State determination device and state determination method

Abstract

The present invention provides a state determination device and a state determination method. The state determination device acquires data related to an industrial machine, generates a plurality of partial time series data by sliding the time series data of physical quantities in the data related to the industrial machine in the direction of the time axis based on the acquired data, extracts a plurality of learning data including the plurality of partial time series data, and performs machine learning using the extracted learning data to generate a learning model.

Description

Technical Field

[0001] The present invention relates to a state determination device and a state determination method, and in particular to a state determination device and a state determination method for assisting maintenance of industrial machinery. Background Art

[0002] Industrial machinery such as injection molding machines are maintained periodically or when an abnormality occurs. When maintaining industrial machinery, the maintenance personnel use the physical quantity representing the operating state of the industrial machinery recorded when the industrial machinery is in motion to determine whether the operating state of the industrial machinery is abnormal, and perform maintenance work such as replacement of abnormal parts. Examples of industrial machinery include injection molding machines, machine tools, mining machinery, woodworking machinery, agricultural machinery, and construction machinery.

[0003] For example, as a maintenance operation for the check valve of the injection cylinder of an injection molding machine, which is a type of industrial machinery, a method of periodically pulling out the screw from the injection cylinder and directly measuring the size of the check valve is known. However, this method requires temporarily stopping production to perform the measurement operation, which has the problem of reduced production efficiency.

[0004] In order to solve such a problem, for example, as a method of diagnosing abnormalities by indirectly detecting the consumption of the check valve of the injection cylinder without temporarily stopping production due to operations such as pulling out the screw from the injection cylinder, there is a known method of diagnosing abnormalities by detecting the rotational torque applied to the screw, or detecting the phenomenon of resin flowing back to the rear of the screw.

[0005] For example, Japanese Patent Publication No. 01-168421 shows that the rotation torque acting on the screw is measured, and if the measured value exceeds the allowable range, it is determined to be abnormal. In addition, Japanese Patent Publication No. 2017-030221 and Japanese Patent Publication No. 2017-202632 show that abnormalities are diagnosed by supervised learning of the load of the drive unit or the resin pressure. And Japanese Patent Publication No. 2018-097616 shows a learning method for clustering feature vectors by performing machine learning on multiple time series data.

[0006] However, in machines with different specifications of elements constituting the drive part of industrial machines such as injection molding machines, there are various types of equipment constituting the machine, parts processed by the machine, etc., and there is a large deviation between the measured values ​​obtained by the machine and the numerical values ​​of the learning data input during machine learning, which causes a problem that the machine learning cannot correctly diagnose. For example, when the type of equipment constituting the movable part of the injection molding machine, the type of raw material of the molded product manufactured by the injection molding machine, i.e., the type of resin, or the type of ancillary equipment of the injection molding machine, i.e., the metal mold, metal mold temperature regulator, resin dryer, etc., is different from the learning conditions when the learning model is generated by machine learning, due to the influence of these differences, the measured values ​​obtained by the machine and the measured values ​​used when generating the learning model deviate, so the state of whether there is an abnormality cannot be correctly determined by machine learning.

[0007] Here, in order to improve the diagnostic accuracy of machine learning, there is a method of preparing a variety of learning conditions for machine learning when generating a learning model for machine learning. However, it is costly to gather a variety of injection molding machines, resins, and ancillary equipment for machine learning. Moreover, when operating the machine, it is also necessary to prepare raw materials such as resins and workpieces, and the cost of raw materials required to obtain learning data is also high. In addition, the operation of obtaining learning data requires a lot of time. Therefore, there is a problem of not being able to efficiently collect learning data. Summary of the invention

[0008] Therefore, there is a need for a state determination device and a state determination method that can efficiently perform machine learning based on measurement values ​​obtained from industrial machinery without incurring high costs and can assist in the maintenance of various industrial machinery using the learning results.

[0009] Therefore, in the state judgment device and method of the present invention, with respect to the time series data (current, speed, etc.) obtained from the industrial machinery, a plurality of learning data are generated by sliding (displacing) the time series data in the direction of the time axis in units of a predetermined number of data or in units of time, and machine learning is performed on the plurality of learning data generated by one time series data, thereby deriving a general learning model that avoids overlearning during machine learning, thereby achieving high-precision estimation of the action state and abnormality, thereby solving the above-mentioned problem.

