State determination device and state determination method

Abstract

The present application provides a state determination device and a state determination method. The state determination device acquires data of an industrial machine, extracts data used for a machine learning-related process from the acquired data in accordance with an extraction condition for extracting the data used for the machine learning-related process, and executes the machine learning-related process using the extracted data.

Description

Technical Field

[0001] This invention relates to a state determination device and a state determination method, and particularly to a state determination device and a state determination method for assisting in the maintenance of injection molding machines. Background Technology

[0002] Regular maintenance is performed on industrial machinery such as injection molding machines, or when abnormalities occur. During maintenance, the maintenance supervisor uses physical quantities recorded during machine operation to indicate its operational status, and determines whether any abnormalities exist, performing maintenance tasks such as replacing faulty components.

[0003] For example, a known method for maintaining the check valve of the injection cylinder in an injection molding machine, a type of industrial machinery, involves periodically pulling the screw out of the injection cylinder and directly measuring the size of the check valve. However, this method requires temporarily stopping production to perform the measurement, thus resulting in a decrease in production efficiency.

[0004] To address this problem, a method is known to indirectly detect the wear of the check valve in the injection cylinder and diagnose abnormalities without temporarily halting production by removing the screw from the injection cylinder. This method involves detecting the rotational torque applied to the screw or detecting the phenomenon of resin flowing backwards from the screw.

[0005] For example, Japanese Patent Application Publication No. 01-168421 discloses a method for measuring the rotational torque of a screw and determining an abnormality when the measured value exceeds the allowable range. Furthermore, Japanese Patent Application Publication Nos. 2017-030221 and 2017-202632 disclose methods for diagnosing abnormalities by performing supervised learning on the load of the drive unit or resin pressure, etc. Further, Japanese Patent Application Publication Nos. 2018-097616 and 2017-188030 disclose methods for performing machine learning using time series data.

[0006] However, in machines where the various components of the drive unit of an injection molding machine differ, there may be significant discrepancies between the measured values ​​obtained from that machine and the values ​​of the learning data input during machine learning, leading to problems with accurate machine learning-based diagnostics. Furthermore, if the type of raw material (resin) used to produce the molded product, or the auxiliary equipment (metal mold, metal mold temperature controller, resin dryer, etc.) of the injection molding machine, differs from the types used in machine learning, problems will arise in the accuracy of machine learning-based diagnostics.

[0007] To address this issue and improve the diagnostic accuracy of machine learning, a variety of learning conditions are needed when generating machine learning models. However, collecting diverse injection molding machines, resins, and auxiliary equipment for machine learning is costly. Furthermore, operating the machinery requires preparing raw materials such as resin and workpieces, and the cost of these raw materials for obtaining learning data is also substantial. Additionally, acquiring learning data takes considerable time. Therefore, there is a problem of ineffective learning data collection.

[0008] Here, when changing the raw material (resin) used in the injection molding machine, replacing the auxiliary equipment (metal mold), starting the operation of peripheral equipment such as the metal mold temperature controller or resin dryer, or changing the operating conditions related to the injection molding machine, such as injection and holding pressure conditions, or when the injection molding machine is in an abnormal operation alarm state, there may be situations where the time series data obtained from the injection molding machine is not suitable for machine learning. However, in the existing technology, there are problems with using inappropriate learning data to derive a learning model or diagnose inappropriate learning data, which may prevent the accurate diagnosis of the machine's operating state. Summary of the Invention

[0009] Therefore, it is desirable to be able to easily filter out inappropriate learning data to perform machine learning with good accuracy, and to use the learning results to assist in the condition determination devices and methods for the maintenance of various industrial machinery.

[0010] Therefore, in the state determination device and method of the present invention, the learning data input into the machine learning is used to perform machine learning on time series data that have changed or are unstable in the operating state and operation state of the injection molding machine, such as time series data in alarms, time series data after the start of machine operation or after the metal mold is changed, or time series data after the setting values ​​of molding conditions such as injection conditions or holding pressure conditions of machine operation are changed, thereby deriving a high-precision learning model and solving the above-mentioned problems.

