WO2024075567A1

WO2024075567A1 - Diagnostic system, information processing device, diagnostic method, and program

Info

Publication number: WO2024075567A1
Application number: PCT/JP2023/034701
Authority: WO
Inventors: 幸広石黒
Original assignee: 三菱電機株式会社
Priority date: 2022-10-03
Filing date: 2023-09-25
Publication date: 2024-04-11

Abstract

This diagnostic system (1) comprises: an operation state information acquisition unit that acquires operation state information that indicates an operation state of a machining facility (2) to be diagnosed; an extraction unit that extracts, from the operation state information that has been acquired by the operation state information acquisition unit, operation state information of a calculation target interval that indicates a time region that is determined according to a preset rule, in a stable machining interval in which a motor rotates in a stable manner; a feature amount calculation unit that calculates a feature amount from the operation state information of the calculation target interval that has been extracted by the extraction unit; and a diagnostic unit that determines, on the basis of the feature amount that has been calculated by the feature amount calculation unit, whether the machining facility 2 is normal or abnormal.

Description

Diagnostic system, information processing device, diagnostic method, and program

This disclosure relates to a diagnostic system, an information processing device, a diagnostic method, and a program.

Technology is known that collects data from machining equipment at manufacturing sites and uses the collected data to diagnose the condition of the machining equipment. For example, Patent Document 1 discloses a numerical control device that determines whether or not an abnormality has occurred in machining equipment based on data on the current flowing through a motor that drives a tool.

JP 2020-013433 A

This numerical control device determines whether an abnormality has occurred by using data from all sections in which the motor rotates at a constant speed. However, if the abnormality occurs for a short period of time, the relatively long period of normal data and the relatively short period of abnormal data are averaged together, and there is a possibility that the abnormality will not be detected. For this reason, there is room for improvement in terms of accurately determining whether an abnormality has occurred in the processing equipment.

This disclosure has been made in consideration of the above-mentioned circumstances, and aims to provide a diagnostic system, information processing device, diagnostic method, and program capable of accurately diagnosing the condition of processing equipment.

In order to achieve the above object, the diagnostic system disclosed herein includes an operating state information acquisition unit that acquires operating state information indicating the operating state of the processing equipment to be diagnosed, an extraction unit that extracts, from the operating state information acquired by the operating state information acquisition unit, operating state information of a calculation target section that indicates a time domain determined by a preset rule within a stable processing section in which the motor rotates stably, a feature amount calculation unit that calculates a feature amount from the operating state information of the calculation target section extracted by the extraction unit, and a diagnosis unit that determines whether the processing equipment is normal or abnormal based on the feature amount calculated by the feature amount calculation unit.

According to the present disclosure, it is possible to determine whether the processing equipment is normal or abnormal based on the operating status information of the calculation target section determined by preset rules within the stable processing section in which the motor of the processing equipment to be diagnosed rotates stably. This makes it possible to diagnose the condition of the processing equipment with high accuracy.

FIG. 1 is a block diagram showing a functional configuration of a diagnostic system according to a first embodiment. FIG. 2 is a diagram showing an example of the rotation speed of a spindle motor, an example of a cutting feed signal, and an example of a current value flowing through a spindle motor, which are acquired by the machining data acquisition device shown in FIG. FIG. 2 is a diagram showing an example of a correspondence table stored in the signal processing device shown in FIG. 1 . FIG. 2 is a block diagram showing an example of a physical configuration of the processed data acquisition device, the signal processing device, the diagnostic device, the learning model generation device, and the control device shown in FIG. 1 . 1 is a flowchart showing a learning model generation process performed by a diagnostic system according to an embodiment. FIG. 2 is a diagram showing an example of learning data generated by the signal processing device shown in FIG. 1 . 1 is a flowchart showing a diagnostic process performed by a diagnostic system according to an embodiment.

Below, the diagnostic system, information processing device, diagnostic method, and program according to the embodiments of the present disclosure will be described with reference to the drawings. Note that the same reference numerals are used to denote the same or corresponding parts in the drawings.

The diagnostic system according to this embodiment collects time-series status data from the processing equipment that indicates the operating status during processing. From the collected status data, the diagnostic system extracts status data for time domains in which abnormalities in the processing equipment are likely to occur according to preset rules, and calculates feature values from the extracted status data. The diagnostic system applies the calculated feature values to a machine learning model to determine whether or not there is an abnormality in the processing equipment.

As shown in FIG. 1, the diagnostic system 1 according to this embodiment includes a processing data acquisition device 3 that acquires processing data including status data from the processing equipment 2 to be diagnosed, a signal processing device 4 that processes the status data and generates input data to be input to a learning model, a storage device 5 that stores various data generated by the signal processing device 4, a diagnostic device 6 that determines whether or not there is an abnormality in the processing equipment 2 based on the input data generated by the signal processing device 4, a learning model generation device 7 that generates a learning model used by the diagnostic device 6, a status indicator 8 that displays the diagnosis result by the diagnostic device 6, and a control device 9 that controls the equipment based on the determination result by the diagnostic device 6.

The processing equipment 2 is a machine for performing processes such as cutting, cutting, and polishing on workpieces, and is, for example, a machining center, lathe, drill press, milling machine, polishing machine, turning center, electric discharge machine, laser processing machine, ultrasonic processing machine, etc. The processing equipment 2 is equipped with sensors such as a current sensor, vibration sensor, microphone, AE (Acoustic Emission) meter, acceleration sensor, torque sensor, rotation sensor, and temperature sensor. The sensors detect physical quantities such as the current value flowing through motors that drive tools such as drills, end mills, face mills, long drills, bit tips, and grinding wheels, the motor rotation speed, rotation torque, and the magnitude of vibration, and output information on the detected physical quantities.

