may be directly connected to and held by the external terminals 101 . Also, the power of the expansion module 2 may be supplied from the connected industrial equipment 1 via the external terminal 101 . This is because, when connecting the extension module 2 to the industrial equipment 1, there is no need to arrange extra cables or separately prepare a power supply cord, and the arrangement can be performed easily. However, there are no technical restrictions on the size and shape of the expansion module 2, and the expansion module 2 is not limited to the stick-like shape shown in FIG. Also, it is possible to connect to the external terminal 101 of the industrial equipment 1 using a cable, or to supply power to the extension module 2 itself from a separate power supply. For example, it may be difficult for the operator to directly access the external terminal 101 because the industrial equipment 1 is built inside the apparatus. It is convenient to connect the expansion module 2 to such an extension cable. Further, as will be described later, the expansion module 2 has a high information processing capability and requires a large amount of power when performing calculations. , it is preferable that power be supplied to the expansion module 2 separately.

In any case, the extension module 2 is directly connected to the external terminal 101 of the industrial equipment 1 when necessary, and is virtually connected to the industrial equipment via a general-purpose information communication network such as TCP/IP. It is not connected to the device 1. This is because, due to the nature of the industrial equipment 1, it is practically often incorporated into some kind of device and cut off from an external general-purpose information communication network for security or other requirements.

Here, the expansion module 2 is an information processing device (so-called computer) that is independent of itself, and has at least a processor and a memory therein. The industrial equipment 1, which is the servo controller shown in this embodiment, also has a function as an information processing equipment including a servo amplifier, a processor for controlling it, and a memory. The performance as an information processing device provided in the industrial equipment 1 is originally designed to satisfy the intended functions of the industrial equipment 1 . That is, as shown in this embodiment, if the industrial equipment 1 is a servo controller, its information processing capability is sufficient as long as it is sufficient for controlling the processing target equipment 3, which is a servo motor. If the device 1 is a PLC, its information processing capability is sufficient as long as it is sufficient for logical operations and input/output in units of several microseconds. Such computing power is usually not provided. The expansion module 2 includes a processor suitable for high-load calculations such as such large-scale parallel calculations. That is, only when the expansion module 2 is connected to the external terminal 101, the industrial equipment 1 is provided with computing power capable of withstanding a high load. The processor installed in the expansion module 2 may be a general CPU (central processing unit) (with excellent processing power), a GPU (graphics processing unit), a VPU, an ASIC (application-specific integrated circuit). ) or any combination of these multiple types.

As will be described later, the calculations performed by the extension module 2 are mainly focused on machine learning in a multi-layered neural network model such as so-called deep learning. Therefore, the design of the expansion module 2 is suitable for the computation in the machine learning model to be applied, and is usually a parallel arrangement of many pipelines so that large-scale parallel computation can be processed at high speed. However, as long as it is suitable for machine learning of the machine learning model to be used, there is no problem.

FIG. 2 is a schematic block diagram showing the hardware configuration of the industrial equipment 1, the expansion module 2, and the controlled equipment 3. As shown in FIG. The industrial equipment 1 includes a servo amplifier 102 and a control circuit 103, receives a command from an external equipment, controls an inverter 106, and uses the direct current converted by an AC/DC converter 105 to drive the equipment 3 to be controlled. It converts to a suitable 3-phase AC and outputs it. A command from an external device and an output from a sensor 301 that detects the state of the control target device 3 are input to the control circuit 13, and based on various information stored in the memory 107, the processor 104 causes the inverter 106 to Determines the control command to be used.

Here, the sensor 301 is a general term for all means for detecting information about the controlled device 3. As in this example, when the controlled device 3 is a servomotor, a rotation angle sensor such as a rotary encoder, a torque sensor, and the like are used. Various detectors may be included such as sensors, motor current means, thermometers.

