Analysis device

Abstract

An analysis device according to the present disclosure is provided with: a data acquisition unit that acquires data detected by an industrial machine; an operation state extraction unit that extracts, from the data, data detected during operation of the industrial machine; an annotation unit that creates a plurality of data set groups by giving, on the basis of a predetermined standard, annotations indicating motion states of the industrial machine to a plurality of data sets segmented by a predetermined standard from the extracted data during the operation; a feature amount extraction unit that extracts a feature amount of data included in each of the data sets; a learning unit that generates decision tree models respectively for the plurality of data set groups; and a display unit that ranks the generated decision tree models on the basis of correct answer rates regarding prediction of the annotations based on predetermined training data and that displays the decision tree models.

Description

[Technical field]

[0001] The present invention relates to an analytical device. [Background technology]

[0002] 2. Description of the Related Art In manufacturing sites such as factories, devices are introduced that monitor the operating status of industrial machines, such as robots and machine tools, installed in manufacturing lines, and manage the operating status of the industrial machines.

[0003] As an example of an apparatus for managing the operating status of industrial machinery, Patent Document 1 discloses an apparatus that, when a product is produced, acquires data on normal occasions, when the product is produced normally, and data on abnormal occasions, when an abnormality occurs in the produced product, and, based on the acquired data, identifies the degree of correlation between the data and an abnormality that occurs in the produced product, and selects data that is effective for predicting the abnormality. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2018-116545 A Summary of the Invention [Problem to be solved by the invention]

[0005] In manufacturing sites, there is a demand to know what symptoms appear in the data collected before an abnormality occurs. In order to find symptoms before an abnormality occurs, when data is collected, it is necessary to compare the data collected under normal conditions with the data collected before the abnormality occurs (before a processing defect or an alarm occurs) and find the symptoms appearing in the data before the abnormality occurs.

[0006] However, such signs that appear in the data before an abnormality may occur only immediately before the abnormality occurs, or may occur several days before. Therefore, when comparing normal data with data before the abnormality occurs, it is necessary to compare which parts of the huge amount of data, which places an excessive burden on the user. In addition, it is difficult to find signs in the data before the abnormality occurs by comparing such a huge amount of data. This type of work is necessary not only for analyzing normal / abnormal data, but also for analyzing signs that appear in data when various events occur at manufacturing sites. Therefore, there is a need for technology to assist in the analysis of data that can be indicative of events occurring at manufacturing sites. [Means for solving the problem]

[0007] The operation data analysis device according to the present disclosure extracts features from operation data acquired from industrial machinery, and supports a user's data analysis work by displaying differences in the extracted features in a decision tree, thereby solving the above-mentioned problems.

[0008] One aspect of the present disclosure is an analysis device that supports analysis of data collected from an industrial machine, the analysis device including a data acquisition unit that acquires the data detected by the industrial machine, and a data acquisition unit that acquires the data detected by the industrial machine while the industrial machine is in operation based on the data. When an operating status extraction unit for extracting data detected during operation based on a predetermined criterion from the extracted operating data; Then, multiple data sets are cut out and the corresponding The analysis device includes an annotation unit that creates a plurality of dataset groups in which annotations indicating the operating state of the industrial machine are added to the datasets according to predetermined criteria; a feature extraction unit that extracts features of the data contained in each of the datasets; a learning unit that generates a decision tree model for each of the plurality of dataset groups using the features as independent variables and the annotations indicating the operating state of the industrial machine as dependent variables; and a display unit that ranks and displays the plurality of decision tree models based on the accuracy rate of predictions of the annotations based on predetermined training data. Effect of the Invention

[0009] According to one aspect of the present disclosure, an industrial machine detects that it is collecting data. Machine It is possible to visualize the conditions that determine differences in states (such as normal operation or before an abnormality occurs) as a decision tree model, making it easier to grasp signs of state changes that appear in the data. [Brief description of the drawings]