The processing data acquisition device 3 acquires processing data including time-series status data indicating the operating status during processing from the processing equipment 2, temporarily stores it, and provides a certain amount of processing data as a bundle to the signal processing device 4. The processing data acquired by the processing data acquisition device 3 includes status data indicating the operating status of the processing equipment 2, such as information on physical quantities detected by sensors and instruction signals output from the control device 9 that controls the operation of the processing equipment 2, and processing-related information, which is information for identifying the processing process, such as the processing program number, tool number, power consumption, coordinate information of the processing axis, hydraulic and pneumatic pressure, battery voltage, and number of days the equipment has been in use. The processing data acquisition device 3 is an example of an operating status information acquisition unit and a processing identification information acquisition unit, the status data is an example of operating status information, and the processing-related information is an example of processing identification information.

The signal processing device 4 processes the time-series status data indicating the operating status of the processing equipment 2 transmitted from the processing data acquisition device 3, and generates input data to be input to the learning model. Specifically, the signal processing device 4 performs signal processing such as amplifying the analog signal acquired from the sensor by the processing data acquisition device 3, filtering using a high-pass filter or a low-pass filter, and A/D (Analog to Digital) conversion based on a sampling frequency and resolution preset by the user. The signal processing device 4 also performs cleansing processing such as removing noise and abnormal values and filling in missing values.

After performing signal processing, the signal processing device 4 extracts data of the calculation target section, which is the section used to calculate the feature amount, from the state data after processing. Specifically, as illustrated in FIG. 3, the signal processing device 4 stores a correspondence table, which is information for identifying which extraction rule is applied from the extraction rules defined by the user in each machining process. The signal processing device 4 refers to this correspondence table, identifies an extraction rule from the machining program number and tool number included in the machining data, and executes the identified extraction rule to extract state data of the calculation target section. As shown in FIG. 3, the correspondence table includes items for the machining program number that identifies the machining program, the tool number that identifies the tool, and the extraction rule ID that identifies the extraction rule. In the illustrated example, it is shown that the extraction rule ID "r1" is applied to the machining process executed by the machining program number "N001" and the tool number "D01".

To explain a specific example, without being limited thereto, FIG. 2 shows the transition of the status data when cutting processing is started, a temporary dimensional defect occurs when the workpiece and the tool come into contact with each other after the actual processing starts, and then normal operation is restored. The status data includes the rotation speed of the spindle motor equipped in the processing equipment 2 shown in FIG. 2, the cutting feed signal output from the control device 9, and the current value flowing through the spindle motor. When cutting processing starts, the processing equipment 2 rotates the spindle motor and accelerates it to a rotation speed according to the rotation speed command by the processing program. In this motor acceleration section, an acceleration torque is generated, and the current flowing through the spindle motor temporarily increases, and when the rotation of the motor reaches a steady state, the current decreases. After the rotation speed of the spindle motor reaches the target rotation speed and the rotation stabilizes, the actual processing starts by the cutting feed signal. When the actual processing starts and the workpiece and the tool come into contact with each other, a load is applied to the spindle motor, so the current flowing through the spindle motor increases again. In this example, it is shown that the current flowing through the spindle motor temporarily increases in the abnormality occurrence section where the dimensional defect occurs. After that, the operation returns to normal, so the current value settles to a steady level. After that, when the actual machining is completed, the machining equipment 2 performs an operation to decelerate the spindle motor. In this motor deceleration section, a deceleration torque is generated, and the current flowing through the spindle motor temporarily increases as in the acceleration section, and the current decreases when the rotation of the motor stops. In this way, the current value flowing through the spindle motor changes in the acceleration section and deceleration section of the spindle motor regardless of the machining phenomenon, so using data from these sections for diagnosis may result in erroneous judgment. Therefore, data from the stable machining section, which is a section in which the motor rotates at a constant speed, excluding the acceleration section and deceleration section of the spindle motor, is used for diagnosis. Also, as in the example shown in the figure, if the feature amount is calculated using all the current data in the stable machining section, when an abnormality occurs temporarily, the difference between the feature amount calculated from the normal data and the feature amount calculated from the data including the abnormality becomes small, and the abnormality may not be detected correctly. Therefore, in each machining process, a calculation target section, which is a time region in which an abnormality is likely to occur, i.e., the probability or frequency of occurrence is high, is specified in advance, and a rule is defined for extracting state data from the calculation target section for each machining process. The rules are arbitrary, but may include, for example, an event that acts as a trigger to identify the interval to be calculated, and information including the time from the trigger to the start and end of the interval to be calculated. If the example in Figure 2 occurs frequently, the trigger that defines the interval to be calculated may be, for example, the cutting feed signal changing from low to high, the start of the interval to be calculated being synchronized with the trigger, and the end being set to the timing when the time t has elapsed from the trigger.