The memory 107 also stores various parameters used when controlling the controlled device 3, such as gain parameters, speed/acceleration constants and time constants that define the speed waveform of the motor, and model constants of the controlled device 3. In addition to storing other useful information, for example, the operating time of the controlled device 3, the total drive amount such as the total number of rotations, the model number and serial number of the controlled device 3, the user information, etc., may be stored. is appropriately referred to and used by the processor 14 according to the type of the control target device 3 and the content of control. Furthermore, as will be described later, the memory 107 may store a learned machine learning model regarding the controlled device 3 .

The industrial equipment 1 is also provided with an I/O 108 as an interface with external equipment, and the external terminal 11 shown in FIG. The I/O 108 includes one or more control circuits and physical terminals for information communication, and the external terminal 101 is one of these physical terminals.

The expansion module 2 includes a memory 201 and a processor 202, and the processor 202 can independently refer to the memory 201 and perform operations. Further, the expansion module 2 is provided with an I/O 203 as an interface, and the I/O 203 includes a control circuit for information communication and a physical device for connecting to the external terminal 101 of the industrial equipment 1. Includes terminals. FIG. 2 shows a state in which the I/O 108 of the industrial equipment 1 and the I/O 203 of the expansion module 2 are connected so as to be able to communicate information. Then, the extension module 2 can appropriately receive information from the sensor 301 collected about the controlled device 3 from the I/O 203 through the industrial device 1 .

FIG. 3 is a functional block diagram showing the internal configuration of the industrial equipment 1 and the expansion module 2 in a state of being connected, focusing on the functions of each equipment.

The industrial equipment 1 includes a motor control unit 109, a servo amplifier 102, and a trained machine learning model 110 in some cases. As hardware, the motor control unit 109 is mainly configured by the control circuit 103 shown in FIG. It contains the necessary motor parameters 111 . Also, the learned machine learning model 110 is stored in the memory 107 shown in FIG. Then, by executing machine learning using the extension module 2 , a trained machine learning model 110 is constructed later and stored in the memory 107 of the industrial equipment 1 . In FIG. 3, dashed lines clearly indicate parts that are not prepared at the beginning and are acquired later, and arrows that indicate the flow of information that is performed when the parts are acquired.

Extension module 2 includes a machine learning model 204 . The machine learning model 204 is configured so that the output from the sensor 301 can be input.

In the configuration shown in FIGS. 1 to 3, the flow until finally the controlled device 3 is steadily controlled by the industrial device 1 will be described.

In order to control the controlled device 3 efficiently, accurately, and at high speed using not only the motor controller but also various industrial devices 1, it is necessary to provide parameters suitable for the mode of use of the controlled device 3. not. Until now, many proposals have been made for methods of estimating such parameters and automatic setting methods, and technical developments have been made. There are cases in which satisfactory parameters cannot be obtained by a uniform setting method. In addition, the usage environment of the control target device 3 may change from moment to moment. For example, if the work performed by the robot arm differs for each task, or if the road surface conditions differ for a vehicle traveling in an external environment, or if the load weight changes, the optimal parameters will naturally change. It is also difficult to reset the parameters every time there is a significant change.

Alternatively, it is difficult to directly measure important internal states such as physical changes in the controlled device 3 . For example, it is difficult to know the state of wear and deterioration over time, and maintenance such as replacing parts at regular operating hours is inevitable, but this means discarding parts that have a useful life and do not need to be replaced. In addition to increasing maintenance costs, this also reduces equipment uptime due to maintenance time requirements. It cannot be said that it is impossible to estimate the internal state from the output of the sensor provided in the controlled device 3 depending on the conditions. , and reliability is not satisfactory.

Machine learning is expected to excel at such indirect estimation of the suitable parameters of the industrial equipment 1 and the internal state of the controlled equipment 3 from the output of the sensor provided in the controlled equipment 3. , if it becomes possible to use it, it will be of great benefit.

Therefore, in the present embodiment, when using machine learning represented by deep learning for such applications, first, the control target device 3 is operated in a state where the industrial device 1 is connected, and the sensor 301 is operated. Learning is performed for the machine learning model 204 using the information obtained from the above as teacher data. By advancing the learning of the machine learning model 204 based on the actual operation data of the control target device 3 prepared according to the specific use, the highly accurate learned machine learning model 110 corresponding to the individual use can be obtained.