[0010] [Figure 1] 1 is a schematic hardware configuration diagram of an analysis device according to an embodiment of the present invention. [Diagram 2] 1 is a block diagram showing schematic functions of an analysis device according to an embodiment of the present invention; [Diagram 3] 1 is a diagram for explaining how to extract a data set and how to add an annotation indicating an operating state of an industrial machine. FIG. [Figure 4] FIG. 4 is a diagram showing a display example by a display unit. [Diagram 5] FIG. 11 is a diagram showing another example of display by the display unit. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. 1 is a schematic hardware configuration diagram showing the main parts of an analytical device according to one embodiment of the present invention. The analytical device 1 of the present invention can be implemented on, for example, a control device that controls an industrial machine 3. The analytical device 1 of the present invention can also be implemented on a computer such as a personal computer provided with the industrial machine 3, a personal computer connected to the industrial machine 3 via a wired / wireless network, a cell computer, a fog computer 6, or a cloud server 7. In this embodiment, an example is shown in which the analytical device 1 is implemented on a personal computer connected to the industrial machine 3 via a network.

[0012] The CPU 11 provided in the analysis device 1 according to this embodiment is a processor that controls the entire analysis device 1. The CPU 11 reads out a system program stored in the ROM 12 via the bus 22, and controls the entire analysis device 1 in accordance with the system program. The RAM 13 temporarily stores temporary calculation data, display data, various data input from outside, and the like.

[0013] The non-volatile memory 14 is composed of, for example, a memory backed up by a battery (not shown) or an SSD (Solid State Drive), and the memory state is maintained even when the power supply of the analysis device 1 is turned off. The non-volatile memory 14 stores data read from an external device 72 via an interface 15, data input via an input device 71, data acquired from the industrial machine 3, and the like. The data stored in the non-volatile memory 14 may be expanded in the RAM 13 at the time of execution / use. In addition, various system programs such as a known analysis program are written in the ROM 12 in advance.

[0014] The interface 15 is an interface for connecting the CPU 11 of the analysis device 1 to an external device 72 such as a USB device. For example, image data of a work stored in advance can be read from the external device 72. In addition, setting data edited in the analysis device 1 can be stored in an external storage means via the external device 72.

[0015] The interface 20 is an interface for connecting the CPU 11 of the analytical device 1 to a wired or wireless network 5. The network 5 is connected to the industrial machine 3, the fog computer 6, the cloud server 7, etc., and exchanges data with the analytical device 1.

[0016] The industrial machines 3 include turning machines, electric discharge machines, robots, transport machines, etc., which are installed at manufacturing sites, etc. The CPU 11 of the analysis device 1 acquires various data via the network 5 and external devices 72, such as the motor current / voltage detected when the industrial machine 3 performs workpiece manufacturing work, the position, speed, and acceleration of the operating parts, video and audio showing the processing status, the temperature around the machine and each part of the machine, the signal status of each part, the setting status of each part of the industrial machine 3, the quality of the operation of the industrial machine 3 input by the worker, the processing result of the product, and setup information.

[0017] The display device 70 displays various data loaded into the memory, data obtained as a result of executing programs, etc., outputted via the interface 17. An input device 71 consisting of a keyboard, pointing device, etc., transmits commands and data based on operations by an operator to the CPU 11 via the interface 18.

[0018] 2 is a schematic block diagram showing functions of the analysis device 1 according to one embodiment of the present invention. Each function of the analysis device 1 according to this embodiment is realized by the CPU 11 of the analysis device 1 shown in FIG. 1 executing a system program and controlling the operation of each part of the analysis device 1.

[0019] The analysis device 1 of this embodiment includes a data acquisition unit 110, an operation state extraction unit 120, a change point detection unit 125, an annotation unit 130, a feature extraction unit 135, a learning unit 145, and a display unit 155. In addition, on the RAM 13 to the non-volatile memory 14 of the analysis device 1, an operation data storage unit 115 which is an area for storing data detected by the industrial machine 3, a feature extraction pattern storage unit 140 which is an area in which a feature extraction pattern used for extracting features is stored in advance, a model storage unit 150 which is an area for storing a model generated as a result of learning, and a training data storage unit 160 which is an area for storing training data used for training the created model are prepared in advance.