[0011] One aspect of the present invention provides a state determination apparatus for determining the operating state of industrial machinery, comprising: a data acquisition unit that acquires data of the industrial machinery; an extraction condition storage unit that stores extraction conditions for extracting data used in machine learning-related processing from the data acquired by the data acquisition unit; a learning data extraction unit that extracts data used in machine learning-related processing from the data acquired by the data acquisition unit according to the extraction conditions stored in the extraction condition storage unit; and a machine learning apparatus that performs machine learning-related processing using the data extracted by the learning data extraction unit.

[0012] The aforementioned machine learning apparatus includes a learning unit that performs machine learning using data extracted by the learning data extraction unit and generates a learning model. Furthermore, the learning unit is capable of performing at least one of supervised learning, unsupervised learning, and reinforcement learning.

[0013] The aforementioned machine learning apparatus may include: a learning model storage unit that stores a learning model generated by machine learning using data extracted by the aforementioned learning data extraction unit; and an estimation unit that estimates the state of the aforementioned industrial machinery using the aforementioned learning model based on the data extracted by the aforementioned learning data extraction unit.

[0014] The aforementioned estimation unit estimates the degree of abnormality related to the operating state of the aforementioned industrial machinery. Furthermore, when the degree of abnormality estimated by the aforementioned estimation unit exceeds a predetermined threshold, the aforementioned state determination device displays a warning message on the display device.

[0015] The aforementioned estimation unit estimates the degree of abnormality related to the operating state of the aforementioned industrial machinery. Furthermore, when the degree of abnormality estimated by the aforementioned estimation unit exceeds a predetermined threshold, the aforementioned state determination device displays a warning icon on the display device.

[0016] The aforementioned estimation unit estimates the degree of abnormality related to the operating state of the industrial machinery. Furthermore, when the degree of abnormality estimated by the aforementioned estimation unit exceeds a predetermined threshold, the aforementioned state determination device outputs at least one of the following commands to the industrial machinery: stop operation, decelerate, or limit the torque of the prime mover.

[0017] The aforementioned industrial machinery is an injection molding machine. Furthermore, the data acquired by the aforementioned data acquisition unit includes at least one of the following information: information indicating any one of the mechanical states of the injection molding machine: running, stopped, heating up, heating finished, metal mold changing, metal mold changing finished, alarm in progress, or production finished; information indicating any one of the operating states of the injection molding machine: whether the injection conditions, holding pressure conditions, metering conditions, mold opening and closing conditions, ejection conditions, or temperature conditions have changed; and information indicating any one of the molding processes of the injection molding machine: mold closing process, mold closing process, injection process, holding pressure process, metering process, mold opening process, ejection process, or standby process.

[0018] The data acquired by the aforementioned data acquisition unit includes at least one of the data acquired from multiple industrial machines connected via wired or wireless networks.

[0019] The machine learning method in the state determination apparatus for acquiring data of industrial machinery according to another aspect of the present invention includes: a data acquisition step for acquiring the data of the industrial machinery; a learning data extraction step for extracting data used in machine learning-related processing from the data acquired in the data acquisition step according to extraction conditions for extracting data used in machine learning-related processing from the data acquired from the industrial machinery; and a step for performing machine learning-related processing using the data extracted in the learning data extraction step.

[0020] By having the above-described structure, this invention can exclude data obtained when the operating state or operational state of industrial machinery changes, and data obtained under unstable forming conditions, from the learning data for machine learning, and can be expected to improve the judgment accuracy of machine learning. Attached Figure Description

[0021] Figure 1 This is a schematic hardware structure diagram of a state determination device for one implementation method.

[0022] Figure 2 This is a schematic functional block diagram of the state determination device according to the first embodiment.

[0023] Figure 3 Examples of extraction conditions.

[0024] Figure 4 This represents an example of extracting learning data from the learning data extraction department.

[0025] Figure 5 Other examples of data extraction for learning purposes in the learning data extraction department.