At this time, in obtaining the actual operation data of the controlled device 3 as teacher data for the machine learning model 204 from the sensor 301, the control itself for operating the controlled device 3 is executed by the industrial device 1, but the sensor obtained is the sensor 301. Data from 301 is sent to expansion module 2 . The extension module 2 performs machine learning on a machine learning model 204 prepared in advance in a memory 201 using a processor 202 mounted therein.

Here, what kind of machine learning model 204 should be selected according to the configuration and application of the controlled device 3, and is not particularly limited. As a representative example, assuming a so-called deep neural network model with a multi-layered structure, for example, a perceptron layer of about 6 to 10 layers, although it depends on the number of perceptron nodes and the number of bits in each node, the machine learning requires Calculations generally have a large calculation load, and the processor of the industrial equipment 1 lacks the calculation capability. That is, it is not possible to operate the controlled device 3 and perform machine learning one after another based on the data obtained from the sensor 301, and even if sufficient data is obtained, it takes a long time to complete the learning. Alternatively, a large memory capacity for holding data necessary for machine learning is temporarily required. Therefore, calculations in such machine learning are performed using the processor 202 mounted on the expansion module 2 having higher calculation performance. The computing power of the processor 202 should be such that the data from the sensor 301, which are successively acquired, can be acquired in real time, or the machine learning model 204 can be learned without much delay from the data acquisition speed, or higher. .

Also, the machine learning model 204 is prepared in advance according to the configuration of the control target device 3 that is assumed to be used by the industrial device 1 and is held in the memory 201 of the expansion module 2 . The machine learning model 204 may be appropriately prepared by an operator who intends to perform machine learning according to the configuration and application of the device 3 to be controlled. , and simply select the most suitable machine learning model 204 .

In order for the operator to prepare the machine learning model 204, as an example, from among a plurality of machine learning models provided via the Internet, etc., the one closest to the configuration and application of the control target device 3 to be used is prepared in advance on a PC or the like. , the extension module 2 is connected to the information processing equipment, and the downloaded machine learning model 204 is transferred to the memory 201, the operator must construct the machine learning model 204 by himself. Machine learning can be used without knowledge and without proficiency in AI technology.

Further, when a plurality of machine learning models 204 are stored in advance in the memory 201 of the expansion module 2, the operator may specify one of them. Alternatively, the expansion module 2 or the industrial device 1 automatically performs machine learning based on information indicating the configuration of the controlled device 3, such as the model number and serial number of the controlled device 3 connected to the industrial device 1. A model 204 may be selected. In this way, even if the operator does not have sufficient knowledge of AI technology, the machine learning model 204 suitable for the configuration and application of the controlled device 3 is automatically selected.

Furthermore, when performing machine learning, the number of machine learning models 204 to be selected is not limited to one, and a plurality of models may be selected. In this case, the extension module 2 simultaneously performs machine learning on the selected plurality of machine learning models 204 in parallel, and among the finally obtained machine learning models 204, the highest performance (for example, prediction accuracy, etc.) ) may be selected and used. In this case, although the computing power of the processor 202 of the extension module 2 needs to be high enough to perform machine learning on a plurality of machine learning models 204 at the same time, highly accurate learned machine learning models can be obtained. more likely.

The procedure from performing machine learning to estimating the parameters of the industrial equipment 1 or estimating the internal state of the controlled equipment 3 will be described according to the flowchart shown in FIG. 5 as follows. That is, when the control target device 3 is connected to the industrial device 1 and the learned machine learning model has not yet been obtained, first, the extension module 2 is connected to the external terminal 101 of the industrial device 1 (ST1). . Subsequently, the controlled device 3 is driven to obtain data necessary for machine learning, and information about the controlled device 3 is acquired from the sensor 301 (ST2). Then, using the obtained information as teacher data, processor 202 of extension module 2 is used to perform machine learning of the machine learning model stored in memory 201 (ST3).