[0010] A state determination device according to one embodiment of the present invention determines the operation state of an industrial machine, and comprises: a data acquisition unit, which acquires data related to the above-mentioned industrial machine; a learning data extraction unit, which generates a plurality of partial time series data by sliding the time series data of physical quantities in the data related to the above-mentioned industrial machine in the direction of the time axis based on the data related to the above-mentioned industrial machine acquired by the above-mentioned data acquisition unit, and extracts a plurality of learning data including the above-mentioned plurality of partial time series data; and a learning unit, which performs machine learning using the learning data extracted by the above-mentioned learning data extraction unit to generate a learning model.

[0011] The state determination device may further include an estimating unit configured to estimate the operating state of the industrial machine using the learning model generated by the learning unit.

[0012] The above-mentioned state judgment device may also have an extraction condition storage unit, which stores the conditions for the above-mentioned learning data extraction unit to extract multiple learning data as the number of data within a predetermined time or time series data range, and the above-mentioned multiple learning data include multiple partial time series data that make the time series data slide in the direction of the time axis.

[0013] The above-mentioned industrial machinery is an injection molding machine. In addition, the time series data obtained by the above-mentioned data acquisition unit includes at least one of the information for identifying the molding processes of the injection molding machine, namely, the mold closing process, the mold clamping process, the injection process, the pressure holding process, the metering process, the mold opening process, the ejection process, the cycle start, and the cycle end. In addition, it includes at least one of the current, voltage, torque, position, speed, acceleration of the prime mover driving the above-mentioned injection molding machine, and the pressure, temperature, flow rate, and flow velocity related to the molding action of the above-mentioned injection molding machine.

[0014] The learning unit may be at least one of supervised learning, unsupervised learning, and reinforcement learning.

[0015] The time-series physical quantity acquired by the data acquisition unit may be at least one of physical quantities possessed by a plurality of industrial machines connected via a wired or wireless network.

[0016] The estimating unit may estimate the degree of abnormality related to the operating state of the industrial machine, and the state determining device may display a warning message on a display device when the degree of abnormality estimated by the estimating unit exceeds a predetermined threshold value determined in advance.

[0017] The estimating unit may estimate the degree of abnormality related to the operating state of the industrial machine, and the state determining device may display a warning icon on a display device when the degree of abnormality estimated by the estimating unit exceeds a predetermined threshold value determined in advance.

[0018] The estimation unit can estimate the degree of abnormality related to the operating state of the industrial machinery. In addition, the state determination device can output at least one of the instructions of stopping, decelerating or limiting the torque of the prime mover to the industrial machinery when the degree of abnormality estimated by the estimation unit exceeds a predetermined threshold value determined in advance.

[0019] The prime mover driving the above industrial machinery is any one of an electric motor, a hydraulic cylinder, a hydraulic motor, and a pneumatic motor, and the transmission mechanism driving the above industrial machinery may include at least one of a ball screw, a gear, a pulley, and a transmission belt.

[0020] Another aspect of the state determination method of the present invention determines the operation state of an industrial machine, comprising: a data acquisition step of acquiring data related to the above-mentioned industrial machine; a learning data extraction step of generating a plurality of partial time series data by sliding the time series data of physical quantities in the data related to the above-mentioned industrial machine in the direction of the time axis based on the data related to the above-mentioned industrial machine acquired in the above-mentioned data acquisition step, and extracting a plurality of learning data including the above-mentioned plurality of partial time series data; and a learning step of performing machine learning using the learning data extracted in the above-mentioned learning data extraction step to generate a learning model.

[0021] The state determination method may further include an estimating step of estimating the operating state of the industrial machine using the learning model generated in the learning step.

[0022] The present invention has the above structure and efficiently uses one time series data, thereby reducing the work of collecting various time series data and efficiently collecting learning data. In addition, multiple learning data are generated from one time series data to perform machine learning, thereby improving the accuracy of machine learning judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a schematic hardware configuration diagram of a state determination device according to one embodiment.

[0024] Figure 2 This is a schematic functional block diagram of a state determination device according to one embodiment.

[0025] Figure 3 The generation of learning data by the learning data extraction unit will be described.

[0026] Figure 4 An example of a process in which the learning data extraction unit generates time-series data after sliding is shown.

[0027] Figure 5Another example of the process in which the learning data extraction unit generates the time-series data after sliding is shown.

