Abnormality detection method and system for machining equipment and storage medium

Abstract

The invention discloses an anomaly detection method and system for machining equipment and a storage medium, and the method comprises the steps: obtaining machining data of the machining equipment in the workpiece machining period, obtaining the type of a workpiece in the machining data, separating the machining data based on the type of the machined workpiece, and selecting a machine learning model corresponding to the workpiece type of the separated processing data to process the separated processing data so as to obtain an anomaly detection result. According to the anomaly detection method of the machining equipment, the anomaly of the machining equipment can be accurately detected, so that the yield of machined workpieces is improved.

Description

Technical Field

[0001] The present disclosure relates to an abnormality detection method, system, and storage medium for machining equipment. Background Art

[0002] Various mechanical processing equipment may experience abnormalities during operation, such as abnormal vibrations. Currently, in order to detect abnormalities in mechanical processing equipment, maintenance engineers can evaluate abnormalities or failures in the mechanical processing equipment by measuring relevant data when the mechanical processing equipment is running / stopped, or they can detect abnormalities based on upper and lower thresholds of processing data set for the mechanical processing equipment. However, depending on the qualifications of the maintenance engineers themselves, there is a great deal of uncertainty in the evaluation of mechanical processing equipment by maintenance engineers. In addition, since mechanical processing equipment may process different workpieces, it may be very challenging to accurately detect abnormalities in mechanical processing equipment based on the unified upper and lower thresholds of processing data set for the mechanical processing equipment. Therefore, there is a need for a method that can detect abnormalities in mechanical processing equipment based on the specific processing conditions of the mechanical processing equipment. Summary of the Invention

[0003] According to an embodiment of the present disclosure, a method for detecting anomalies in a mechanical processing device is provided, comprising: obtaining processing data of the mechanical processing device during processing of a workpiece, obtaining the type of the workpiece in the processing data, separating the processing data based on the type of the processed workpiece, and selecting a machine learning model corresponding to the type of the workpiece in the separated processing data to process the separated processing data to obtain an anomaly detection result.

[0004] According to the abnormality detection method of an embodiment of the present disclosure, the processing data includes at least one of mechanical power output and electrical power received by the machining equipment during machining of the workpiece.

[0005] According to the abnormality detection method of the embodiment of the present disclosure, the type of workpiece in machining data is obtained by using a scheduling history in a machining device and the machining data is separated.

[0006] According to an abnormality detection method of an embodiment of the present disclosure, obtaining processing data of a mechanical processing device during processing of a workpiece includes: obtaining full-period processing data including processing period processing data and stop period processing data, removing the stop period processing data from the full-period processing data to obtain processing period processing data; and wherein, selecting a machine learning model corresponding to the workpiece type of the separated processing data to process the separated processing data includes: processing the processing period processing data in the separated processing data.

[0007] According to the abnormality detection method of the embodiment of the present disclosure, the stop period processing data is removed from the full period processing data through a machine learning classification model or based on a threshold.

[0008] According to the abnormality detection method of an embodiment of the present disclosure, obtaining the processing data of the mechanical processing equipment during processing of the workpiece also includes: removing the impact pulse data of the mechanical processing equipment at startup and shutdown from the processing period processing data to obtain the effective processing period processing data; and wherein, selecting the machine learning model corresponding to the workpiece type of the separated processing data to process the separated processing data includes: processing the effective processing period processing data in the separated processing data.

[0009] According to the abnormality detection method of an embodiment of the present disclosure, it also includes: obtaining the processing part of the processed workpiece, separating the processing data based on the type and processing part of the processed workpiece, and selecting a machine learning model corresponding to the workpiece type and processing part of the separated processing data to process the separated processing data to obtain an abnormality detection result.

[0010] According to the abnormality detection method of an embodiment of the present disclosure, it also includes: obtaining the processing stage of the processed workpiece, separating the processing data based on the type, processing part and processing stage of the processed workpiece, and selecting a machine learning model corresponding to the workpiece type, processing part and processing stage of the separated processing data to process the separated processing data to obtain an abnormality detection result.