When machine learning has progressed sufficiently, a trained machine learning model 110 is obtained in the memory 201 of the extension module 2 . The trained machine learning model 110 is transferred and stored in the memory 107 of the industrial equipment 1 (ST4). In this way, the industrial equipment 1 is provided with a trained machine learning model 110 that has been trained for a specific control target equipment 3 without relying on its own processor 104 .

The learned machine learning model 110 receives information obtained from the sensor 301 and outputs some desired information. Therefore, a desired output is obtained by appropriately inputting information about the controlled device 3 obtained from the sensor 301 to the trained machine learning model 110 (ST5). The information to be output may be one or more parameters to be used in the industrial equipment 1, may be an estimate of the internal state of one or more controlled equipment 3, and both There may be.

To give a more specific example, the operation waveforms such as the speed, torque, and current of the controlled device 3, which is a servo motor, are input, and gain parameters suitable for control are output, or signs of wear and aging deterioration are detected at an early stage. It is possible to build a learned machine learning model 110 that estimates, outputs a warning signal prompting an operator to inspect or replace parts, or estimates the remaining life until a failure occurs.

Since this learned machine learning model 110 is stored in the memory 107 of the industrial equipment 1 in this embodiment, the parameters of the industrial equipment or the industrial equipment using the learned machine learning model 110 Estimation of the internal state of the controlled appliance does not require the expansion module 2. Therefore, when the industrial equipment 1 is used regularly, the expansion module 2 does not need to be connected all the time, and can be removed.

FIG. 4 is a functional block diagram showing the internal configuration of the industrial equipment 1 equipped with the trained machine learning model 110. As shown in FIG. If the industrial equipment 1 has acquired the learned machine learning model 110, as shown in FIG. It is possible to estimate a suitable parameter or estimate the internal state of the device 3 to be controlled.

Here, the trained machine learning model 110 is also typically a deep neural network model having multiple perceptron layers, and requires a considerable amount of calculations to output an input value. However, compared to performing machine learning, the required amount of calculation is small, and the estimated value to be obtained using the trained machine learning model 110 often does not require real-time or urgency. Sufficient practicability can be ensured by computation using the surplus computing power of the processor 104, such as between normal control of the control target device 3. FIG.

As described above, since the expansion module 2 is not necessarily required during steady operation, even when operating a large number of industrial devices 1, the expansion module 2 is temporarily connected only to the industrial device 1 for which machine learning is to be performed. It is economical because the expansion module 2 can be reused.

As described above, in the present embodiment, the learned machine learning model 110 outputs gain parameters suitable for control, or estimates signs of wear and aging deterioration at an early stage, and conducts inspections and parts replacement. This is intended to output a warning signal prompting the operator or to estimate the remaining life until failure occurs. It is difficult to obtain a learning effect for infrequent events, and the efficiency of learning is not necessarily good.

Therefore, the machine learning model 204 stored in the memory 201 of the expansion module 2 is preferably an intermediate learning model that has undergone a certain amount of learning in advance according to the configuration of the industrial equipment 1 and its control target equipment 3. . This utilizes a machine learning method called so-called transfer learning, and uses a representative configuration close to the configuration of the assumed control target device 3 to perform intermediate learning and learning for events with low occurrence frequency in advance. It refers to using a learning model that has undergone the above as the machine learning model 204 . By using such an intermediately learned machine learning model 204, the efficiency of learning is high, and the accuracy of the machine learning model 204 can be improved at an early stage, and the problem of overfitting due to overlearning can be alleviated. A trained machine learning model 110 is obtained that is capable of correctly estimating events with a low occurrence frequency that are unlikely to be included in teacher data obtained from the device 1 and its control target device 3 .

Here, the events with low frequency of occurrence include, for example, failures such as wear and aging deterioration caused by long-time operation of the controlled device 3 and their signs, and accidental events such as foreign matter being caught and damaged. Conceivable. In particular, it is considered difficult to learn such phenomena by machine learning using the new industrial equipment 1 and the control target equipment 3. Therefore, it is decided to use the intermediate learning model as the machine learning model 204. has great advantages.