[0020] The data acquisition unit 110 acquires data detected by the industrial machine 3. The data detected by the industrial machine 3 includes, for example, the current / voltage of the motor detected when the industrial machine 3 is in operation, the position and speed of the operating part, acceleration, video and audio showing the processing status, the temperature around the machine and each part of the machine, the state of the signal of each part, the setting state of each part of the industrial machine 3, the quality of the operation of the industrial machine 3 input by the worker, the processing result of the product, and setup information. The data acquisition unit 110 may acquire data from the industrial machine 3 via the network 5, for example. In addition, data detected by the industrial machine 3 that has been recorded in advance may be acquired from the external device 72. The data acquisition unit 110 stores the acquired data in the operation data storage unit 115.

[0021] The operation status extraction unit 120 extracts signal data indicating whether the industrial machine 3 is operating or not from the data detected by the industrial machine 3 stored in the operation data storage unit 115, and identifies the portion of the data detected while the industrial machine 3 is operating. Generally, while the industrial machine 3 is operating, a predetermined signal indicating that the industrial machine 3 is operating is in an ON state. The operation status extraction unit 120 extracts data within the range in which the signal is ON as a set of data detected while the industrial machine 3 is operating.

[0022] The change point detection unit 125 detects a change point at which a change appears in the trend of data movement from among the data detected by the industrial machine 3 stored in the operation data storage unit 115. Examples of the change point include change points that can be extracted from data, such as a change in the machining program, a change in the tool, or a change in the offset value, change points extracted from signal data, such as a sudden change in temperature or a decrease in the torque of the motor of the industrial machine 3, and change points due to human work, such as a change in the setup time or a change in the execution time of the machining program. In many cases, similar trends are not observed in the detected data movement before and after the change point. Therefore, the change point detection unit 125 detects the change point as one criterion when analyzing data. In many cases, a group of data detected between change points is analyzed as a set.

[0023] The annotation unit 130 performs annotation by adding the operation state of the industrial machine 3 at the time when the data extracted by the operation state extraction unit 120 was detected as an annotation to the data detected during the operation of the industrial machine 3. Examples of the annotation of the operation state added by the annotation unit 130 may be, for example, the annotation of "normal operation" and the annotation of "before the occurrence of an abnormality". Other examples of the annotation of the operation state added by the annotation unit 130 may be based on the content input by the operator with respect to the result of the operation of the industrial machine 3, such as "before the occurrence of a machining defect". In addition, for the abnormality of the industrial machine 3, a more detailed annotation may be added based on a detected alarm signal, such as before the occurrence of a tool-related alarm or before the occurrence of a spindle-related alarm, and for the machining defect, a more detailed annotation according to the defect content, such as before the occurrence of chipping of the workpiece or before the occurrence of large roughness of the cutting surface, may be added. These annotations may be automatically added according to the content of the operation data, or may be added by a user by manually specifying the annotation.

[0024] When a predetermined event is detected, it is not possible to know from the beginning when the symptom of the event appears in the detected data. Therefore, the annotation unit 130 extracts a data set for each operation, for example, and creates a data set group in which the data set for one operation is annotated with "before the occurrence of the event" going back from the time when the predetermined event was detected. In addition, a data set group is created in which the data set for two operations is annotated with "before the occurrence of the event" going back from the time when the predetermined event was detected. Furthermore, a data set group is created in which the data set for three operations is annotated with "before the occurrence of the event" going back from the time when the predetermined event was detected. In this way, the annotation unit 130 creates a predetermined number of n data set groups. It is preferable that the number of data set groups to be created can be set in advance based on a user's designation. The range of the data set extraction is not limited to the number of operations. For example, the data set may be extracted based on a time, such as 30 minutes before, 1 hour before, etc. going back from the time when the predetermined event was detected. In addition, the change point detected by the change point detection unit 125 may be used as the range of the data set extraction. The range of the data set to be cut out may be set based on a user specification.