[0011] According to the abnormality detection method of an embodiment of the present disclosure, the processing position and processing stage of the workpiece in the processing data are obtained based on the mechanical processing principle or machine learning classification model of the processing workpiece.

[0012] According to an embodiment of the present disclosure, there is provided an abnormality detection system for mechanical processing equipment, comprising: a sensor unit, comprising one or more sensors, configured to collect processing data about the mechanical processing equipment; an abnormality detection model library, comprising one or more abnormality detection models configured to perform abnormality detection on the mechanical processing equipment; and a processor, configured to: obtain processing data of the mechanical processing equipment during processing of a workpiece from the sensor unit, obtain the type of the workpiece in the processing data, separate the processing data based on the type of the processed workpiece, select a machine learning model corresponding to the workpiece type of the separated processing data from the abnormality detection model library, process the separated processing data, and obtain an abnormality detection result.

[0013] According to an embodiment of the present disclosure, a non-transitory storage medium is provided, on which instructions are stored. When executed by a processor, the instructions enable the processor to execute the above-mentioned abnormality detection method for machining equipment.

[0014] According to the abnormality detection method for mechanical processing equipment disclosed in the present invention, by processing processing data using a machine learning abnormality detection model related to the workpiece type, processing location, and processing stage, accurate abnormality detection of the mechanical processing equipment can be performed, thereby improving the yield of the processed workpiece. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and other aspects, features and advantages of certain embodiments of the present disclosure will become more apparent from the following description taken in conjunction with the accompanying drawings, in which:

[0016] Figure 1 is a flow chart of an anomaly detection method according to an embodiment of the present disclosure;

[0017] Figure 2 is an example graph of machining data of a type of workpiece according to an embodiment of the present disclosure;

[0018] Figure 3 is another flow chart of an anomaly detection method according to an embodiment of the present disclosure;

[0019] Figure 4 is an example graph of machining data of two workpieces according to an embodiment of the present disclosure;

[0020] Figure 5 is a schematic diagram of a system according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0021] Before proceeding with the following detailed description, it may be advantageous to set forth the definitions of certain words and phrases used throughout this disclosure. The terms "include" and "comprising" and their derivatives mean to include, but are not limited to. The term "controller" or "control unit" means any device, system, or part thereof that controls at least one operation. Such a controller may be implemented in hardware, or a combination of hardware and software and / or firmware. The functionality associated with any particular controller may be centralized or distributed, whether local or remote. The phrase "at least one of," when used with a list of items, means that different combinations of one or more of the listed items may be used, and only one item from the list may be required. For example, "at least one of A, B, C" includes any one of the following combinations: A, B, C, A and B, A and C, B and C, A and B and C.

[0022] Definitions for other specific words and phrases are provided throughout this disclosure. Those of ordinary skill in the art should understand that in many, if not most instances, such definitions apply to prior and future uses of such defined words and phrases.

[0023] The various embodiments of the principles of the present disclosure described below in conjunction with the accompanying drawings are merely illustrative and should not be construed as limiting the scope of the present disclosure in any way. It will be understood by those skilled in the art that the principles of the present disclosure can be implemented in any appropriately arranged system or device. In some cases, the actions described in the present disclosure can be performed in a different order and still achieve the desired result. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired result. In specific embodiments, multitasking and parallel processing may be advantageous.

[0024] The text and drawings are provided as examples only to aid understanding of the present disclosure. They should not be construed as limiting the scope of the claims appended hereto in any way. Although certain embodiments and examples have been provided, it will be apparent to those skilled in the art based on the contents of this disclosure that modifications may be made to the embodiments and examples shown without departing from the scope of this disclosure.

[0025] The mechanical processing equipment used throughout this disclosure may include various general and special mechanical processing equipment, including but not limited to machine tools (for example, grinding machines, lathes, drilling machines, boring machines, gear processing machines, thread processing machines, milling machines, planers, broaching machines, sawing machines and other machine tools, etc.), processing and manufacturing equipment (for example, food processing equipment, textile equipment, chemical equipment, assembly equipment, etc.), etc.