Further, the timing at which the industrial equipment 1 shown in FIG. It is preferable to set it as the case where the time has passed, or the case where the passage of a certain period of time has been detected.

As a first example, a case in which parameter estimation is required is considered to be a case in which the state of the controlled device 3 connected to the industrial device 1, such as the magnitude of the load, has changed. In such cases, it is believed that some physical change has been made to the configuration of the device itself using the industrial equipment 1, and such changes are generally considered to be made while the device is stopped. Therefore, for example, it is conceivable that it is performed when it is detected that the power of the industrial equipment 1 is turned on or off. In addition, a specific state of the industrial equipment 1 can be detected, such as when a change of a certain amount or more in the load of the controlled equipment 3 is detected, or when it is detected that the control contents of the controlled equipment 3 have been changed. When it is detected, it is preferable to estimate the parameters and the internal state of the controlled device 3 .

As a second example, since it is considered that changes in characteristics accompanying continuous use of the control device 3, failures due to wear and deterioration, or detection of signs thereof do not occur suddenly, For example, it is considered sufficient to estimate the parameters and the internal state of the control target device 3 each time the elapse of 24 hours is detected. In addition, the measurement for a certain period of time at this time may be based on the passage of real time, or the passage of the integrated amount of the time when the power of the industrial equipment 1 is on or the time when the controlled equipment 3 is in operation. can be done by

In this way, when a specific state of the industrial equipment 1 is detected, when the lapse of a certain period of time is detected, or both, by estimating the parameters and the internal state of the controlled equipment 3, the industrial equipment Parameters and the internal state of the control target device 3 can be estimated without imposing an unnecessary load on the processor 104 of the device 1 .

FIG. 6 is a diagram for explaining the industrial equipment 1 and the expansion module 2 according to the second embodiment of the present invention, and shows the functions when the industrial equipment 1 and the expansion module 2 are connected. It is a block diagram.

In this embodiment, the industrial equipment 1, the expansion module 2, and the controlled equipment 3 have the same physical configuration as in the previous embodiment, and their functions are also basically the same. The same numbers are assigned to the parts, and redundant explanations are omitted. Also, FIGS. 1 and 2 for the previous embodiment shall be incorporated into this embodiment.

This embodiment is completely the same as the previous embodiment up to the point of performing machine learning on the machine learning model stored in the memory 201 of the extension module 2 . 5, the extension module 2 is connected to the external terminal 10 of the industrial device 1 (ST1), the control device 3 is driven, and information is acquired from the sensor 301 (ST2). , and using the obtained information as teacher data, the processor 202 of the extension module 2 is used to perform learning for the machine learning model 204 on the memory 201 of the extension module 2 (ST3). However, in this embodiment, the finally obtained learned machine learning model 205 is continuously held in the memory 201 of the extension module 2 .

Therefore, the estimation of the parameters of the industrial equipment 1 and the internal state of the control target equipment 3 performed by the industrial equipment 1 at appropriate timing (as already described in the first embodiment) can be performed as shown in FIG. 2, the industrial equipment 1 and the expansion module 2 are connected.

In the case of this embodiment, it is necessary to connect the extension module 2 to the external terminal 101 of the industrial equipment 1 when estimating the parameters of the industrial equipment 1 and the internal state of the controlled equipment 3. Since the trained machine learning model 205 with a large amount of information does not need to be held in the memory 107 of the industrial equipment 1, there is an advantage that the capacity of the memory 107 of the industrial equipment 1 is at least good. Further, the calculation itself for estimating the parameters of the industrial equipment 1 and the internal state of the controlled equipment 3 is executed by the processor 202 mounted on the expansion module 2 having excellent calculation performance. Without imposing an extra load on the processor 104, especially when the computational performance of the processor 104 of the industrial equipment 1 is low, the parameters of the industrial equipment 1 and the estimated values of the internal state of the controlled equipment 3 can be quickly obtained. There are benefits to be gained.

FIG. 7 is a diagram for explaining the industrial equipment 1 and the expansion module 2 according to the third embodiment of the present invention, and shows the functions when the industrial equipment 1 and the expansion module 2 are connected. It is a block diagram.