To design such rules, the user specifies the type of machining equipment and tools, the machining content, etc., and specifies the time region in each machining process where abnormalities are likely to occur, and sets the extraction rules. For example, the user may determine the calculation target section by capturing the machining phenomenon with a high-speed camera to specify the timing of abnormality occurrence, observing the behavior of the status data to specify the timing of abnormality occurrence, or setting multiple candidate sections for the calculation target section and selecting the one with the best accuracy of the diagnosis results obtained by applying each section data to the learning model as the calculation target section. Then, for example, if it is specified that abnormalities are likely to occur immediately after the start of actual machining, an extraction rule is defined to extract status data for 5 seconds from the time the tool and workpiece come into contact, and in the case of machining where abnormalities are likely to occur at the end of actual machining, an extraction rule is defined to extract status data for 5 seconds before and after the tool and workpiece separate. Note that two or more calculation target sections may be set for one process.

In the example of Figure 3, a combination of a machining program number and a tool number is used as information for identifying each machining process, but any information that can identify the machining process can be used.

The signal processing device 4 calculates the feature amount from the extracted state data of the calculation target section. The feature amount is, for example, the maximum value, minimum value, average value, standard deviation, variance, kurtosis, skewness, etc. of the current value. In addition to the current value, the feature amount may be calculated from the frequency, magnitude, temperature, etc. of vibration, sound, or AE. When obtaining the feature amount from vibration, sound, or AE, frequency analysis may be performed to calculate the feature amount from the magnitude of a specific frequency. The signal processing device 4 calculates the feature amount for each cycle, which is the period from the start to the end of processing of one workpiece. The signal processing device 4 generates input data by associating a combination of the calculated multiple feature amounts with the cycle number indicating which cycle it is. The signal processing device 4 associates the generated input data with processing-related information including the processing program number, tool number, etc. output from the processing data acquisition device 3 and stores them in the storage device 5. In addition, when multiple processing steps are performed in one cycle, the feature amount may be calculated for each processing step, and input data may be generated for each processing step. The signal processing device 4 is an example of an extraction unit and a feature calculation unit.

Returning to FIG. 1, the storage device 5 stores input data including the feature values and cycle counts sent from the signal processing device 4 in association with machining-related information including the machining program number, tool number, etc.

The diagnostic device 6 diagnoses the processing equipment 2 using the input data generated by the signal processing device 4 and stored in the storage device 5 and the learning model generated by the learning model generation device 7. Specifically, the diagnostic device 6 applies the input data generated based on the status data acquired sequentially during processing of the processing equipment 2 to the learning model, and outputs a diagnostic result that determines whether the processing equipment 2 is normal or not. The diagnostic device 6 transmits the diagnostic result to the status display 8 and the control device 9. The diagnostic device 6 may be an edge computer installed at the production site, or may be a computer connected via the Internet from an office away from the production site. The diagnostic device 6 is an example of a diagnostic unit.

The learning model generating device 7 generates a learning model for determining whether or not there is an abnormality in the processing equipment 2. Specifically, the learning model generating device 7 generates a learning model by machine learning by providing the learning model with learning data including input data generated by the signal processing device 4 and label information indicating whether each input data is normal data or abnormal data. The user determines whether the processing process of each cycle performed by actual processing or simulation is normal or not, and creates learning data by assigning label information indicating normal or abnormal to the feature amount corresponding to each cycle. The learning model is a model according to a machine learning algorithm such as a decision tree, k-nearest neighbor method, or support vector machine, and outputs judgment result information that judges whether the processing equipment 2 to be inspected is normal or abnormal using the feature amount calculated from the status data as input. Note that the input data included in the learning data does not have to be status data acquired from the processing equipment 2, and for example, simulated data generated by a computer may be used depending on the type of processing equipment 2 and the processing content. Furthermore, the label information included in the learning data is not limited to two patterns, normal and abnormal, and three or more patterns of labels may be assigned by further subdividing the abnormal label into dimensional abnormality, poor surface quality, etc. The learning model generation device 7 is an example of a learning model generation unit.

The status display 8 displays the diagnosis results from the diagnostic device 6. The status display 8 is, for example, an operation panel installed on the processing equipment 2. The status display 8 notifies the user of the occurrence of an abnormality by displaying information indicating that an abnormality has been diagnosed on a screen. The status display 8 may also include a lamp, a speaker, etc., and may light up a lamp and emit an alarm from the speaker when an abnormality is diagnosed. Furthermore, the status display 8 may have a function of automatically sending an email indicating that an abnormality has been diagnosed to a pre-set email address.

The control device 9 controls the operation of the processing equipment 2. The control device 9 is, for example, a numerical control device that controls the processing equipment 2 by commands based on an NC (Numerical Control) program. When the control device 9 receives diagnosis result information from the diagnosis device 6 indicating that an abnormality has occurred in the processing equipment 2, the control device 9 stops the processing equipment 2. Note that the control device 9 may not completely stop the processing equipment 2, but may limit some of the functions of the processing equipment 2 depending on the degree of the abnormality occurring in the processing equipment 2. Note that the control device 9 is an example of a control unit.

In the diagnostic system 1 having the functional configuration described above, the processed data acquisition device 3, the signal processing device 4, the diagnostic device 6, the learning model generation device 7, and the control device 9 physically comprise, as shown in FIG. 4, a CPU (Central Processing Unit) 11 that executes processing according to a program, a RAM (Random Access Memory) 12 which is a volatile memory, a ROM (Read Only Memory) 13 which is a non-volatile memory, a storage unit 14 that stores data, an input unit 15 that accepts input of information, and a communication unit 16 that transmits and receives information, which are connected via an internal bus 99.