[0026] Figure 6 Other examples of data extraction used in the learning data extraction department.

[0027] Figure 7 This is a schematic functional block diagram of the state determination device according to the second embodiment.

[0028] Figure 8 An example of displaying an abnormal state. Detailed Implementation

[0029] Figure 1 This is a schematic hardware structure diagram of the main components of a state determination device for a machine device in one embodiment.

[0030] The state determination device 1 of this embodiment can be installed, for example, on a control device that controls industrial machinery. Alternatively, it can be installed as a personal computer set up alongside the control device, 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, an example will be described where the state determination device 1 of this embodiment is installed as a computer connected via a network to a control device that controls an injection molding machine, which is industrial machinery. Furthermore, in the following embodiments, an injection molding machine is used as an example of industrial machinery; however, the industrial machinery to which the state determination device 1 of the present invention is applied can include injection molding machines, machine tools, robots, mining machinery, woodworking machinery, agricultural machinery, construction machinery, etc.

[0031] The CPU 11 included in the state determination device 1 of this embodiment is the processor that controls the entire state determination device 1. The CPU 11 reads the 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, etc.

[0032] The non-volatile memory 14 is composed of, for example, a memory that is backed up by a battery (not shown) or an SSD (Solid State Drive), and maintains its storage state even when the power supply to the state determination device 1 is cut off. The non-volatile memory 14 stores setting area or input data from the input device 71 that stores setting information related to the operation of the state determination device 1, static data (machine type, 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 detected during the molding operation of the injection molding machine 2 (nozzle temperature, position, speed, acceleration, current, voltage, torque of the prime mover driving the nozzle, temperature of the metal mold, resin flow rate, flow velocity, pressure, etc.), time series data of information indicating the operation state or operating state of the injection molding machine 2 (identifying the mold closing process, mold closing process, injection process, pressure holding process, metering process, mold opening process, ejection process, cycle start, cycle end information, information indicating the alarm generation state, etc.), and data read from an external storage device (not shown) or from other computers via the network 7, etc. The program and various data stored in the non-volatile memory 14 can be expanded onto the RAM 13 during execution / use. In addition, a system program containing a known analysis program for analyzing various data and a program for controlling the exchange with the machine learning device 100 described later is pre-written into the ROM 12.

[0033] The status determination device 1 is connected to a wired / wireless network 7 via interface 16. At least one injection molding machine 2 and a management device 3 for managing the production operations of the injection molding machine 2 are connected to the network 7 and exchange data with the status determination device 1.

[0034] An injection molding machine 2 is a machine used to produce products made from resins such as plastics. It melts the resin and injects it into a metal mold to form the product. The injection molding machine 2 consists of various components, including a nozzle, a prime mover (electric motor, etc.), a transmission mechanism, a reducer, and movable parts. The status of each part is detected by sensors, and the operation of each part is controlled by a control device. The prime mover used in the injection molding machine 2 can be, for example, an electric motor, a hydraulic cylinder, a hydraulic electric motor, or a pneumatic electric motor. Furthermore, the transmission mechanism used in the injection molding machine 2 can be a ball screw, gear, pulley, belt, etc.

[0035] Various data read into the memory, data obtained as a result of executing programs, and data output from the machine learning device 100 (described later) are output to the display device 70 via the interface 17 and displayed. In addition, the input device 71, which consists of a keyboard and pointing devices, transmits instructions and data based on the operator's operation to the CPU 11 via the interface 18.

[0036] Interface 21 is used to connect the state determination device 1 and the machine learning device 100. The machine learning device 100 includes a processor 101 for unified control of the entire machine learning device 100, a ROM 12 storing system programs, RAM 103 for temporary storage in various machine learning-related processes, and non-volatile memory 104 for storing learning models, etc. The machine learning device 100 can observe various information obtainable by the state determination device 1 via interface 21 (e.g., various data such as the type of injection molding machine 2, the quality and material of the metal model, the type of resin, the nozzle temperature, the position, speed, acceleration, current, voltage, torque of the prime mover driving the nozzle, the temperature of the metal model, the flow rate, velocity, pressure, etc. of the resin, and time series data of information representing the operating state and working state of the injection molding machine 2, etc.). In addition, the state determination device 1 obtains the processing results output from the machine learning device 100 via interface 21, stores and displays the obtained results, and sends them to other devices via network 7, etc.