Also in this embodiment, the industrial equipment 1, the expansion module 2, and the controlled equipment 3 have the same physical configuration as in the previous embodiment, and their functions are also basically the same. The same numbers are assigned to the parts to be replaced, and redundant explanations are omitted. Also, FIGS. 1 and 2 for the previous embodiment shall be incorporated into this embodiment.

Also in this embodiment, the extension module 2 is connected to the external terminal 101 of the industrial equipment 1, and the machine learning model stored in the memory 201 of the extension module 2 is subjected to machine learning. 2 embodiment. Also, the obtained learned machine learning model 205 is held in the memory 201 of the extension module 2 as in the second embodiment.

In this embodiment, the memory 201 of the extension module 2 can hold a plurality of trained machine learning models 205 as shown in FIG. In addition, the memory 201 holds information identifying the industrial equipment 1 that was connected when the learned machine learning model 205 was obtained, and the plurality of learned machine learning models 205 are each It is possible to identify which industrial equipment 1 is supported.

Here, when a certain learned machine learning model 205 corresponds to a specific industrial device 1, when performing machine learning to obtain the learned machine learning model 205, the extension module 2 is connected. , refers to the industrial equipment 1 that drives the controlled equipment 3 and collects teacher data. The information for identifying the industrial equipment 1 may be any information as long as it can uniquely identify the industrial equipment 1. In this embodiment, the serial number of the industrial equipment 1 and the , is a table 206 that holds a correspondence relationship with the obtained learned machine learning model.

That is, in the present embodiment, when the industrial equipment 1 and the extension module 2 are connected to perform machine learning on the machine learning model 204, the identification information 112 stored in the industrial equipment 1 is transferred to the extension module 2. , and the correspondence with the machine learning model 204 being learned is stored in the table 206 . The timing of transmission of this identification information 112 to the expansion module 2 may be at any point during machine learning, and may be transmitted first, or may be completed after machine learning is completed. It may be at the time or after the machine learning model 205 is obtained.

The extension module 2 is not necessarily used only for a specific single industrial device 1, but is used for a plurality of industrial devices 1. That is, the expansion module 2 is temporarily connected to a specific industrial equipment 1, then removed and then connected to another industrial equipment 1, and so on. Then, in the memory 201 of the expansion module 2, the learned machine learning model 205 is created and stored for each connected industrial device 1 as long as its storage capacity permits. FIG. 7 shows, as an example, that three learned machine learning models A to C are held.

Here, for an industrial device 1, in order to estimate its internal parameters or the internal state of the controlled device 3 connected to the industrial device 1, the learned machine acquired for the industrial device 1 A learning model 205 is required. Therefore, when estimating the internal parameters or the internal state of the controlled device 3 , the extension module 2 that already holds the learned machine learning model 205 in the memory 201 is connected to the external terminal 101 of the industrial device 1 .

Then, when a command is given to the industrial device 1 directly or indirectly via an external device to instruct the estimation of the internal parameters or the internal state of the controlled device 3 connected to the industrial device 1, Information about the controlled device 3 obtained by the sensor 301 and the identification information 112 of the industrial device 1 are sent to the extension module 2 .

The processor 202 of the extension module 2 compares the sent identification information 112 with the table 206 to identify the learned machine learning model 205 obtained for the currently connected industrial equipment 1 . Here, it is assumed that the trained machine learning model A corresponds.

The processor 202 inputs the information about the device 3 to be controlled that has been sent to the learned machine learning model A, and sends the obtained desired output to the industrial device 1 . The processor 104 of the industrial equipment 1 performs appropriate processing according to the type and content of the obtained output. In the example shown in FIG. 7, the output from the learned machine learning model A is the estimated value of the motor parameter 111, which is the internal parameter of the industrial equipment 1. Therefore, according to this output, the processor 104 changes the motor parameter 111 update. In addition to this, if the obtained output is an estimated value of the internal state, the processor 104 can display the information corresponding to the value to the operator. process.