[0026] Figure 1 Flowchart of an abnormality detection method according to an embodiment of the present disclosure.

[0027] At S102, processing data of the machining equipment during machining of the workpiece can be obtained. The processing data can be collected by sensors and can include data such as power, sound, vibration, force, speed, etc., but the present disclosure is not limited thereto. For example, the processing data can include at least one of the mechanical power output (e.g., the power of the grinding wheel spindle) and the electrical power received by the machining equipment during machining of the workpiece. For example, the processing data can be sampled at a high sampling rate of at least 100 Hz.

[0028] In S104, the type of the workpiece in the processing data may be obtained. For example, the type of the processed workpiece may include screws, gears, etc., but the present disclosure is not limited thereto.

[0029] At S106, the processing data may be separated based on the type of the workpiece being processed. A machining device may process different types of workpieces. Therefore, processing data corresponding to each type of workpiece may be separated from processing data including multiple types of workpieces.

[0030] At S108, a machine learning model corresponding to the workpiece type of the separated processing data may be selected and used to process the separated processing data to obtain anomaly detection results. The anomaly may include abnormal vibration or tremor, but the present disclosure is not limited thereto. The selected machine learning model may be used to process the processing data separated based on the workpiece type in step S106. For example, the selected machine learning model may be built or trained using previously extracted features of processing data of a specific workpiece type.

[0031] For example, mechanical processing equipment can process different types of workpieces, and the processing data may include processing data for processing three types of workpieces: A, B, and C. Due to the different types of workpieces, the pattern of the processing data can vary greatly, so that the traditional method of detecting anomalies in the processing data through a unified threshold may deviate greatly from the actual situation. According to the method disclosed in the present invention, the processing data can be separated based on the type of workpiece being processed, and the data can be processed using a corresponding machine learning model to detect anomalies. For example, the processing data of a workpiece of type A can be separated from the processing data, and the processing data of the workpiece of type A can be processed using a machine learning model corresponding to the workpiece of type A. In this way, the accuracy of anomaly detection can be improved, thereby improving the yield of the processed workpiece.

[0032] Figure 2 FIG. 1 is an example graph of machining data of one type of workpiece according to an embodiment of the present disclosure.

[0033] A machining device can process different types of workpieces, and the machining data can include machining data for different types of workpieces. The machining data can be separated by using the scheduling history of the machining device to determine the type of workpiece in the machining data. For example, the machining device can store a scheduling history that records the types and machining times of all workpieces machined by the machining device during a certain period of time. The machining data can be separated based on the workpiece type and machining time. For example, machining data for different types of workpieces can be separated based on the scheduling history, and further, machining data for a single workpiece within machining data for workpieces of the same type can be separated based on the scheduling history.

[0034] Figure 2 The full-time curve diagram of the processing data of one type of workpiece after separation is shown in FIG. Figure 2As shown, the full-time curve includes both processing and stop periods. While machining a workpiece, the machining equipment is not always in the processing state. In other words, the machining equipment may not be processing a workpiece for a certain period of time, but may be in a shutdown or standby state. The period between two processing periods is the stop period. The machining period includes the active processing period, the processing start period, and the processing end period. During the processing start and end periods, shock pulses are generated due to the startup and shutdown of the machining equipment.

[0035] The full-time processing data may include processing data during the processing period and processing data during the stop period. For example, processing data during the processing period may refer to data of the machining equipment actually processing the workpiece. Processing data during the stop period may refer to data of the machining equipment itself during the time when the machining equipment is not processing the workpiece. For example, processing data during the stop period may approach 0. In order to obtain more accurate anomaly detection results, processing data during the stop period may be identified and removed from the full-time processing data to obtain processing data during the processing period. By processing the processing data during the processing period in the separated processing data using a selected machine learning model corresponding to the workpiece type of the separated processing data, more accurate anomaly detection results can be obtained.

[0036] The stopped period processed data can be removed from the full period processed data based on a machine learning classification model or a threshold. For example, processed data with an amplitude lower than a threshold can be determined as stopped period processed data and removed from the full period processed data to obtain the processing period processed data.