[0025] An example of the operation of the operating state extraction unit 120 and the annotation unit 130 will be described with reference to Fig. 3. Fig. 3 is a diagram showing an example of data detected from the industrial machine 3 when a machining defect of a workpiece is detected. In this example, OperationA data set is cut out for each number of times. When the data illustrated in FIG. 3 is detected, the operation state extraction unit 120 refers to the operation signal in the detected data, and extracts a data group of the section in which the operation signal is ON as a data set detected during operation. In the example of FIG. 3, at least three data sets 311, 312, and 313 are extracted. It is assumed that an operator reports that a machining defect occurred in the workpiece machined when the data set 311 of these data sets was detected. In this case, the annotation unit 130 creates a data set group in which the data set 311 extracted by the operation state extraction unit 120 is annotated as "before the occurrence of the machining defect" and the data sets 312 and 313 are annotated as "normal operation". The annotation unit 130 also creates a data set group in which the data set 311 and 312 extracted by the operation state extraction unit 120 are annotated as "before the occurrence of the machining defect" and the data set 313 is annotated as "normal operation". In this way, the annotation unit 130 creates a plurality of data set groups.

[0026] The feature extraction unit 135 extracts a predetermined feature for each data included in the annotated data set. Examples of the predetermined feature include an average value, a change slope, a variance value, a maximum change amount, a value after Fourier transform, and an abnormality degree. The feature extraction unit 135 may extract a plurality of feature values ​​from one data. The feature extraction pattern storage unit 140 stores a feature value to be extracted in advance for each data included in the data set. The feature extraction unit 135 refers to the feature extraction pattern storage unit 140 to determine a feature value to be extracted for each data.

[0027] The learning unit 145 generates a decision tree model for each data set group created by the annotation unit 130, in which the feature quantities of each data extracted by the feature quantity extraction unit 135 are used as independent variables and the annotations of the operating state of the industrial machine 3 added by the annotation unit 130 are used as dependent variables. The decision tree learning algorithm used by the learning unit 145 may be known ID3, CART, C4.5, or the like. The hyperparameters of the decision tree (such as the depth of the tree and the number of nodes) may be set in advance. Note that the algorithm for creating the decision tree model is already well known, so a detailed description of the generation process will be omitted in this specification. Each decision tree model generated by the learning unit 145 is stored in the model storage unit 150.

[0028] The display unit 155 evaluates the decision tree model generated by the learning unit 145 using the training data, and displays the evaluation result on the display device 70. The training data storage unit 160 may store a plurality of training data created in advance by, for example, a user. In this case, the training data is data to which the user has given correct annotations to a specific data set. In addition, the training data may be a data set to which the annotation unit 130 has given annotations. The display unit 155 predicts annotations of the operating state of the industrial machine 3 for each piece of training data stored in the training data storage unit 160, using each decision tree model generated by the learning unit 145. Next, if the predicted annotation matches the annotation given to the training data, it is determined to be correct, and if it is different, it is determined to be incorrect, and calculates the accuracy rate of the decision tree model for the entire training data. Then, the result is displayed on the display device 70 as an evaluation result.

[0029] The display unit 155 may display on the display device 70 the accuracy rate and parameters such as how the data set is extracted for each decision tree model generated by the learning unit 145. FIG. 4 shows an example of displaying the accuracy rate for each model by the display unit 155. As exemplified in FIG. 4, the display unit 155 displays a display that allows the user to understand how the data set is extracted for each model and how annotations are added, and a display that allows the user to understand the correspondence between the accuracy rate when prediction is performed on training data. By displaying in this manner, the user can quickly understand how the data is extracted and how annotations are added to generate a decision tree model closer to the correct answer.