The CPU 11 executes various processes by reading out the programs stored in the storage unit 14 into the RAM 12 and executing them. The main functions provided by the programs are: processing the status data sent from the processing data acquisition device 3 by the signal processing device 4 to extract data for the calculation target section and calculate feature quantities from the extracted data; diagnosis processing by the diagnosis device 6 to determine whether the processing equipment 2 is normal; and generation of a learning model by the learning model generation device 7.

RAM 12 is used as a work area for CPU 11. ROM 13 stores the control program executed by CPU 11, BIOS (Basic Input Output System), etc.

The storage unit 14 includes a hard disk drive, stores the programs executed by the CPU 11, and stores various data used when executing the programs. The storage unit 14 stores the processing data acquired from the processing equipment 2 by the processing data acquisition device 3, the input data generated by the signal processing device 4, the learning model generated by the learning model generation device 7, and the learning data used to learn the learning model.

The input unit 15 is a user interface equipped with a keyboard, mouse, communication device, etc. The communication unit 16 is a network termination device or a wireless communication device that connects to the network, and a serial interface or a LAN (Local Area Network) interface that connects to them.

Next, the operation of the diagnostic system 1 will be described. In the following, the processing equipment 2 to be diagnosed by the diagnostic system 1 is a cutting machine, and an example will be described in which the presence or absence of an abnormality in the processing equipment 2 is determined based on feature values calculated from the current flowing through a spindle motor with a drill attached at the tip.

(Advance preparation)
The diagnostic system 1 calculates the feature quantity from the current data in the stable machining section using the current data of the calculation target section, which is a time region previously specified by the user. Here, the process of specifying the calculation target section will be described. In the actual machining process, the user observes the tendency of abnormality occurrence in the machining equipment 2 by, for example, using a video taken by a high-speed camera, and specifies the time region where abnormality is likely to occur. For example, the user specifies that abnormality is likely to occur in 5 seconds from the time when the tool and the workpiece come into contact. Next, the user sets an extraction rule for extracting current data in the specified time region in the signal processing device 4. For example, in order to extract current data in the specified time region, the user sets an extraction rule for extracting current data for 5 seconds from the time when the cutting signal is output in the signal processing device 4. The user specifies the calculation target section, which is a time region where abnormality is likely to occur, for each machining process according to the type of machining equipment and tool owned and the machining content, and sets an extraction rule for extracting current data in the specified calculation target section.

Next, the user creates a correspondence table, which is information for identifying which extraction rule is applied in each machining process, and stores it in the signal processing device 4. The signal processing device 4 refers to this correspondence table, identifies an extraction rule from the machining program number and tool number included in the machining data, and executes the identified extraction rule to extract current data in the calculation target section. As shown in FIG. 3, the correspondence table includes items of a machining program number that identifies a machining program, a tool number that identifies a tool, and an extraction rule ID that identifies an extraction rule. In the illustrated example, the machining program number "N001" is information that identifies the machining program as a cutting machining program, and the tool number "D01" is information that identifies the tool as a drill. The extraction rule ID "r1" is applied to the machining process executed by the machining program number "N001" and the tool number "D01", and indicates that the current data in the calculation target section is extracted. The extraction rule of the extraction rule ID "r1" is, for example, a rule that extracts state data for 5 seconds from the time when the cutting feed signal is transmitted as the calculation target section.

(Learning model generation process)
Next, a learning model generation process in which the diagnostic system 1 generates a learning model for determining whether or not there is an abnormality in the machining equipment 2 will be described with reference to Fig. 5. The learning model generation process is executed before the first diagnostic process is performed, when it becomes necessary to update the learning model due to replacement of the machining equipment 2 or a tool, when it is desired to improve the accuracy of the determination by using more data, etc. In the following, a case will be described in which, in a machining process in which only cutting is performed, one cycle is defined as the period from the start to the end of machining of one workpiece, and a learning model is generated based on a feature amount calculated from current data for each cycle.

Prior to the execution of the learning model generation process, the user presets in the signal processing device 4 the sampling frequency and resolution for A/D conversion of the analog signal output from the current sensor that measures the current flowing through the spindle motor. The higher the sampling frequency, the more accurately the analog signal can be traced, and the digital signal closer to the true signal can be input to the computer for use in learning. In addition, since the frequency range F_range of the FFT performed after A/D conversion is determined by the following formula 1, it is better to increase the sampling frequency when a wide frequency range is desired.
F_range=fs/2.56 (fs: sampling frequency) Equation 1
In general, it is considered advisable to set the sampling frequency to about 10 times the frequency of the analog signal in order to suppress the occurrence of aliasing, but in order to obtain a more accurate learning model, it is preferable to set it to 1,000 times or more the frequency of the analog signal.

When the user operates the input unit 15 of the learning model generation device 7 to request the start of the learning model generation process, the learning model generation device 7 starts the process.

When cutting of the workpiece by the machining equipment 2 is started, the machining data acquisition device 3 acquires machining data including status data indicating the current value flowing through the spindle motor, the motor rotation speed, and the cutting feed signal, as well as machining-related information such as the machining program number and tool number (step S11). For each cycle, which is the period from the start to the end of machining of one workpiece, the machining data acquisition device 3 lumps together the status data and outputs it to the signal processing device 4 together with the machining-related information.