[0037] In another embodiment, the processing data during the stop period may be determined based on a machine learning classification model, thereby removing the processing data during the stop period from the processing data during the full period.

[0038] A machine learning classification model can be trained using previously acquired processing data during the processing period and processing data during the stop period. Features of the processing data during the processing period and processing data during the stop period can be extracted using root mean square (RMS), peak-to-peak (P2P), average, standard deviation, Envelope 3 (Envolop3), time-frequency transforms (such as FFT transforms), numerical operations, and the like.

[0039] In one embodiment, the machine learning classification model may include a distribution-based machine learning model (such as 3sigma, Z-score, boxplot, etc.), a distance-based machine learning model (such as K-nearest neighbor (KNN)), a density-based machine learning model (such as Local Outlier Factor (LOF), Connectivity-Based Outlier Factor (COF), Stochastic Outlier Selection (SOS)), etc.), a clustering-based machine learning model (such as Density-Based Spatial Clustering of Applications with Noise (DBSCAN)), etc.), a tree-based machine learning model (such as Isolation Forest (iForest)), a dimensionality reduction-based machine learning model (such as Principal Component Analysis (PCA), AutoEncoder, etc.), a classification-based machine learning model (such as One-Class Support Vector Machine (SVM), etc.), a classification-based machine learning model (such as One-Class Support Vector Machine (SVM), etc.), a classification-based machine learning model (such as CNN), a distance-based machine learning model (such as K-nearest neighbor (KNN)), a density-based machine learning model (such as Local Outlier Factor (LOF), Connectivity-Based Outlier Factor (COF), Stochastic Outlier Selection (SOS)), etc.), a clustering-based machine learning model (such as Density-Based Spatial Clustering of Applications with Noise (DBSCAN)), etc.), a tree-based machine learning model (such as Isolation Forest (iForest)), etc.), a dimensionality reduction-based machine learning model (such as Principal Component Analysis (PCA), AutoEncoder, etc.), a classification-based machine learning model (such as One-Class Support Vector Machine (SVM), etc.), a distance-based machine learning model (such as K-nearest neighbor (KNN)), a density-based machine learning model (such as Local Outlier Factor (LOF), Connectivity-Based Outlier Factor (COF), Stochastic Outlier Selection (SOS)), etc.), a clustering-based machine

[0014] The present invention also provides a method for classifying machine learning models, such as SVM, prediction-based machine learning models (such as moving average, autoregressive integrated moving average model (ARIMA), etc.). Although the example embodiments show some machine learning classification models, those skilled in the art will appreciate that the above description is merely exemplary and not exhaustive, and other existing or future developed machine learning classification models may be used, all of which are within the contemplation of the present disclosure.

[0040] like Figure 2 As shown, the processing data during the processing period may include shock pulse data and effective processing period processing data. The mechanical processing equipment may generate a large shock pulse when it is started and shut down. The shock pulse may be significantly higher than the processing data during normal processing of the workpiece by the mechanical processing equipment. In order to obtain more accurate abnormality detection results, the shock pulse data of the mechanical processing equipment during startup and shutdown may be identified, and the shock pulse data of the mechanical processing equipment during startup and shutdown may be removed from the processing data during the processing period to obtain the effective processing period processing data. By processing the effective processing period processing data in the separated processing data using a selected machine learning model corresponding to the workpiece type of the separated processing data, a more accurate abnormality detection result may be obtained.

[0041] In one embodiment, based on the findpeaks function, shock pulse data can be identified from the machining data during the machining period, and shock pulse data at startup and shutdown of the machining equipment can be removed from the machining data during the machining period to obtain effective machining data during the machining period.

[0042] In another embodiment, shock pulse data from the start-up and shutdown of the machining equipment can be removed from the machining data during the machining period based on a machine learning classification model or a threshold. For example, machining data with an amplitude exceeding a threshold can be identified as shock pulse data and removed from the full-period machining data to obtain the effective machining data.

[0043] In another embodiment, the shock pulse data may be determined based on a machine learning classification model, thereby removing the shock pulse data from the machining period machining data.