[0037] Figure 2 This is a schematic functional block diagram of the state determination device 1 and the machine learning device 100 in the first embodiment.

[0038] The state determination device 1 of this embodiment has the structure (learning mode) required when the machine learning device 100 learns during the machine learning phase. Figure 1 The CPU 11 of the state determination device 1 and the processor 101 of the machine learning device 100 shown execute various system programs and control the various actions of the state determination device 1 and the machine learning device 100, thereby realizing... Figure 2 The various functional blocks shown.

[0039] 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, which includes a learning unit 110. Furthermore, the non-volatile memory 104 of the state determination device 1 includes an acquisition data storage unit 50 for storing data acquired from external machinery or the like, and an extraction condition storage unit 52 for storing conditions for extracting learning data from the acquired data. The non-volatile memory 104 of the machine learning device 100 includes a learning model storage unit 130 for storing the learning model constructed by the machine learning process of the learning unit 110.

[0040] The data acquisition unit 30 acquires various data input from the injection molding machine 2 and the input device 71. The data acquisition unit 30 acquires static data such as the type of injection molding machine 2, the quality and material of the metal mold, and the type of resin; nozzle temperature; position, speed, acceleration, current, voltage, and torque of the prime mover driving the nozzle; time series data of various physical quantities related to the molding action of the injection molding machine 2, such as the temperature of the metal mold, the flow rate, flow velocity, and pressure of the resin; and information on the mechanical status of the injection molding machine 2, such as running, stopping, heating up, heating up finished, metal mold changing in progress, metal mold changing finished, alarm in progress, and production finished. It also identifies information indicating the operating status of the injection molding machine 2, such as whether there have been changes in injection conditions, holding pressure conditions, metering conditions, mold opening and closing conditions, ejection conditions, and temperature conditions. Furthermore, it identifies information on the molding processes of the injection molding machine 2, including mold closing process, mold closing process, injection process, holding pressure process, metering process, mold opening process, ejection process, standby process, cycle start, and cycle end, as well as information indicating the alarm generation status and maintenance-related information input by the operator. All of this data is stored in the data acquisition and storage unit 50. When acquiring time series data, the data acquisition unit 30 stores the time series data acquired within a predetermined time range (e.g., the range of one cycle of molding operations) in the acquisition data storage unit 50 based on changes in signal data and other time series data acquired from the injection molding machine 2. The data acquisition unit 30 can acquire data from the management device 3 or other computers via an external storage device not shown or via a wired / wireless network 7.

[0041] The learning data extraction unit 32 extracts the acquisition data used for machine learning from the acquisition data acquired by the data acquisition unit 30 (and stored in the acquisition data storage unit 50) based on the extraction conditions stored in the extraction condition storage unit 52 for the stage of machine learning performed in the learning unit 110. In other words, the learning data extraction unit 32 excludes acquisition data that is not suitable for machine learning from the acquisition data acquired by the data acquisition unit 30 based on the extraction conditions stored in the extraction condition storage unit 52.

[0042] Figure 3 The extraction conditions stored in the extraction condition storage unit 52 are shown as an example.

[0043] The extraction condition storage unit 52 stores at least one extraction condition that is organized and managed, for example, according to condition categories. The extraction conditions stored in the extraction condition storage unit 52 can be conditions specifying the acquired data used in machine learning, or conditions specifying acquired data that is not used (excluded) in machine learning. The extraction conditions stored in the extraction condition storage unit 52 include conditions for classifying the acquired data based on predetermined data values ​​contained in the acquired data, and a specification of whether to use the acquired data classified by the conditions as learning data.

[0044] Figure 4 This section describes an example of how data is extracted by the learning data extraction unit 32 based on the extraction conditions of the mechanical state stored in the extraction condition storage unit 52.