[0028] Figure 6 An example of a display indicating an abnormal status. DETAILED DESCRIPTION

[0029] Figure 1 This is a schematic hardware configuration diagram showing main components of a state determination device including a machine learning device according to one embodiment.

[0030] The state determination device 1 of the present embodiment can be installed on a control device for controlling an industrial machine, or can be installed as a personal computer arranged in parallel with the control device for controlling the industrial machine, a management device 3 connected to the control device via a wired / wireless network, an edge computer, a fog computer, a cloud server, or other computer. Hereinafter, the state determination device 1 of the present embodiment is installed as a computer connected to a control device via a network, and the control device controls an injection molding machine as an industrial machine.

[0031] In addition, in the following embodiments, an injection molding machine is used as an example of an industrial machine for explanation, but the industrial machine that is the state judgment object of the state judgment device 1 of the present invention can be an injection molding machine, a machine tool, a robot, a mining machine, a woodworking machine, an agricultural machine, a construction machine, etc.

[0032] The CPU 11 included in the state determination device 1 of the present embodiment is a processor that controls the entire state determination device 1. The CPU 11 reads a system program stored in the ROM 12 via the bus 20, and controls the entire state determination device 1 according to the system program. The RAM 13 temporarily stores temporary calculation data, various data input by the operator via the input device 71, and the like.

[0033] The nonvolatile memory 14 is composed of, for example, a memory or SSD (Solid State Drive) backed by a battery (not shown), and maintains the storage state even when the power supply of the state determination device 1 is turned off. The nonvolatile memory 14 stores: a setting area for storing setting information related to the operation of the state determination device 1, data input by the input device 71, static data (machine model, quality and material of the metal mold, type of resin, etc.) obtained from the injection molding machine 2 via the network 7, time series data of physical quantities (nozzle temperature, position, speed, acceleration, current, voltage, torque, metal mold temperature, resin flow rate, flow velocity, pressure, etc.) detected during the molding operation of the injection molding machine 2, and data read from an external storage device (not shown) or other computers via the network 7. The program and various data stored in the nonvolatile memory 14 can be expanded in the RAM 13 when executed / used. In addition, system programs including a well-known analysis program for analyzing various data, a program for controlling exchanges with a machine learning device 100 described later, and the like are written in advance in the ROM 12 .

[0034] The state determination device 1 is connected to a wired / wireless network 7 via an interface 16. At least one injection molding machine 2 and a management device 3 for managing the production operation of the injection molding machine 2 are connected to the network 7 and exchange data with the state determination device 1.

[0035] The injection molding machine 2 is used to produce products molded by resins such as plastics. It is a machine that melts the resin as a material and fills (injects) it into a metal mold to perform molding. The injection molding machine 2 is composed of various equipment such as a nozzle, a prime mover (electric motor, etc.), a transmission mechanism, a speed reducer, and a movable part. The state of each part is detected by a sensor, and the movement of each part is controlled by a control device. As the prime mover used by the injection molding machine 2, for example, an electric motor, a hydraulic cylinder, a hydraulic motor, a pneumatic motor, etc. are used. In addition, as the transmission mechanism used by the injection molding machine 2, a ball screw, a gear, a pulley, a transmission belt, etc. are used.

[0036] The data read into the memory, the data obtained as a result of executing the program, the data output from the machine learning device 100 described later, etc. are output and displayed on the display device 70 via the interface 17. In addition, the input device 71 composed of a keyboard and a pointing device transmits instructions, data, etc. based on the operation performed by the operator to the CPU 11 via the interface 18.

[0037] The interface 21 is an interface for connecting the state determination device 1 to the machine learning device 100. The machine learning device 100 includes a processor 101 for comprehensively controlling the entire machine learning device 100, a ROM 12 storing system programs, etc., a RAM 103 for temporarily storing various processes related to machine learning, and a non-volatile memory 104 for storing learning models, etc. The machine learning device 100 can observe various information that can be obtained by the state determination device 1 through the interface 21 (for example, various data such as the model of the injection molding machine 2, the quality and material of the metal mold, the type of resin, the temperature of the nozzle, the position of the prime mover driving the nozzle, the speed, acceleration, current, voltage, torque, the temperature of the metal mold, the flow rate, flow velocity, pressure, etc. of the resin. Time series data of various physical quantities). In addition, the state determination device 1 obtains the processing results output from the machine learning device 100 through the interface 21, stores or displays the obtained results, or sends them to other devices through the network 7, etc.