In this way, the memory 201 of the expansion module 2 can store a plurality of learned machine learning models 205, and according to the identification information 112 of the industrial device 1 to which the expansion module 2 is connected, By selecting the learned machine learning model 205 to be used for estimating the internal state of the control target device 3 connected to the industrial device 1, using one extension module 2, the plurality of industrial devices 1 parameters and the internal state of the device can be appropriately executed by simply connecting to the external terminal 101 of the industrial device 1 without re-learning.

In this embodiment, the selection of the learned machine learning model 205 is automatically performed based on the identification information 112 of the industrial equipment 1, but the operator who is the user explicitly selects the learned machine learning model. 205 may be selected.

Regarding the third embodiment, the flow of learning the machine learning model 204 and estimating the parameters and the internal state of the device using the trained machine learning model 205 will be described with reference to the flowcharts shown in FIGS. .

FIG. 8 is a flow diagram showing a procedure for performing learning on the machine learning model 204 in the third embodiment.

First, the expansion module 2 is connected to the external terminal 101 of the industrial equipment 1 (ST11). At this stage, the machine learning model 204 held in the memory 201 of the extension module 2 has not been machine-learned yet. However, this machine learning model 204 may be an intermediate learning model, as already mentioned.

The industrial equipment 1 drives the equipment 3 to be controlled, and acquires information as teacher data from the sensor 3 (ST12). Furthermore, the obtained information is transmitted to the extension module 2, and the processor 202 of the extension module 2 performs machine learning on the machine learning model 204 using the obtained information as teacher data to obtain the learned machine learning model 205 ( ST13).

Also, the industrial equipment 1 transmits its own identification information 112 to the expansion module 2 (ST14). The timing of transmitting the identification information 112 may be earlier, for example, immediately after connecting the expansion module 2 to the external terminal 101 of the industrial equipment 1 in ST11.

Finally, the obtained correspondence relationship between the learned machine learning model 205 and the identification information 112 is stored in the memory 201 (ST15). In the example shown in FIG. 7, this correspondence is stored in the form of a table.

FIG. 9 is a flowchart showing a procedure for estimating parameters and the internal state of the device using the learned machine learning model 205 in the third embodiment.

First, if not yet connected, the expansion module 2 is connected to the external terminal 101 of the industrial equipment 1 (ST21). Here, the connected extension module 2 has already performed machine learning using the industrial equipment 1 in the procedure shown in FIG. there is

Subsequently, the industrial equipment 1 drives the controlled equipment 3 and acquires information from the sensor 301 (ST22). The acquired information and the identification information 112 of the industrial equipment 1 itself are sent to the expansion module 2 (ST23).

Processor 202 of extension module 2 identifies corresponding trained machine learning model 205 based on received identification information 112 (ST24). Furthermore, the transmitted information is input to the identified trained machine learning model 205 to obtain a desired output (ST25).

The expansion module 2 transmits the obtained output value to the industrial equipment 1 (ST26). The processor 104 of the industrial equipment 1 performs predetermined processing according to the type and value of the received output value.

In each embodiment described above, the machine learning model 204 (intermediate (including learning models) are prepared, or a suitable one is explicitly or automatically selected from a plurality of machine learning models 204 pre-stored in the extension module 2.

However, when preparing suitable machine learning models 204 for various configurations of the controlled device 3 and preparing intermediate learning models corresponding to their uses, depending on the case, the type of machine learning model 204 to be selected may be More cases are expected. In such a case, storing all of the many types of machine learning models 204 in advance in the memory 201 of the expansion module 2 is not a good idea in terms of the capacity of the memory 201. It is considered difficult for a non-skilled operator to select the optimum one from among them.

Therefore, an embodiment having a configuration for automatically selecting and acquiring a suitable model from a large number of machine learning models 204 including intermediate learning models will be described below as a fourth embodiment.