Next, the signal processing device 4 performs signal processing on the received state data, and extracts state data of the calculation target section from the processed state data (step S12). Specifically, the signal processing device 4 converts the analog signal included in the state data transmitted from the machining data acquisition device 3 into a digital signal at a preset sampling frequency. Note that the analog signal before being converted into a digital signal may be subjected to signal amplification processing, high-pass filtering, and low-pass filtering. Next, the signal processing device 4 acquires the machining program number and tool number included in the machining data output from the machining data acquisition device 3 in step S11, and identifies the extraction rule by referring to the correspondence table. For example, when the signal processing device 4 acquires information on the program number "N001" and the tool number "D01", it identifies the extraction rule ID "r1" as the extraction rule by referring to the correspondence table shown in FIG. 3. Next, the signal processing device 4 acquires the rotation speed of the spindle motor included in the state data after signal processing, and identifies the stable machining section in which the rotation speed is a constant speed section. Next, the signal processing device 4 reads out the extraction rule with the specified extraction rule ID "r1" and executes the extraction rule to extract status data for 5 seconds from the time the cutting feed signal is sent as the calculation target section, and extracts the current data for the calculation target section from the current data for the stable machining section.

Returning to FIG. 5, the signal processing device 4 then performs a cleansing process on the current data in the extracted calculation target section (step S13). Specifically, the signal processing device 4 removes noise and abnormal values from the current data and performs a process to complement missing values.

Then, the signal processing device 4 calculates feature quantities from the processing data cleansed in step S13 (step S14). Specifically, the signal processing device 4 calculates feature quantities by determining the average, maximum, minimum, standard deviation, variance, kurtosis, skewness, etc. of the current values in the cleansed calculation target section. The signal processing device 4 calculates feature quantities for each cycle from the start to the end of cutting processing of one workpiece. The signal processing device 4 sequentially accumulates the calculated feature quantities and the current data used to calculate the feature quantities in the storage device 5, together with the number of cycles indicating which processing has been performed.

Next, the learning model generation device 7 generates learning data (step S15). Specifically, based on the input from the user, the learning model generation device 7 generates learning data by assigning labels to the feature amounts for each cycle stored in the storage device 5 in step S14, indicating whether the data is data when the processing equipment 2 is operating normally or data when an abnormality is present. As illustrated in FIG. 6, the learning data includes a "cycle count" indicating the number of times the workpiece has been processed, "feature amount A", "feature amount B", ... indicating a data set of multiple feature amounts for each cycle calculated by the signal processing device 4 in step S14, and a "state label" indicating the label assigned to the feature amount for each cycle. The learning model generation device 7 stores the generated learning data in the storage device 5.

Returning to FIG. 5, next, the learning model generation device 7 generates a learning model (step S16). Specifically, the learning model is a model according to a machine learning algorithm such as a decision tree, k-nearest neighbor method, or support vector machine, and outputs judgment result information that judges whether the processing equipment 2 to be inspected is normal or abnormal using the feature amount for each cycle calculated from the current data as input. The learning model generation device 7 generates a learning model by performing machine learning using learning data including the feature amount for each cycle contained in the learning data generated in step S15 and label information indicating normality or abnormality assigned to the feature amount for each cycle. The learning model generation device 7 stores the generated learning model in the storage device 5.

In this way, a learning model for determining whether processing equipment 2 is normal or not is completed.

(Diagnosis Processing)
Next, the operation of the diagnosis process for determining whether or not the processing equipment 2 is normal using the learning model in an actual processing situation will be described with reference to FIG.

The learning model generated by the above-mentioned learning model generation process is stored in the storage device 5, and the learning model stored in the storage device 5 is pre-installed in the diagnostic device 6 by the user.

Similar to the learning model generation process, the diagnostic system 1 starts the diagnostic process when the machining equipment 2 starts machining the workpiece. Steps S21 to S24 of the diagnostic process are similar to steps S11 to S14 of the learning model generation process shown in FIG. 5, and sequentially acquire machining data and calculate feature values for each cycle of the cutting process from the current data included in the machining data.

Returning to FIG. 7, next, the diagnostic device 6 judges whether the target processing equipment 2 is normal or abnormal (step S25). Specifically, the diagnostic device 6 inputs the feature amount calculated in step S24 into a learning model, and obtains judgment result information indicating whether the processing equipment 2 is normal or abnormal.

If the judgment result information indicates that the processing equipment 2 is normal (step S26; Yes), the diagnostic device 6 determines whether the processing process has ended, and if it determines that it has ended (step S27; Yes), it ends the diagnostic process. If the diagnostic device 6 determines that the processing process has not ended (step S27; No), it returns to step S21 and obtains the processing data for the next cycle (step S21).

Returning to step S26, if the judgment result information indicates that the processing equipment 2 is abnormal (step S26; No), the diagnostic device 6 notifies the status display 8 and the control device 9 that it has determined that the processing equipment 2 is abnormal (step S28), and ends the process. Thereafter, the status display 8 notifies the user of the occurrence of an abnormality by displaying on the screen information indicating that the processing equipment 2 has been diagnosed as abnormal. The control device 9 stops the processing equipment 2.

As described above, the diagnostic system 1 extracts status data for the calculation target section, which is a time region in which abnormalities are likely to occur, from the status data for the section from the start to the end of machining of one workpiece, and calculates the feature values. The diagnostic system 1 determines whether or not there is an abnormality in the machining equipment 2 based on the calculated feature values. The feature values calculated from the status data for the calculation target section have a larger difference between normal and abnormal values than feature values calculated from all data within the machining section, making it possible to accurately determine whether or not there is an abnormality.