[0045] Consider when Figure 4 When the acquired data is stored in the acquired data storage unit 50, the learning data extraction unit 32 extracts the waveform data of the current value for each cycle as learning data. At this time, if the extraction condition "remove the acquired data in the alarm from the learning data" is set in the extraction condition storage unit 52, the learning data extraction unit 32 will operate by not extracting the current value data acquired in that cycle as learning data when an alarm is generated during the forming process cycle. Specifically, in Figure 4 In the example case, the learning data extraction unit 32 does not extract the current value data obtained in the (i+2)th and (i+3)th cycles when the alarm is detected as learning data, but extracts the current value data obtained before the (i+1)th cycle and after the (i+4)th cycle as learning data.

[0046] Figure 5 This section describes an example of how learning data extraction unit 32 extracts data based on extraction conditions of the operation state stored in extraction condition storage unit 52.

[0047] Consider storing in Data Storage Unit 50 Figure 5In the illustrated case of acquiring data, the learning data extraction unit 32 extracts waveform data of the voltage value for each cycle as learning data. At this time, if the extraction condition "removing the data acquired in the 10 cycles starting from the injection condition change from the learning data" is set in the extraction condition storage unit 52, then when the injection condition is changed during the molding process cycle (the injection condition change signal is on), the learning data extraction unit 32 operates by not using the voltage value data acquired during the 10 cycles starting from the cycle in which the change occurred as learning data. Specifically, in Figure 5 In the example case, the learning data extraction unit 32 does not extract the voltage value data obtained from the (i+1)th cycle after the change of injection conditions (up to the (i+10)th cycle) as learning data, but extracts the current value data obtained before the i-th cycle and after the (i+11)th cycle as learning data.

[0048] Figure 6 This section describes an example of how data is extracted by the learning data extraction unit 32 based on the extraction conditions of the forming process stored in the extraction condition storage unit 52.

[0049] Consider when Figure 6 When the acquired data is stored in the acquired data storage unit 50, the learning data extraction unit 32 extracts the waveform data of the current value for each cycle as learning data. At this time, if the extraction condition "only extracts the acquired data from the injection and holding pressure processes as learning data" is set in the extraction condition storage unit 52, the learning data extraction unit 32 operates by extracting the current value data acquired during the injection and holding pressure processes in each molding process as learning data. Specifically, in Figure 6 In the case of the example, the learning data extraction unit 32 determines the duration of the injection process and the holding pressure process in each molding process based on the start signal and end signal of each process, and extracts the current value data obtained during the period as learning data.

[0050] Multiple extraction conditions can be set in the extraction condition storage unit 52. In this case, conflicts may arise regarding whether to use more than two extraction conditions as learning data. The learning data extraction unit 32 can then prioritize not using these conditions as learning data. The extraction condition storage unit 52 stores the extraction conditions and their priority order, allowing it to resolve conflicts based on this priority order.

[0051] During the machine learning phase of the machine learning device 100, the preprocessing unit 34 generates learning data used by the machine learning device 100 based on the learning data extracted by the learning data extraction unit 32. The preprocessing unit 34 generates learning data that transforms (numerizes, samples, etc.) the data input from the learning data extraction unit 32 into a unified form that is processed 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 form as learning data; when the machine learning device 100 performs supervised learning, it generates a set of state data S in a predetermined form and label data L as learning data; and when the machine learning device 100 performs reinforcement learning, it generates a set of state data S in a predetermined form and decision data D as learning data.

[0052] The learning unit 110 in the machine learning device 100 performs machine learning using the learning data extracted by the learning data extraction unit 32 and the learning data generated by the preprocessing unit 34. The learning unit 110 performs machine learning using the data obtained from the injection molding machine 2 through known machine learning methods 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. Examples of unsupervised learning methods performed by the learning unit 110 include autoencoder and k-means methods; examples of supervised learning methods include multilayer perceptron, recurrent neural network, long short-term memory, and convolutional neural network methods; and examples of reinforcement learning methods include Q-learning.