[0038] Figure 2 This is a schematic functional block diagram of the state determination device 1 and the machine learning device 100 according to one embodiment.

[0039] The state determination device 1 of the present embodiment has a configuration required when the machine learning device 100 performs learning at the stage of performing machine learning. Figure 1 The CPU 11 of the state determination device 1 and the processor 101 of the machine learning device 100 execute their respective system programs to control the operation of each part of the state determination device 1 and the machine learning device 100, thereby achieving Figure 2 The individual functional blocks shown.

[0040] The state determination device 1 of this embodiment includes a data acquisition unit 30, a learning data extraction unit 32, a preprocessing unit 34, and a machine learning device 100, and the machine learning device 100 includes a learning unit 110 and an estimation unit 120. In addition, an acquisition data storage unit 50 for storing data acquired from an external machine and the like and an extraction condition storage unit 52 for storing conditions for extracting learning data from the acquired data are provided in the nonvolatile memory 14 of the state determination device 1. A learning model storage unit 130 is provided in the nonvolatile memory 104 of the machine learning device 100, and the learning model storage unit 130 stores a learning model constructed by machine learning of the learning unit 110.

[0041] The data acquisition unit 30 acquires various data input from the injection molding machine 2 and the input device 71. For example, the data acquisition unit 30 acquires static data such as the model of the injection molding machine 2, the quality and material of the metal mold, and the type of resin, time series data of various physical quantities such as the temperature of the nozzle, the position, speed, acceleration, current, voltage, torque, temperature of the metal mold, flow rate, flow velocity, and pressure of the prime mover driving the nozzle, information for identifying the molding process of the injection molding machine 2, namely, the mold closing process, the mold clamping process, the injection process, the pressure holding process, the metering process, the mold opening process, the ejection process, the cycle start, and the cycle end (the information acquired in association with time, so these are also referred to as a kind of time series data), and information related to the maintenance work of the injection molding machine input by the operator, and stores the various data in the acquired data storage unit 50. When acquiring the time series data, the data acquisition unit 30 makes the time series data acquired in a predetermined time range (for example, the range of the molding process of one cycle) into one time series data based on the signal data acquired from the injection molding machine 2 and the changes of other time series data, and stores it in the acquired data storage unit 50. The data acquisition unit 30 may acquire data from an external storage device (not shown) or from the management device 3 or other computers via the wired / wireless network 7 .

[0042] The learning data extraction unit 32 extracts data used for learning from the acquired data acquired by the data acquisition unit 30 (and stored in the acquired data storage unit 50) according to the extraction conditions stored in the extraction condition storage unit 52 at the stage of machine learning performed by the learning unit 110. The time width Wd (for example, the time that matches the range of a molding process of one cycle) of one time series data (partial time series data) to be extracted and the sliding amount Δt that causes the time series data to slide (displace) are pre-set in the extraction condition storage unit 52. The setting value of the sliding amount Δt is, for example, set to a value smaller than the time width Wd, or set to a time that matches the molding process of the injection molding machine 2, that is, the mold closing process, the mold clamping process, the injection process, the pressure holding process, the metering process, the mold opening process, and the ejection process. In addition, the setting unit of the sliding amount Δt can be set in time units or in data quantity units of the acquired data.

[0043] like Figure 3 As shown, the learning data extraction unit 32 generates a plurality of time series data after sliding the time series data included in the acquired data on the time axis for each acquired data stored in the acquired data storage unit 50, and extracts a plurality of acquired data including the generated plurality of time series data as learning data. The generation of time series data after sliding the time series data on the time axis here refers to the following steps: Figure 4As shown, for a series of target time series data, partial time series data is generated with a time width Wd in which the start time is shifted by a predetermined sliding amount Δt.

[0044] The acquired data includes, for example, static data that does not change over time and time series data that records changes with time. The learning data extraction unit 32 generates a plurality of partial time series data after sliding on the time axis from the time series data, and extracts a plurality of acquired data by combining the partial time series data with the static data.