In addition, because the diagnostic system 1 determines the presence or absence of an abnormality using less data, there is no need to store and process a large amount of data. This improves the processing speed of the diagnostic process and makes it possible to save on storage capacity.

　As long as it does not deviate from the spirit of this disclosure, it is possible to select and discard the configurations described in the above embodiments, or to change them to other configurations as appropriate.

In the above embodiment, the functions of the diagnostic system 1 have been described as being executed by individual devices, namely the processed data acquisition device 3, the signal processing device 4, the storage device 5, the diagnostic device 6, the learning model generation device 7, the status display device 8, and the control device 9, but this is not limited thereto, and the functions of each device may be executed by an information processing device that is a single computer. Also, each process of the diagnostic system 1 may be executed by multiple computers that execute the functions of several devices, such as executing the functions of the processed data acquisition device 3 and the signal processing device 4 by a single computer, and executing the functions of the diagnostic device 6 and the learning model generation device 7 by a single computer.

In the above explanation, the diagnostic system 1 is provided with a learning model generating device 7, but the diagnostic system 1 does not need to be provided with a learning model generating device 7. For example, the teacher data may be supplied to a machine learning device prepared on another computer to generate a learning model, which may then be set in the diagnostic device 6.

In the above embodiment, the diagnostic device 6 uses a learning model to determine whether or not there is an abnormality in the processing equipment 2, but this is not limited to the above. For example, the presence or absence of an abnormality in the processing equipment 2 may be determined by a diagnostic rule in which a threshold is set for a feature amount calculated from the state data, and it is determined that an abnormality has occurred in the processing equipment 2 if the feature amount is equal to or greater than the threshold. In this case, in the process of setting the calculation target section, the user determines the threshold value for the feature amount using the state data in which the abnormality has occurred, and sets the diagnostic rule in the diagnostic device 6. In step S25, the diagnostic device 6 reads out the set diagnostic rule and determines whether or not there is an abnormality by comparing the feature amount with the threshold.

The status display 8 may not only display the diagnosis results, but also the processed data stored in the storage device 5. Furthermore, since the feature values calculated from the status data are stored sequentially in the storage device 5 during the diagnosis process, the feature values may be displayed in real time using trend graphs, scatter diagrams, etc.

The diagnostic device 6 may not only determine whether an abnormality is currently occurring in the processing equipment 2, but may also predict when an abnormality will occur. Specifically, the diagnostic device 6 creates a trend graph of the feature quantities, performs linear regression analysis or nonlinear regression analysis, and fits an approximation curve to predict future feature quantity values and calculate the time until an abnormality occurs. When the time lag becomes short and exceeds a preset threshold, the status display 8 may notify the operator of an alarm or display a message urging maintenance. The status display 8 may also display the threshold along with the trend graph.

In addition, in the above embodiment, a case has been described in which one processing step is performed in one cycle, but this is not limiting, and multiple processing steps such as cutting, cutting, and polishing may be performed in one cycle. In this case, an extraction rule for extracting a calculation section for each processing step is set, and a learning model is generated by calculating feature values from the state data of each calculation section. Then, the diagnostic device 6 inputs each feature value into each learning model and determines whether or not there is an abnormality in the processing equipment 2.

In addition, the information stored in the storage device 5 may be managed collectively by a cloud server on the network, and the signal processing device 4, diagnostic device 6, and learning model generation device 7 may access the cloud server as necessary to read and write information. In this case, the diagnostic system 1 may not need to include a storage device 5. Furthermore, the signal processing of data by the signal processing device 4 and the learning model generation processing by the learning model generation device 7 may be performed on the cloud using information stored in the cloud server.

The above explanation has focused on an example in which the signal processing device 4 sets one or more calculation intervals for each period from the start to the end of machining of one workpiece, but if one cycle includes multiple machining processes, a calculation interval may be set for each machining process, and the quality of the machining may be determined for each machining process.

In addition, the signal processing device 4 has been described with a focus on an example in which the state data of the calculation target section is extracted based on a preset extraction rule, but a section in which the difference between normal processing and abnormality occurrence is likely to appear may be specified as the calculation target section based on the output value from the sensor installed in the processing equipment 2. Specifically, for example, in the above-mentioned advance preparation stage, the signal processing device 4 obtains the difference in the sensor output value between the normal processing section and the abnormality occurrence section from multiple time-series state data including the sensor output value when the abnormality occurs. The difference in the sensor output value is obtained, for example, from the difference between the average value, maximum value, minimum value, etc. of the current value of the normal processing section and the abnormality occurrence section. Next, the signal processing device 4 specifies the calculation target section based on the abnormality occurrence section of the state data in which the obtained difference is large. For example, the abnormality occurrence section of the state data in which the difference is the largest may be specified as the calculation target section from multiple state data, or a section such as 3 seconds before and after or 5 seconds before and after the abnormality occurrence section may be specified as the calculation target section, or a time region including each abnormality occurrence section of multiple state data in which the difference is large, or a time region overlapping each abnormality occurrence section may be specified as the calculation target section. For example, if the abnormality occurrence intervals of the three status data with large differences are between 10 seconds and 15 seconds after the start of actual machining, between 8 seconds and 18 seconds after the start of actual machining, and between 12 seconds and 20 seconds after the start of actual machining, the calculation target interval may be specified as the time region that includes each of the abnormality occurrence intervals, between 8 seconds and 20 seconds after the start of actual machining, or the calculation target interval may be specified as the time region that overlaps each of the abnormality occurrence intervals, between 12 seconds and 15 seconds after the start of actual machining. The signal processing device 4 may then specify the calculation target interval for each machining process and set an extraction rule for extracting the status data of the specified calculation target interval. In the process of step S12 or step S22, the signal processing device 4 may extract the status data of the calculation processing interval specified for each machining process from the status data acquired by the machining data acquisition device 3.