[0053] The learning unit 110, for example, performs unsupervised learning based on learning data, and can generate the distribution of data obtained under normal conditions as a learning model. The learning data is obtained by the learning data extraction unit 32 and the preprocessing unit 34 processing the data obtained from the injection molding machine 2 under normal operating conditions.

[0054] In addition, the learning unit 110 assigns a normal label to the data obtained from the injection molding machine in a normal operating state, and assigns an abnormal label to the data obtained from the injection molding machine 2 before and after an abnormality occurs. Supervised learning is performed using the learning data obtained by the learning data extraction unit 32 and the preprocessing unit 34, and a discrimination boundary between normal data and abnormal data is generated as a learning model.

[0055] In the state determination device 1 of the first embodiment with the above-described structure, regarding the acquired data obtained from the injection molding machine 2, the learning data extraction unit 32 extracts learning data from the acquired data included in the acquired data according to the extraction conditions stored in the extraction condition storage unit 52. The extraction condition storage unit 52 can set extraction conditions so that the operator can extract appropriate data as learning data in accordance with the purpose of machine learning at that time. In this way, the learning data extracted by the learning data extraction unit 32, such as time series data excluding alarm time series data, time series data after machine operation has started or after the metal mold has been changed, or time series data after the setting values ​​of molding conditions such as injection and holding pressure conditions of machine operation have been changed, etc., which are time series data of states that change or are unstable in the operating state or operation state of the injection molding machine, can be used for machine learning only on the time series data of predetermined processes required for determining the operating state. Compared with the case where a learning model generated by conventional methods is used, the state determination accuracy of the injection molding machine 2 using such a learning model is expected to be improved.

[0056] Figure 7 This is a schematic functional block diagram of the state determination device 1 and the machine learning device 100 in the second embodiment.

[0057] The state determination device 1 of this embodiment has the structure (estimation mode) required for the machine learning device 300 to perform estimation. Figure 1 The CPU 11 of the state determination device 1 and the processor 101 of the machine learning device 100 shown execute various system programs to control the actions of each part of the state determination device 1 and the machine learning device 300, thereby realizing... Figure 7 The various functional blocks shown.

[0058] The state determination device 1 of this embodiment, like the first embodiment, includes a data acquisition unit 30, a learning data extraction unit 32, a preprocessing unit 34, and a machine learning device 100, which includes an estimation unit 120. Furthermore, the non-volatile memory 14 of the state determination device 1 is provided with an acquisition data storage unit 50 for storing data acquired from external machinery, etc., and an extraction condition storage unit 52 for storing conditions for extracting learning data from the acquired data. The non-volatile memory 14 of the machine learning device 100 is provided with a learning model storage unit 130 for storing the learning model constructed by the machine learning unit 110.

[0059] The data acquisition unit 30 of this embodiment has the same function as the data acquisition unit 30 of the first embodiment.

[0060] The basic operation of the learning data extraction unit 32 in this embodiment is the same as that in the first embodiment. However, the data extracted by the learning data extraction unit 32 is estimation data used by the machine learning device 100 to estimate the state of the injection molding machine 2, which is different from the first embodiment.

[0061] In this embodiment, the preprocessing unit 34, during the state estimation stage of the injection molding machine 2 using the learning model of the machine learning device 100, converts (numerizes, samples, etc.) the estimation data extracted by the learning data extraction unit 32 into a unified form that is processed in the machine learning device 100, and generates state data S in a predetermined form used for estimation by the machine learning device 100.