Also in the fourth embodiment of the present invention, the industrial equipment 1, the expansion module 2 and the controlled equipment 3 have the same physical configuration as the previous embodiment, and their functions are also basically the same. Therefore, the same numbers are given to the parts common to both, and overlapping explanations are omitted. Also, FIGS. 1 and 2 for the previous embodiment shall be incorporated into this embodiment. Further, the procedure of machine learning for the machine learning model 204 and the procedure of estimating the parameters of the industrial equipment 1 and the internal state of the controlled equipment 3 are any of the first to third embodiments described above. I don't mind.

The flow of selecting and acquiring the machine learning model 204 in this embodiment will be described below with reference to the flowchart shown in FIG. First, a learned machine learning model has not yet been obtained for the industrial equipment 1 to which the specific control target equipment 3 is connected, and the machine learning model 204 suitable for such a configuration has not yet been specified. In this state, the operator first connects the expansion module 2 to the external terminal 101 of the industrial equipment 1 (ST31).

At this point, the memory 201 of the expansion module 2 does not store the machine learning model 204, or may store some machine learning model 204 (eg, high availability). Assume that no machine learning model 204 suitable for the configured configuration is stored.

In this case, the extension module 2 collects information necessary for specifying the machine learning model 204 from the industrial equipment 1 and stores it in the memory 201 (ST32). The information necessary to identify the machine learning model 204 includes information about the industrial equipment and information about the equipment to be controlled. and information identifying its configuration. The information specifying the industrial equipment 1 may be the model number of the industrial equipment 1 or the like. The information specifying the controlled device 3 and its configuration includes the model number of the controlled device 3, information indicating performance, such as motor capacity, and information indicating the application and combination of the controlled device 3. It may be information indicating performance to be performed.

After that, the operator once removes the expansion module 2 from the industrial equipment 1 (ST33), and connects the expansion module 2 to a computer connected to an external information communication network such as the Internet (ST34). The expansion module 2 transmits the information required to identify the collected machine learning model 204 to the server connected to the information communication network via the external information communication network (ST35).

The server to which such information is sent stores machine learning models, including many types of intermediate learning models. Then, the server selects a suitable machine learning model based on the received information and transmits it to the extension module 2 (ST36).

Extension module 2 stores machine learning model 204 sent from the server in memory 201 (ST37).

After that, machine learning is performed on the machine learning model 204 in the same procedure as described in the first to third embodiments, and the parameters in the industrial equipment 1 and the internal state of the controlled equipment 3 can be estimated.

1 industrial device, 2 expansion module, 3 controlled device, 101 external terminal, 102
Servo amplifier, 103 control circuit, 104 processor, 105 AC/DC converter, 106 inverter, 107 memory, 108 I/O, 109 motor control unit, 110 learned machine learning model, 111 motor parameter, 112 identification information, 201 memory , 202 processor, 203 I/O, 204 machine learning model, 205
Trained machine learning model, 206 tables, 301 sensors.

Claims (4)

An expansion module having at least a processor and memory and connected to an external terminal of a servo controller,
When the memory does not store a machine learning model suitable for the assumed configuration, the processor is connected to the external terminal of the servo controller to control the servo controller from the servo controller. gather the information necessary to identify the machine learning model, including the displacement of the motor to be driven;
When the processor is detached from the external terminal of the servo controller and connected to an external server so as to be able to communicate information, the processor transmits information necessary to identify the machine learning model to the server, receiving an intermediate learning model from a server and storing it in said memory;
The extension module , wherein the processor is connected to the external terminal of the servo controller again to learn the intermediate learning model using the information obtained from the servo controller as teacher data. 2. The expansion module according to claim 1, wherein the intermediate learning model is a learning model that has been trained for infrequently occurring events. The infrequently occurring event is at least one of wear and aging caused by long-term operation of the motor, and at least one of wear and aging caused by long-term operation of the motor. 3. The expansion module according to claim 2, including at least one event selected from predictors of the resulting failure, and accidental events including at least one of foreign object bite and breakage. 4. The method according to any one of claims 1 to 3 , wherein the information necessary to identify the machine learning model includes information about the servo controller and information identifying a device controlled by the servo controller and its configuration. Extension module as described.

Images