The signal processing device 4 may also divide the stable processing section shown in FIG. 2 into a plurality of sections, calculate the feature value for each of the divided sections, and identify the calculation target section according to the calculated feature value for each section. For example, the signal processing device 4 divides the plurality of state data acquired by the processing data acquisition device 3 into a first section, a second section, a third section, ... by a set number such as 5 or 10. Next, the signal processing device 4 calculates the feature value for each divided section for each state data. Next, the signal processing device 4 may determine the variation in the feature value for each section of the plurality of state data, the first section, the second section, the third section, ..., and compare the respective variations to identify the section with the largest variation as the calculation target section, or determine the average value of the feature value for each section and identify the section with the largest difference in the average value with other sections as the calculation target section.

Furthermore, the processed data acquisition device 3, the signal processing device 4, the storage device 5, the diagnostic device 6, and the learning model generation device 7 can be realized using a normal computer system, rather than using dedicated devices. For example, a program for realizing each function can be stored and distributed on a computer-readable recording medium such as a CD-ROM (Compact Disc Read Only Memory) or a DVD-ROM (Digital Versatile Disc Read Only Memory), and a computer that can realize each of the above-mentioned functions can be configured by installing this program on a computer.

In addition, if each function is realized by sharing the responsibilities of an OS (Operating System) and an application, or by cooperation between an OS and an application, only the application may be stored on the recording medium.

This disclosure allows for various embodiments and modifications without departing from the broad spirit and scope of the disclosure. Furthermore, the above-described embodiments are intended to explain the disclosure and do not limit the scope of the disclosure. In other words, the scope of the disclosure is indicated by the claims, not the embodiments. Furthermore, various modifications made within the scope of the claims and within the scope of the disclosure equivalent thereto are considered to be within the scope of the disclosure.

This application is based on Japanese Patent Application No. 2022-159730, filed on October 3, 2022. The entire specification, claims, and drawings of Japanese Patent Application No. 2022-159730 are incorporated herein by reference.

The various aspects of this disclosure are summarized below as appendices.

(Appendix 1)
an operation status information acquisition unit that acquires operation status information indicating an operation status of the processing equipment to be diagnosed;
an extracting unit that extracts, from the operation state information acquired by the operation state information acquiring unit, operation state information of a calculation target section that indicates a time region determined by a preset rule within a stable processing section in which the motor rotates stably;
a feature amount calculation unit that calculates a feature amount from the motion state information of the calculation target section extracted by the extraction unit;
a diagnosis unit that determines whether the processing equipment is normal or abnormal based on the feature amount calculated by the feature amount calculation unit;
A diagnostic system comprising:

(Appendix 2)
The diagnosis unit inputs the feature amount calculated by the feature amount calculation unit into a preset machine learning model, and based on the output obtained, determines whether the processing equipment to be diagnosed is normal or abnormal.
2. The diagnostic system of claim 1.

(Appendix 3)
a learning model generation unit that generates the machine learning model based on learning data including the features calculated by the feature calculation unit and label information assigned to each feature;
3. The diagnostic system of claim 2, further comprising:

(Appendix 4)
The diagnostic device further includes a processing identification information acquisition unit that acquires processing identification information that identifies processing content performed by the diagnostic target processing equipment,
the extraction unit specifies a rule for extracting the operation state information of the calculation target section from the operation state information of the processing equipment to be diagnosed based on correspondence information that associates the processing content with a rule for extracting the operation state information of the calculation target section, and executes the specified rule to extract the operation state information to be used in calculating the feature amount.
4. The diagnostic system of any one of claims 1 to 3.

(Appendix 5)
A control unit that controls an operation of the processing equipment to be diagnosed,
When the diagnosis unit determines that the processing equipment to be diagnosed is abnormal, the diagnosis unit notifies the control unit of a determination result indicating that the processing equipment is abnormal;
When the control unit receives the determination result from the diagnosis unit, the control unit stops operation of a part or all of the processing equipment to be diagnosed.
5. The diagnostic system of any one of claims 1 to 4.

(Appendix 6)
The operational status information includes a sensor output indicating a physical quantity detected by a sensor installed in the processing equipment,
the extraction unit obtains a difference between the sensor output in a normal processing section and the sensor output in the abnormality occurrence section from a plurality of pieces of operation status information including operation states of the abnormality occurrence section of the processing equipment, identifies the calculation target section based on the abnormality occurrence section of the operation status information including a sensor output in which the obtained difference is large, and extracts operation status information of the identified calculation target section from the operation status information acquired by the operation status information acquisition unit.
6. The diagnostic system of any one of claims 1 to 5.

(Appendix 7)
The extraction unit divides the stable processing section of the operation status information acquired by the operation status information acquisition unit into a plurality of sections, calculates a feature amount for each of the divided sections, compares the calculated feature amounts for each of the sections, and based on a comparison result, identifies the calculation target section, and extracts operation status information of the identified calculation target section.
7. The diagnostic system of any one of claims 1 to 6.