[0062] The estimation unit 120 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 this embodiment, when the learning model stored in the learning model storage unit 130 is a learning model generated through unsupervised learning (with parameters determined), the estimation unit 120 inputs the state data S obtained by the preprocessing unit 34 into the learning model, thereby estimating how much the state data S deviates from the state data obtained during normal operation, and calculating the state-related anomaly degree of the injection molding machine as the estimation result. Conversely, when the learning model stored in the learning model storage unit 130 is a learning model generated through supervised learning, the state data S obtained by the preprocessing unit 34 is input into the learning model, thereby estimating and calculating the category (normal / abnormal, etc.) to which the operation state of the injection molding machine belongs. The estimation results (state-related anomaly degree of the injection molding machine and the category to which the operation state of the injection molding machine belongs, etc.) estimated by the estimation unit 120 can be displayed and output to the display device 70, or transmitted and output to a host computer or cloud computer via a wired / wireless network (not shown) for use. Furthermore, when the state determination device 1 determines a predetermined state by the estimation unit 120 (for example, when the degree of abnormality estimated by the estimation unit 120 exceeds a predetermined threshold, or when the category of the injection molding machine's operating state estimated by the estimation unit 120 becomes "abnormal"), for example... Figure 8 For example, warning messages or icons can be displayed on the display device 70, and commands can be output to the injection molding machine to stop, slow down, or limit the torque of the prime mover driving the injection molding machine.

[0063] In the state determination device 1 of the second embodiment with the above-described structure, regarding the acquired data obtained from the injection molding machine 2, the learning data extraction unit 32 extracts estimation data from the acquired data contained in the acquired data storage unit 50 according to the extraction conditions stored in the extraction condition storage unit 52. The extraction condition storage unit 52 can set extraction conditions so that the operator can extract appropriate data as estimation data in accordance with the current state determination purpose of the injection molding machine 2. In this way, the estimation data extracted by the learning data extraction unit 32, such as time series data after excluding alarms, after the start of machine operation, after changing the metal mold, or after changing the setting values ​​of molding conditions such as injection conditions or holding pressure conditions during machine operation—time series data representing changes in the operating state or unstable molding state of the injection molding machine—can only be used for state determination based on time series data suitable for the operation state determination of the injection molding machine 2, which is expected to improve the accuracy of the machine learning-based determination of the operation state of the injection molding machine 2.

[0064] The state determination device 1 of the first and second embodiments described above can be applied to situations involving state determination of industrial machinery such as robots and machine tools. However, it is also suitable for industrial machinery that exhibits unstable movements within a predetermined range, such as at the start of manufacturing or during the initiation of operating actions at the start of manufacturing restart. In particular, even when an injection molding machine is producing under the same injection conditions, there are many cases where unstable movements occur within a predetermined range at the start of machine operation or after a change in injection conditions. However, if operation continues directly, the movement will converge to a stable and normal state. Therefore, such operating states are not considered abnormal states and are not subject to maintenance / inspection. Thus, the state determination device of the present invention is particularly useful for injection molding machines with such characteristics.

[0065] The embodiments of the present invention have been described above. However, the present invention is not limited to the examples of the above embodiments, and can be implemented in various ways with appropriate modifications.

[0066] For example, in the above embodiment, the state determination device 1 and the machine learning device 100 are described as devices with different CPUs (processors). However, the machine learning device 100 can be implemented using the CPU 11 of the state determination device 1 and the system program stored in the ROM 12. In addition, when multiple injection molding machines 2 are interconnected via a network, the operating state of the multiple injection molding machines can be determined by a single state determination device 1, or the state determination device 1 can be installed on the control device of the injection molding machine.

Claims

1. A state determination device for determining the operating state of an injection molding machine, characterized in that, The status determination device has the following features: The data acquisition unit acquires time-series data within the molding process of one cycle of the injection molding machine as data. The extraction condition storage unit stores extraction conditions that associate changes in the operating state or operational state of the injection molding machine with a predetermined number of cycles. These extraction conditions are used to extract data used for machine learning-related processing from the data obtained by the data acquisition unit. The learning data extraction unit extracts data used in machine learning-related processing from the data acquired by the data acquisition unit, according to the extraction conditions stored in the extraction condition storage unit; and The machine learning device performs machine learning-related processing on the data extracted by the aforementioned learning data extraction unit. The extraction conditions stored in the extraction condition storage unit are conditions that specify the data used in machine learning, or conditions that specify data not used in machine learning. The extraction conditions stored in the extraction condition storage unit include conditions for classifying the acquired data based on predetermined data values ​​contained in the acquired data, and a specification of whether to use the acquired data classified by the conditions as learning data. The extraction condition storage unit also stores the priority order among the extraction conditions. When there are multiple extraction conditions, the priority order is used to determine whether the data should be used as learning data or not. If the extraction condition for removing data acquired during an alarm is set in the aforementioned extraction condition storage unit, then data acquired during the period in which an alarm was generated will not be extracted as learning data.