[0044] A machine learning classification model can be trained using previously acquired effective machining period processing data and shock pulse data. Features of the effective machining period processing data and shock pulse data can be extracted using root mean square (RMS), peak-to-peak (P2P), average, standard deviation, Envolop3, time-frequency transforms (such as FFT transforms), and numerical operations.

[0045] The types of machine learning classification models used to identify impact pulse data may be similar to the types of machine learning classification models used to identify stop period processing data, and will not be described in detail here.

[0046] Figure 3 is another flow chart of an anomaly detection method according to an embodiment of the present disclosure.

[0047] In S301, processing data of a machining device during machining of a workpiece may be obtained. For example, the processing data may include at least one of the mechanical power output (e.g., the power of a grinding wheel spindle) and the electrical power received by the machining device during machining of the workpiece. For example, the processing data may be sampled at a high sampling rate of at least 100 Hz. For other descriptions of the processing data, please refer to the description of the processing data. Figure 1 Description of S102.

[0048] In S302, the type of the workpiece in the processing data may be obtained. For example, the type of the workpiece to be processed may include screws, gears, etc., but the present disclosure is not limited thereto.

[0049] In S303, the processing data can be separated based on the type of the workpiece being processed. The machining equipment can process different types of workpieces. Therefore, the processing data corresponding to each type of workpiece can be separated from the processing data including multiple types of workpieces.

[0050] Reference Figure 4 describe Figure 3 . Figure 4 is an example graph of processing data of a workpiece according to an embodiment of the present disclosure. Figure 4 As shown, the processing data may include processing data for two types of workpieces (e.g., workpiece type A and workpiece type B). The types of workpieces in the processing data can be obtained and separated by using the scheduling history of the machining equipment. For example, the machining equipment may store a scheduling history that records the types and processing times of all workpieces processed by the machining equipment during a certain period of time. Based on the workpiece type and processing time, the processing data can be separated into the processing data for workpiece type A and the processing data for workpiece type B.

[0051] In one embodiment, as described above, the processing data of the workpiece type A can be obtained from the processing data of the processing period and the effective processing period. The process of obtaining the processing data of the processing period and the effective processing period can be the same as Figure 2 The description is similar to that in , and will not be repeated here.

[0052] Return to reference Figure 3 In step S304, a processing portion of the workpiece in the processing data may be obtained. For example, the workpiece may have multiple processing portions to be processed. For example, the processing portion of the workpiece may include an upper tooth and a lower tooth, but the present disclosure is not limited thereto.

[0053] In S305, the processing data can be further separated based on the processing part of the workpiece. Figure 4 As shown, the effective processing period processing data of workpiece type A can be further separated into upper gear processing data and lower gear processing data. Figure 4 In the embodiment, only the processing data of workpiece type A is further separated, but those skilled in the art will appreciate that similar separation can be performed on the processing data of workpiece type B.

[0054] return Figure 3 In S304, the processing stage of the workpiece in the processing data may be further obtained. For example, the processing stage may include rough processing, semi-fine processing, and fine processing, but the present disclosure is not limited thereto.

[0055] In S305, the processing data can be further separated based on the processing stage of the workpiece being processed. Figure 4 As shown, the upper gear processing data of workpiece type A can be further divided into rough processing stage, semi-fine processing stage, and fine processing stage. Figure 4In the figure, only the upper gear processing data of workpiece type A is further separated, but those skilled in the art will appreciate that similar separation can be performed on the lower gear processing data of workpiece type A and the processing data of workpiece type B.

[0056] Return to reference Figure 3 In S306, a machine learning model corresponding to the workpiece type and processing part of the separated processing data can be selected to process the separated processing data to obtain an abnormality detection result. Figure 4 As shown, an anomaly can be detected during the machining of the upper teeth of workpiece type A. By processing the separated machining data using a machine learning model selected based on the workpiece type and machining location of the separated machining data, more accurate anomaly detection results can be obtained. For example, the selected machine learning model can be built or trained using previously extracted features from machining data of a specific machining location of a specific workpiece type.