[0045] For example, when the acquired data (FN-1, RE1, ECi) including the model name: FN-1, the resin type: RE1 as static data, and the current: ECi as time series data is used as the extraction object of the learning data, when the extraction condition stored in the extraction condition storage unit 52 is set to generate partial time series data after each sliding by the sliding amount Δt with the time width Wd, the learning data extraction unit 32 generates partial time series data ECi1 to ECi2 in which the time series data ECi slides on the time axis by the time width Wd of Δt each time. n , extract the n data (FN-1, RE1, ECi1)~(FN-1, RE1, ECi1) after combining these n partial time series data with the static data FN-1 and RE1 respectively n ) as learning data.

[0046] As another example, when the acquired data (FN-1, RE1, ECi, PR) including the model name: FN-1, the resin type: RE1 as static data, and the current: ECi, the pressure: PR as time series data is used as the extraction object of the learning data, when the extraction condition stored in the extraction condition storage unit 52 is set to generate partial time series data after each sliding amount Δt with a time width Wd, the learning data extraction unit 32 (1) generates partial time series data ECi1 to ECi2 with a time width Wd after each sliding amount Δt on the time axis based on the time series data ECi. n , and (2) generate partial time series data PR1 to PR1 by sliding the time series data PR on the time axis by a time width Wd of Δt each time n , (3) extracting n data (FN-1, RE1, ECi1, PR1), (FN-1, RE1, ECi2, PR2) ~ (FN-1, RE1, ECi n , P.R. n ) as learning data.

[0047] In this way, when the acquired data includes a plurality of time series data, the learning data is generated by grouping the time series data that have slid the same amount of time series data as a group for the partial time series data generated from each time series data. This means that when a plurality of time series data are included, the changes of each time series data at the same time are learned.

[0048] In addition, if Figure 5 As shown, the extraction condition storage unit 52 may further include an extraction start position St of partial time series data generated based on the time series data included in the acquired data. The extraction start position St may be set, for example, by the start timing of a predetermined process or cycle in the operation of the injection molding machine 2, or may be set by adding a predetermined time width Δt to the start timing of the process or cycle. d The extraction start position.

[0049] By setting the extraction start position St of the partial time series data in accordance with the time width Wd and the sliding amount Δt of the partial time series data, for example, a plurality of partial time series data including waveforms showing a predetermined process (for example, Figure 4 Injection process in which the waveform of the current value vibrates up and down).

[0050] The preprocessing unit 34 generates learning data used in the learning of the machine learning device 100 based on the learning data extracted by the learning data extraction unit 32 during the machine learning stage of the machine learning device 100. The preprocessing unit 34 generates learning data by converting (digitizing, sampling, etc.) the data input from the learning data extraction unit 32 into a uniform format for processing in the machine learning device 100. For example, when the machine learning device 100 performs unsupervised learning, the preprocessing unit 34 generates state data S in a predetermined format for the learning as learning data, when the machine learning device 100 performs supervised learning, the preprocessing unit 34 generates a combination of the state data S in a predetermined format for the learning and label data L as learning data, and when the machine learning device 100 performs reinforcement learning, the preprocessing unit 34 generates a combination of the state data S in a predetermined format for the learning and judgment data D as learning data.

[0051] In addition, during the estimation stage of the machine learning device 100, the preprocessing unit 34 converts (digitizes, samples, etc.) the acquired data acquired by the data acquisition unit 30 (and stored in the acquired data storage unit 50) into a unified form for processing in the machine learning device 100, and generates state data S in a predetermined form used in the estimation of the machine learning device 100.

[0052] The learning unit 110 of the machine learning device 100 performs machine learning using the learning data generated by the preprocessing unit 34 based on the learning data extracted by the learning extraction unit 32. The learning unit 110 performs machine learning using the data obtained from the injection molding machine 2 by a well-known machine learning method such as unsupervised learning, supervised learning, and reinforcement learning, thereby generating a learning model, and storing the generated learning model in the learning model storage unit 130. As the unsupervised learning method performed by the learning unit 110, for example, an autoencoder method and a k-means method are listed, as the supervised learning method, for example, a multilayer perceptron method, a recurrent neural network method, a Long Short-Term Memory method, a convolutional neural network method, etc. are listed, and as the reinforcement learning method, for example, Q learning is listed.