2. The state determination device according to claim 1, characterized in that, The aforementioned machine learning device includes a learning unit that performs machine learning using the data extracted by the aforementioned learning data extraction unit and generates a learning model.

3. The state determination device according to claim 2, characterized in that, The aforementioned learning department performs at least one of the following machine learning methods: supervised learning, unsupervised learning, and reinforcement learning.

4. The state determination device according to claim 1, characterized in that, The aforementioned machine learning device has the following features: The learning model storage unit stores learning models generated through machine learning using data extracted from the aforementioned learning data extraction unit; and The estimation unit estimates the state of the injection molding machine using the learning model based on the data extracted by the learning data extraction unit.

5. The state determination device according to claim 4, characterized in that, The aforementioned estimation unit estimates the degree of abnormality related to the operating state of the injection molding machine. When the degree of abnormality estimated by the aforementioned estimation unit exceeds a predetermined threshold, the aforementioned status determination device displays a warning message on the display device.

6. The state determination device according to claim 4, characterized in that, The aforementioned estimation unit estimates the degree of abnormality related to the operating state of the injection molding machine. When the degree of abnormality estimated by the above-mentioned estimation unit exceeds a predetermined threshold, the above-mentioned status determination device displays a warning icon on the display device.

7. The state determination device according to claim 4, characterized in that, The aforementioned estimation unit estimates the degree of abnormality related to the operating state of the injection molding machine. When the degree of abnormality estimated by the above-mentioned estimation unit exceeds a predetermined threshold, the above-mentioned state determination device outputs at least one of the following commands to the injection molding machine: stop operation, decelerate operation, or limit the torque of the prime mover.

8. The state determination device according to claim 1, characterized in that, The data acquired by the aforementioned data acquisition unit includes at least one of the following information: information indicating any one of the mechanical states of the injection molding machine, such as running, stopping, heating up, heating up finished, metal mold changing in progress, metal mold changing finished, alarm in progress, and production finished; information indicating whether there has been a change in any one of the operating states of the injection molding machine, such as injection conditions, holding pressure conditions, metering conditions, mold opening and closing conditions, ejection conditions, and temperature conditions; and information indicating any one of the molding processes of the injection molding machine, such as mold closing process, mold closing process, injection process, holding pressure process, metering process, mold opening process, ejection process, and standby process.

9. The state determination device according to claim 1, characterized in that, The data acquired by the aforementioned data acquisition unit includes at least one of the data acquired from multiple injection molding machines connected via wired or wireless networks.

10. A machine learning method in a state determination device for determining the operating state of an injection molding machine, characterized in that, The state determination method includes the following steps: The time-series data obtained within the molding process of one cycle of the above-mentioned injection molding machine is used as a data acquisition step; According to the extraction conditions of extracting data used for machine learning-related processing from the data obtained from the injection molding machine by associating the changes in the operating state or operation state of the injection molding machine with a predetermined number of cycles, a learning data extraction step is performed to extract data used for machine learning-related processing from the data obtained in the data acquisition step. as well as The steps involved performing machine learning-related processing on the data extracted in the aforementioned learning data extraction steps. The extraction criteria mentioned above specify either the data to be used in machine learning or the data not to be used in machine learning. The above extraction conditions include conditions for classifying the acquired data based on at least predetermined data values ​​contained in the acquired data, and a specification of whether to use the acquired data classified according to the conditions as learning data. When multiple extraction criteria exist, the order of priority among these criteria determines whether the data should be used for learning purposes or not. If the extraction criteria for removing data acquired during alarms are set from the learning data, then data acquired during the period in which an alarm was generated will not be extracted as learning data.

Images