[0057] Furthermore, a machine learning model corresponding to the workpiece type, processing location, and processing stage of the separated processing data can be selected to process the separated processing data to obtain anomaly detection results. Figure 4 As shown, an anomaly can be detected during the semi-fine machining phase in the upper gear machining data for workpiece type A. By processing the separated machining data using a machine learning model selected to correspond to the workpiece type, machining location, and machining phase of the separated machining data, more accurate anomaly detection results can be obtained. For example, the selected machine learning model can be built or trained using previously extracted features of machining data for a specific machining location of a specific workpiece type and a specific machining phase.

[0058] In S304-S305, the processing position and processing stage of the workpiece in the processing data can be obtained based on the mechanical processing principle of the workpiece or the machine learning classification model.

[0059] In one embodiment, the number of times a workpiece is processed and the processing location and processing stage corresponding to the number of processing times can be determined based on the machining principle of the type of workpiece being processed. By comparing the pulse count of the processing signal in the processing data of the effective processing period with the number of processing times based on the machining principle, the processing location and processing stage of the workpiece can be determined. For example, it can be determined based on the machining principle that a workpiece of workpiece type A needs to be processed 6 times, of which the first three times are the upper tooth processing process and the last three times are the rear tooth processing process. In addition, it can be determined based on the machining principle that the three processes included in the upper tooth processing process are, in order, a rough processing process, a semi-fine processing process, and a fine processing process. In this way, the processing location and processing stage of the processed workpiece can be obtained, and the processing data corresponding to the workpiece type can be further separated.

[0060] In another embodiment, the processing position and processing stage of the workpiece in the processing data can be determined by a machine learning classification model. For example, the machine learning classification model for determining the processing position and processing stage of the workpiece in the processing data may include a decision tree, a random forest, a logistic regression, a naive Bayes, etc. Although the example embodiment shows some classification models using machine learning, it should be understood by those skilled in the art that the above description is merely exemplary and not exhaustive, and other existing or future developed models for classification can be used to determine the processing position and processing stage of the workpiece in the processing data, all of which are within the scope of the present disclosure. The machine learning classification model can be trained by the features of the processing data of the processing position and processing stage of the workpiece previously obtained. For example, the features of the processing data of each processing position and processing stage of each type of workpiece can be extracted by root mean square (RMS), peak to peak (P2P), average value, standard deviation, Envolop3, time-frequency transform (such as FFT transform), numerical operation, etc.

[0061] A data association model can also be used to determine the machining location and machining stage of a workpiece in the machining data. The data association model can correspond to a data / signal preprocessing method. For example, the data association model can be used to determine the correlation between the machining data and previously acquired machining data, thereby selecting the machining location and machining stage corresponding to the highest correlation based on the correlation ranking.

[0062] Figure 5 is a schematic diagram of a system according to an embodiment of the present disclosure.

[0063] although Figure 5 The storage device is not shown, but one of ordinary skill in the art will understand that the processor may include one or more storage devices on which instructions and / or data are stored. Figure 5 The processor 520 and the sensor unit 510 are functionally shown as being within a single box, but one of ordinary skill in the art will appreciate that the processor 520 and the sensor unit 510 may actually include multiple processors 520 and multiple sensor units 510, which may or may not be housed within the same physical housing. Therefore, references to the processor 520 and the sensor unit 510 will be understood to include references to a collection of processors 520 and a collection of sensor units 510 operating in parallel or not.

[0064] Processor 520 may be any conventional processing unit, such as a commercial CPU. Alternatively, processor 520 may be a dedicated device, such as an ASIC or other hardware-based control unit.

[0065] The sensor unit 510 may include Figure 5 The various sensors shown in FIG. 5 may be more or less types of sensors. The sensor unit 510 may collect processing data about the machining equipment.

[0066] In one embodiment, the vibration sensor 511 may include but is not limited to mechanical vibration sensors, optical vibration sensors, and electrical vibration sensors (such as inductive vibration sensors, eddy current vibration sensors, capacitive vibration sensors, resistive strain vibration sensors, and piezoelectric vibration sensors).