[0053] The learning unit 110 can generate the distribution of data acquired in a normal state as a learning model by, for example, performing unsupervised learning based on learning data obtained by processing the acquired data acquired from the injection molding machine 2 in a normal operating state by the learning data extraction unit 32 and the preprocessing unit 34. The estimation unit 120 described later can use the learning model generated in this way to estimate how much the state data S acquired by processing the acquired data acquired from the injection molding machine 2 by the preprocessing unit 34 deviates from the state data acquired when the injection molding machine 2 is operating in a normal state, and calculate the abnormality as the estimation result.

[0054] In addition, the learning unit 110 assigns a normal label to the acquired data acquired from the injection molding machine in a normal operating state, and assigns an abnormal label to the acquired data acquired from the injection molding machine 2 before and after the abnormality occurs, and performs supervised learning on the acquired data obtained by processing the acquired data using the learning data extraction unit 32 and the preprocessing unit 34, so that the discrimination boundary between normal data and abnormal data can be generated as a learning model. The estimation unit 120 described later uses the learning model generated in this way, inputs the state data S obtained by processing the acquired data acquired from the injection molding machine 2 by the preprocessing unit 34 into the learning model, and can estimate whether the state data S belongs to normal data or abnormal data, and calculate the label value (normal / abnormal) and its reliability as the estimation result.

[0055] The estimation unit 120 of the machine learning device 100 estimates the state of the injection molding machine using the learning model stored in the learning model storage unit 130 based on the state data S generated by the preprocessing unit 34. In the estimation unit 120 of the present embodiment, the state data S obtained by the preprocessing unit 34 is input to the learning model generated (with parameters determined) by the learning unit 110, thereby estimating and calculating the abnormality related to the state of the injection molding machine, or estimating and calculating the class (normal / abnormal, etc.) to which the operating state of the injection molding machine belongs. The result estimated by the estimation unit 120 (the abnormality related to the state of the injection molding machine, the class to which the operating state of the injection molding machine belongs, etc.) is displayed and output on the display device 70, or sent and output to a host computer or a cloud computer via a wired / wireless network (not shown). In addition, when the result estimated by the estimation unit 120 is a predetermined state (for example, when the abnormality estimated by the estimation unit 120 exceeds a predetermined threshold value, when the operation state of the injection molding machine estimated by the estimation unit 120 is classified as "abnormal", etc.), the state determination device 1 Figure 6 As shown, a warning message or an icon may be displayed and output on the display device 70, and an instruction to stop the operation, slow down the operation, or limit the torque of the prime mover may be output to the injection molding machine.

[0056] In the state determination device 1 having the above-mentioned structure, the learning data extraction unit 32 generates a plurality of partial time series data by sliding the time series data included in the acquired data acquired from the injection molding machine on the time axis according to the extraction conditions stored in the extraction condition storage unit 52, thereby generating a plurality of learning data from one acquired data. In this way, a large amount of learning data can be generated from a predetermined number of acquired data obtained from the limited operation of the injection molding machine 2, so that the learning unit 110 included in the machine learning device 100 can efficiently promote learning to assist the maintenance of various industrial machines without incurring high costs, and can generate a learning model that can flexibly cope with waveform deviations in the time axis direction.

[0057] The state determination device 1 of the present embodiment can be applied to the case of determining the state of industrial machinery such as robots or machine tools, but can be appropriately used for industrial machinery that performs unstable behavior, such as when the operation starts or when the operating conditions change. In particular, regarding the operation of an injection molding machine, even when operating under the same injection conditions, the operation of the machine is sometimes delayed depending on the internal or external state of the machine. Even in such a case, the molding action of the injection molding machine itself is normal, so in order not to determine such data as abnormal, learning data for learning as normal action is required. Even if the state determination device 1 of the present embodiment does not specifically obtain data when a delay occurs in the operation of the machine, it can generate a plurality of learning data by sliding the time series data based on the generally obtainable acquisition data, so it is particularly useful for the state determination of the injection molding machine.

[0058] Although the embodiment of the present invention has been described above, the present invention is not limited to the examples of the above-described embodiment, and can be implemented in various forms by adding appropriate changes.

[0059] For example, in the above embodiment, the state determination device 1 and the machine learning device 100 are described as devices having different CPUs (processors), but the machine learning device 100 may be implemented by the CPU 11 and the system program stored in the ROM 12 of the state determination device 1. In addition, when a plurality of injection molding machines 2 are connected to each other via a network, the operation states of the plurality of injection molding machines may be determined by one state determination device 1, and the state determination device 1 may be installed on the control device of the injection molding machine.