[0030] The display unit 155 may display the decision tree model and the details of the accuracy rate as the evaluation result based on the user's designation or automatically for the decision tree model with a high accuracy rate. FIG. 5 is an example of display of the evaluation result of the decision tree model by the display unit 155. The example shown in FIG. 5 is a decision tree model created for a set of data in which each data set is annotated as a normal operation and a state before the occurrence of an abnormality. The decision tree learning algorithm treats the maximum value of the spindle torque and the average value of the servo motor temperature as independent variables that affect the prediction result, and predicts that each data set is either in a normal operation state or a state before the occurrence of an abnormality based on the judgment of each independent variable. The display unit 155 predicts the state using this decision tree model for the training data prepared in advance. The prediction result is shown in a pie chart in FIG. 5. In the decision tree model shown in FIG. 5, it can be seen at a glance that the accuracy rate is poor when the maximum value of the spindle torque is less than 100 N·m and the servo motor temperature is 29°C or higher. In addition, when the maximum spindle torque is 100 N·m or more and the average servo motor temperature is less than 26°C, the accuracy rate is also low.

[0031] In addition, when each node of the decision tree model is selected in the display of Fig. 5, the data set classified into that node may also be displayed. By displaying these in parallel, the user can easily interpret the classification rules in each decision tree model.

[0032] In this way, while viewing the decision tree model displayed by the display unit 155 and the results of predictions for the training data, the user can change the way the data set is extracted and add / delete features to be extracted and stored in the feature extraction pattern storage unit 140. By repeating this process, a decision tree model that makes a more appropriate prediction for the training data can be obtained. By checking how annotations have been added to a decision tree model that makes an appropriate prediction, the user can easily determine at what timing an indication of a state change appears.

[0033] The analysis device 1 according to the present embodiment having the above configuration can visualize, as a decision tree model, data conditions that are signs of a change in the state (normal operation, before an abnormality occurs, etc.) of the industrial machine 3. By referring to this decision tree model, the user can easily grasp the signs of state changes appearing in the data. By visualizing the decision tree models generated for each method of extracting the data set, it becomes possible to grasp when the signs of each state change appearing in the data.

[0034] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described examples of the embodiments, and can be embodied in various forms by making appropriate modifications. [Explanation of symbols]

[0035] 1 Analyzer 3. Industrial machinery 5. Network 6. Fog Computer 7. Cloud Server 11 CPU 12 ROM 13 RAM 14 Non-volatile memory 15,17,18,20 Interface 22 Bus 70 Display device 71 Input Devices 72 External equipment 110 Data Acquisition Section 115 Operational data storage unit 120 Operation status extraction unit 125 Change point detection unit 130 Annotation Section 135 Feature Extraction Unit 140 Feature extraction pattern storage unit 145 Learning Department 150 Model Memory Section 155 Display section 160 Training data storage unit 311~313 Dataset

Claims

1. An analysis device that supports analysis of data collected from an industrial machine, a data acquisition unit that acquires the data detected by the industrial machine; an operating state extraction unit that extracts data detected when the industrial machine is in operation from the data; an annotation unit that extracts a plurality of data sets from the extracted data during operation based on a predetermined criterion, and creates a plurality of data set groups by adding annotations indicating the operating state of the industrial machine based on the predetermined criterion to the plurality of extracted data sets; a feature extraction unit that extracts features of data included in each of the data sets; a learning unit that generates a decision tree model for each of the plurality of dataset groups using the feature amounts as independent variables and the annotations indicating the operational states of the industrial machines as dependent variables; a display unit that ranks and displays the plurality of decision tree models based on a correct answer rate related to the prediction of the annotation based on predetermined training data; An analytical device comprising:

2. The predetermined criterion for extracting the data set from the extracted operation data is the number of times or time of operation of the industrial machine. The analytical device of claim 1 .

3. a change point detection unit that detects a change point in the operation of the industrial machine from the data acquired by the data acquisition unit, The predetermined criterion for extracting the data set from the extracted in-operation data is a change point detected by the change point detection unit. The analytical device of claim 1 .

4. Each of the plurality of data sets is annotated as a change in an operating state of the industrial machine at a different time. The analytical device of claim 1 .

5. the display unit displays the decision tree model and a correct answer rate based on the training data for each terminal node of the decision tree model. The analytical device according to claim 1.

6. the display unit displays the training data classified into the terminal node when the terminal node of the decision tree model is selected. The analytical device according to claim 1.

Images