[0067] In one embodiment, the sound sensor 512 may include, but is not limited to, a piezoelectric ceramic acoustic sensor, a capacitive acoustic sensor, a magnetoelectric acoustic sensor, and the like.

[0068] In one embodiment, the power sensor 513 can obtain at least one of the mechanical power output (e.g., the power of the grinding wheel spindle) and the electrical power received by the machining equipment during machining of the workpiece. Alternatively, at least one of the mechanical power output (e.g., the power of the grinding wheel spindle) and the electrical power received by the machining equipment during machining of the workpiece can be obtained by performing arithmetic operations on sensor data obtained by other sensors.

[0069] In one embodiment, the speed 514 may include, but is not limited to, a photoelectric speed sensor, a magnetoelectric speed sensor, a Hall speed sensor, and the like.

[0070] In one embodiment, the mechanical sensor 515 may include but is not limited to a multi-component force sensor (e.g., a two-component, three-component, four-component, six-component force sensor), a torque sensor (e.g., a dynamic torque sensor and a static torque sensor), and an acceleration sensor (e.g., a single-axis acceleration sensor, a three-axis acceleration sensor).

[0071] In one embodiment, the voltage sensor 516 may include but is not limited to a voltage transformer, a Hall voltage sensor, a fiber optic voltage sensor, etc.

[0072] In one embodiment, the current sensor 517 may include, but is not limited to, a shunt, an electromagnetic current transformer, an electronic current transformer, and the like.

[0073] In one embodiment, the temperature sensor 518 may include, but is not limited to, contact temperature sensors (such as bimetallic thermometers, glass liquid thermometers, pressure thermometers, resistance thermometers, thermistors, and thermocouples) and non-contact temperature sensors (such as various non-contact temperature sensors based on radiation thermometry including luminance method, radiation method, and colorimetry).

[0074] The connection between the processor 520 and the sensor unit 510 may be any connection capable of transmitting at least the machining data of the machining equipment output by the sensor to the processor 520. In one embodiment, the connection includes one or both of a cable connection and a wireless connection.

[0075] The cable connection form of the connection part can include a cable that transmits analog signals (e.g., voltage, 4-20mA current) or digital signals (pulse, CAN, RS485, etc.). The cable form of the connection part is more suitable for applications that require high-performance acquisition and high reliability.

[0076] Wireless connectivity can include various configurations and protocols, including Bluetooth TM ,Bluetooth TM LE short-range communication protocols, sub-GHz, WirelessHART, infrared links, ZigBee, radio frequency identification (RFID), WiFi, the Internet, the World Wide Web, an intranet, a virtual private network, a wide area network, a local area network, a private network using a communication protocol proprietary to one or more companies, Ethernet and HTTP, various cellular communication technologies such as GSM, CDMA, UMTS, EV-DO, WiMAX, LTE or 5th generation "5G" cellular technology and other cellular technologies developed in the future, and various combinations of the foregoing. The connection part in the form of a wireless connection is more suitable for requirements such as ease of installation and small size.

[0077] As reference Figure 1 As described, the processor 520 can obtain processing data of the machining equipment during processing of the workpiece from the sensor unit 510, obtain the type of the workpiece in the processing data, separate the processing data based on the type of the processed workpiece, and select a machine learning model corresponding to the workpiece type of the separated processing data from the anomaly detection model library 530 to process the separated processing data to obtain an anomaly detection result.

[0078] Figure 5 One or more of the processor 520 and the anomaly detection model library 530 can be cloud-based, which may be advantageous when local computing power and storage space are insufficient or communication is smooth. Figure 5One or more of the processor 520 and the anomaly detection model library 530 can be edge-based, which may be advantageous when local computing power and storage space are sufficient or communication is blocked.

[0079] According to an embodiment of the present disclosure, a computer-readable medium is provided, on which computer instructions are stored. When the computer instructions are executed by a processor, the processor executes a method for detecting a fault or anomaly of a mechanical device.

[0080] Examples of storage media for providing program code include floppy disks, hard disks, magneto-optical disks, optical disks (e.g., CD-ROMs, CD-Rs, CD-RWs, DVD-ROMs, DVD-RAMs, DVD-RWs), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer via a communication network.