Claims

1. A state determination device for determining the operating state of an industrial machine, characterized in that: The above-mentioned state determination device comprises: a data acquisition unit configured to acquire data related to the industrial machine, wherein the data related to the industrial machine includes static data and time series data of physical quantities; a learning data extraction unit that generates, based on the data related to the industrial machinery acquired by the data acquisition unit, a plurality of partial time series data in which the time series data of the physical quantity in the data related to the industrial machinery is slid in a time axis direction by a predetermined sliding amount at a predetermined time width, and extracts a plurality of learning data in which the plurality of partial time series data are respectively combined with the static data; a learning unit that performs machine learning using the learning data extracted by the learning data extraction unit to generate a learning model; and an estimating unit that estimates the operating state of the industrial machine using the learning model generated by the learning unit, The above industrial machinery is an injection molding machine, The data acquisition unit acquires time series data including at least one of information for identifying the molding processes of the injection molding machine, namely, a mold closing process, a mold clamping process, an injection process, a pressure holding process, a metering process, a mold opening process, an ejection process, a cycle start, and a cycle end. The time series data of the physical quantity includes at least one of the current, voltage, torque, position, speed, acceleration of the prime mover driving the injection molding machine, and the pressure, temperature, flow rate, and flow velocity related to the molding action of the injection molding machine.

2. The state determination device according to claim 1, characterized in that: The state determination device further includes an extraction condition storage unit that stores the condition under which the learning data extraction unit extracts the plurality of learning data as the number of data within a predetermined time width or a range of time-series data.

3. The state determination device according to claim 1, characterized in that: The learning unit is at least one of supervised learning, unsupervised learning, and reinforcement learning.

4. The state determination device according to claim 1, characterized in that: The time-series physical quantity acquired by the data acquisition unit is at least one of the physical quantities possessed by a plurality of industrial machines connected via a wired or wireless network.

5. The state determination device according to claim 1, characterized in that: The estimating unit estimates a degree of abnormality related to an operating state of the industrial machine. The state determination device displays a warning message on a display device when the abnormality degree estimated by the estimation unit exceeds a predetermined threshold value determined in advance.

6. The state determination device according to claim 1, characterized in that: The estimating unit estimates a degree of abnormality related to an operating state of the industrial machine. The state determination device displays a warning icon on a display device when the abnormality degree estimated by the estimation unit exceeds a predetermined threshold value determined in advance.

7. The state determination device according to claim 1, characterized in that: The estimating unit estimates a degree of abnormality related to an operating state of the industrial machine. The state determination device outputs at least one of a command to stop operation, decelerate operation, or limit torque of a prime mover to the industrial machine when the abnormality estimated by the estimation unit exceeds a predetermined threshold value determined in advance.

8. The state determination device according to claim 1, characterized in that: The prime mover driving the above industrial machinery is one of an electric motor, a hydraulic cylinder, a hydraulic motor, and a pneumatic motor. The transmission mechanism for driving the above industrial machinery includes at least one of a ball screw, a gear, a pulley, and a transmission belt.

9. A state determination method for determining the operating state of an industrial machine, characterized in that: The above-mentioned state determination method includes: A data acquisition step of acquiring data related to the industrial machinery, wherein the data related to the industrial machinery includes static data and time series data of physical quantities; a learning data extraction step, based on the data related to the industrial machinery acquired in the data acquisition step, generating a plurality of partial time series data in which the time series data of the physical quantity in the data related to the industrial machinery is slid in the time axis direction by a predetermined sliding amount at a predetermined time width, and extracting a plurality of learning data in which the plurality of partial time series data are respectively combined with the static data; a learning step of performing machine learning using the learning data extracted in the learning data extraction step to generate a learning model; and an estimating step of estimating the operating state of the industrial machine using the learning model generated in the learning step, The above industrial machinery is an injection molding machine, In the data acquisition step, time series data including at least one of information for identifying the molding processes of the injection molding machine, namely, mold closing process, mold clamping process, injection process, pressure holding process, metering process, mold opening process, ejection process, cycle start and cycle end are acquired. The time series data of the physical quantity includes at least one of the current, voltage, torque, position, speed, acceleration of the prime mover driving the injection molding machine, and the pressure, temperature, flow rate, and flow velocity related to the molding action of the injection molding machine.

Images