(Appendix 8)
an extraction unit that extracts, from operation status information indicating the operation status of the processing equipment to be diagnosed, operation status information of a calculation target section that indicates a time region determined by a preset rule within a stable processing section in which a motor rotates stably;
a diagnosis unit that judges whether the processing equipment to be diagnosed is normal or abnormal based on the operation state information of the calculation target section extracted by the extraction unit;
An information processing device comprising:

(Appendix 9)
A step of extracting operation status information of a calculation target section, which indicates a time region determined by a preset rule within a stable machining section in which a motor rotates stably, from operation status information indicating an operation status of the machining equipment to be diagnosed;
and determining whether the processing equipment to be diagnosed is normal or abnormal based on the extracted operating state information of the calculation target section.

(Appendix 10)
On the computer,
A process of extracting operation status information of a calculation target section, which indicates a time region determined by a preset rule within a stable machining section in which a motor rotates stably, from operation status information indicating an operation status of the machining equipment to be diagnosed;
A process of determining whether the processing equipment to be diagnosed is normal or abnormal based on the extracted operation state information of the calculation target section;
A program that executes the following.

1 Diagnostic system, 2 Machining equipment, 3 Machining data acquisition device, 4 Signal processing device, 5 Storage device, 6 Diagnostic device, 7 Learning model generation device, 8 Status display device, 9 Control device, 99 Internal bus, 11 CPU, 12 RAM, 13 ROM, 14 Memory unit, 15 Input unit, 16 Communication unit.

Claims

an operation status information acquisition unit that acquires operation status information indicating an operation status of the processing equipment to be diagnosed;
an extracting unit that extracts, from the operation state information acquired by the operation state information acquiring unit, operation state information of a calculation target section that indicates a time region determined by a preset rule within a stable processing section in which the motor rotates stably;
a feature amount calculation unit that calculates a feature amount from the motion state information of the calculation target section extracted by the extraction unit;
a diagnosis unit that determines whether the processing equipment is normal or abnormal based on the feature amount calculated by the feature amount calculation unit;
A diagnostic system comprising:
The diagnosis unit inputs the feature amount calculated by the feature amount calculation unit into a preset machine learning model, and based on the output obtained, determines whether the processing equipment to be diagnosed is normal or abnormal.
The diagnostic system of claim 1 .
a learning model generation unit that generates the machine learning model based on learning data including the features calculated by the feature calculation unit and label information assigned to each feature;
The diagnostic system of claim 2 , further comprising:
The diagnostic device further includes a processing identification information acquisition unit that acquires processing identification information that identifies processing content performed by the diagnostic target processing equipment,
the extraction unit specifies a rule for extracting the operation state information of the calculation target section from the operation state information of the processing equipment to be diagnosed based on correspondence information that associates the processing content with a rule for extracting the operation state information of the calculation target section, and executes the specified rule to extract the operation state information to be used in calculating the feature amount.
A diagnostic system according to any one of claims 1 to 3.
A control unit that controls an operation of the processing equipment to be diagnosed,
When the diagnosis unit determines that the processing equipment to be diagnosed is abnormal, the diagnosis unit notifies the control unit of a determination result indicating that the processing equipment is abnormal;
When the control unit receives the determination result from the diagnosis unit, the control unit stops operation of a part or all of the processing equipment to be diagnosed.
A diagnostic system according to any one of claims 1 to 4.
The operational status information includes a sensor output indicating a physical quantity detected by a sensor installed in the processing equipment,
the extraction unit obtains a difference between the sensor output in a normal processing section and the sensor output in the abnormality occurrence section from a plurality of pieces of operation status information including operation states of the abnormality occurrence section of the processing equipment, identifies the calculation target section based on the abnormality occurrence section of the operation status information including a sensor output in which the obtained difference is large, and extracts operation status information of the identified calculation target section from the operation status information acquired by the operation status information acquisition unit.
A diagnostic system according to any one of claims 1 to 5.
The extraction unit divides the stable processing section of the operation status information acquired by the operation status information acquisition unit into a plurality of sections, calculates a feature amount for each of the divided sections, compares the calculated feature amounts for each of the sections, and based on a comparison result, identifies the calculation target section, and extracts operation status information of the identified calculation target section.
A diagnostic system according to any one of claims 1 to 6.
an extraction unit that extracts, from operation status information indicating the operation status of the processing equipment to be diagnosed, operation status information of a calculation target section that indicates a time region determined by a preset rule within a stable processing section in which a motor rotates stably;
a diagnosis unit that judges whether the processing equipment to be diagnosed is normal or abnormal based on the operation state information of the calculation target section extracted by the extraction unit;
An information processing device comprising:
A step of extracting operation status information of a calculation target section, which indicates a time region determined by a preset rule within a stable machining section in which a motor rotates stably, from operation status information indicating an operation status of the machining equipment to be diagnosed;
A step of determining whether the processing equipment to be diagnosed is normal or abnormal based on the extracted operation state information of the calculation target section;
A diagnostic method comprising:
On the computer,
A process of extracting operation status information of a calculation target section, which indicates a time region determined by a preset rule within a stable machining section in which a motor rotates stably, from operation status information indicating an operation status of the machining equipment to be diagnosed;
A process of determining whether the processing equipment to be diagnosed is normal or abnormal based on the extracted operation state information of the calculation target section;
A program that executes the following.