[0081] The disclosed anomaly detection method, system, and storage medium for machining equipment can provide accurate fault and anomaly detection results when machining different workpieces. The present invention can also detect early-stage failures in mechanical equipment and provide maintenance recommendations to customers, thereby avoiding unplanned failures. By processing machining data using anomaly detection models tailored to workpiece type, machining location, and machining stage, accurate anomaly detection can be performed on machining equipment, thereby improving the yield rate of machined workpieces.

[0082] The text and drawings are provided as examples only to aid understanding of the present disclosure. They should not be construed as limiting the scope of the present disclosure in any way. Although certain embodiments and examples have been provided, it will be apparent to those skilled in the art based on what is disclosed herein that the embodiments and examples shown may be modified without departing from the scope of the present disclosure.

[0083] Although the present disclosure has been described with exemplary embodiments, various changes and modifications may be suggested to one skilled in the art. It is intended that the present disclosure encompass such changes and modifications as fall within the scope of the appended claims.

[0084] Any description in the present invention should not be construed as implying that any particular element, step, or function is essential to be included in the scope of the claims. The scope of the patented subject matter is limited only by the claims.

Claims

1. A method for detecting abnormalities in machining equipment, comprising: Obtain processing data of machining equipment during the processing of workpieces, Get the type of workpiece in the processing data, Separate machining data based on the type of workpiece being machined, A machine learning model corresponding to the workpiece type of the separated processing data is selected to process the separated processing data to obtain an anomaly detection result.

2. The method according to claim 1, wherein The machining data includes at least one of mechanical power output and electrical power received by the machining equipment during machining of a workpiece.

3. The method according to claim 1, wherein The types of workpieces in machining data are obtained by using a scheduling history in a machining device and the machining data are separated.

4. The method according to claim 1, wherein The processing data of the machining equipment during the machining of the workpiece includes: Obtain full-time processing data including processing data during processing period and processing data during stop period, removing the stop period processing data from the full period processing data to obtain the processing period processing data; and The selecting of a machine learning model corresponding to the workpiece type of the separated processing data to process the separated processing data includes: processing the processing period processing data in the separated processing data.

5. The method according to claim 4, wherein Remove the stopped period processing data from the full period processing data through machine learning classification models or based on thresholds.

6. The method according to claim 4, wherein: Obtaining processing data of the machining equipment during the processing of the workpiece also includes: Removing the impact pulse data of the machining equipment at startup and shutdown from the machining data during the machining period to obtain the effective machining data during the machining period; and The selecting of a machine learning model corresponding to the workpiece type of the separated processing data to process the separated processing data includes: processing the effective processing period processing data in the separated processing data.

7. The method according to claim 1, further comprising: Obtain the processing part of the workpiece to be processed, Separate the processing data based on the type and processing part of the workpiece to be processed. A machine learning model corresponding to the workpiece type and processing position of the separated processing data is selected to process the separated processing data to obtain an abnormality detection result.

8. The method according to claim 7, further comprising: Obtain the machining stage of the machined workpiece, Separate processing data based on the type of workpiece being processed, processing location and processing stage, A machine learning model corresponding to the workpiece type, processing position, and processing stage of the separated processing data is selected to process the separated processing data to obtain an anomaly detection result.

9. The method according to claim 7 or 8, wherein Based on the mechanical processing principle or machine learning classification model of the workpiece, the processing position and processing stage of the workpiece in the processing data are obtained.

10. An abnormality detection system for machining equipment, comprising: a sensor unit including one or more sensors configured to collect processing data about the machining equipment; an anomaly detection model library, comprising one or more anomaly detection models configured to perform anomaly detection on machining equipment; and The processor is configured to: Obtain processing data of the machining equipment during processing of the workpiece from the sensor unit, Get the type of workpiece in the processing data, Separate machining data based on the type of workpiece being machined, A machine learning model corresponding to the workpiece type of the separated processing data is selected from the anomaly detection model library to process the separated processing data to obtain an anomaly detection result.

Images