Equipment abnormity monitoring method and electronic equipment

Abstract

The invention provides an equipment abnormity monitoring method and electronic equipment, and the method comprises the steps: obtaining the processing data of processing equipment, generating a first data line diagram corresponding to the processing data, and enabling the first data line diagram to comprise an initial change curve of the processing data along with the time change; performing time sequence prediction on the initial change curve according to the time sequence characteristics of the initial change curve to obtain prediction data; the first data line graph is updated based on the prediction data, a second data line graph is obtained, and the second data line graph comprises an updated initial change curve; determining an abnormal score of the processing equipment based on a distribution position and a change parameter of the updated initial change curve; and determining an abnormity monitoring result of the processing equipment according to the abnormity score. According to the invention, the equipment abnormity monitoring efficiency and accuracy can be improved.

Description

Technical Field

[0001] This application relates to the field of anomaly monitoring technology in the manufacturing industry, specifically to a method and electronic device for monitoring equipment anomalies. Background Technology

[0002] In the context of intelligent manufacturing upgrades, the processing of workpieces by machining equipment faces a prominent contradiction between the explosive growth of processing data and the mismatch between quality control capabilities. Related equipment anomaly monitoring methods typically compare processing data with anomaly thresholds after processing to obtain the anomaly monitoring results. This monitoring method can lead to delayed anomaly response and production capacity losses. Summary of the Invention

[0003] In view of the above, it is necessary to propose a method and electronic device for monitoring equipment anomalies, which can solve the technical problem that the inability of related technologies to predict the processing data of equipment leads to the inability to respond to equipment anomalies in a timely manner, resulting in reduced equipment capacity.

[0004] A first aspect of this application provides a method for monitoring equipment anomalies, comprising: acquiring processing data of a processing equipment; generating a first data line graph corresponding to the processing data, the first data line graph including an initial change curve obtained based on the processing data; performing time-series prediction on the initial change curve to obtain predicted data; updating the first data line graph based on the predicted data to obtain a second data line graph; performing anomaly assessment on the processing equipment based on the second data line graph to obtain an anomaly score for the processing equipment; and determining the anomaly monitoring result of the processing equipment based on the anomaly score.

[0005] According to an embodiment of this application, the processing data includes the workpiece size and processing time of the same type of workpiece processed by the processing equipment.

[0006] According to an embodiment of this application, the method further includes: preprocessing the processing data, the preprocessing including: determining abnormal data in the processing data based on an anomaly detection algorithm, cleaning the abnormal data, wherein the abnormal data includes outliers and noise data; sorting the data-cleaned processing data according to the processing time to obtain sorted processing data; and grouping the sorted processing data according to a division rule that the processing data corresponding to a preset number of workpieces are grouped together to obtain multiple data groups.

[0007] According to an embodiment of this application, generating the first data line graph corresponding to the processing data includes: determining the change curve of the workpiece size with the processing time based on the sorted processing data to obtain the initial change curve; determining a first size mean, a first control value, and a second control value based on the workpiece size in each of the plurality of data groups; determining the center line of the first data line graph based on the first size mean; determining a first upper control line based on the first control value; and determining a first lower control line based on the second control value; and generating the first data line graph based on the initial change curve, the center line, the first upper control line, and the first lower control line.

[0008] According to an embodiment of this application, the step of determining a first average size, a first control value, and a second control value based on the workpiece dimensions in each of the plurality of data groups, determining the center line of the first data graph based on the first average size, determining a first upper control line based on the first control value, and determining a first lower control line based on the second control value includes: determining the average value of all workpiece dimensions in the plurality of data groups to obtain the first average size; determining the center line based on the first average size; determining the difference between the maximum and minimum workpiece dimensions in each data group to obtain the size range; determining the average of the size ranges of the plurality of data groups based on the size ranges of each data group to obtain the range average; determining the first control value based on the sum of the first average size and the range average, determining the second control value based on the difference between the first average size and the range average; determining the first upper control line based on the first control value, and determining the first lower control line based on the second control value.

[0009] According to an embodiment of this application, the step of performing time-series prediction on the initial change curve based on its temporal features to obtain predicted data includes: obtaining original distribution data based on the initial change curve; performing data augmentation on the original distribution data using the data augmentation module of the time-series prediction model to obtain augmented data; extracting features from the augmented data using the encoder of the time-series prediction model to obtain multi-scale temporal features; determining the weight of the temporal features at each scale based on the self-attention mechanism of the time-series prediction model; fusing the multi-scale temporal features according to the weights to obtain fused features; and decoding the fused features using the decoder of the time-series prediction model to obtain the predicted data.

[0010] According to an embodiment of this application, the data augmentation module using a time-series prediction model performs data augmentation on the original distributed data to obtain augmented data, including: normalizing the original distributed data to obtain normalized distributed data; performing data augmentation on the normalized distributed data to obtain synthetic data, wherein the data augmentation objective includes minimizing the distribution difference between the synthetic data and the normalized distributed data; and obtaining the augmented data based on the set of the normalized distributed data and the synthetic data.

[0011] According to an embodiment of this application, updating the first data line graph based on the predicted data to obtain the second data line graph includes: updating the initial change curve based on the prediction curve corresponding to the predicted data, and updating the time range corresponding to the first data line graph; obtaining the second data line graph based on the updated initial change curve and the updated time range.

[0012] According to an embodiment of this application, the second data line graph includes an updated initial change curve, a center line, a first upper control line, and a first lower control line. Determining the anomaly score of the processing equipment based on the distribution position and change parameters of the updated initial change curve includes: determining the distribution position and change parameters of each data point in the updated initial change curve based on the updated initial change curve, the center line, the first upper control line, and the first lower control line; performing multi-dimensional anomaly evaluation on the data points of the updated initial change curve based on the distribution position and change parameters to obtain an anomaly score for each dimension; and determining the anomaly score of the processing equipment based on the anomaly scores of all dimensions.

[0013] According to an embodiment of this application, if the processing equipment includes multiple processing devices, determining the abnormal score of the processing device based on the abnormal scores of all dimensions includes: determining the extreme significance level value of each processing device among the multiple processing devices based on the abnormal scores of all processing devices; normalizing the abnormal score of each processing device in each dimension to obtain the normalized abnormal score of each processing device in each dimension; determining the weight of each dimension based on the extreme significance level value and the normalized abnormal score; and determining the weighted sum of all normalized abnormal scores corresponding to all dimensions based on the weight of each dimension and the normalized abnormal score of each processing device in each dimension to obtain the abnormal score of each processing device.

[0014] According to an embodiment of this application, determining the abnormal monitoring result of the processing equipment based on the abnormal score includes: if the abnormal score is greater than a preset score threshold, determining that the abnormal monitoring result indicates that the processing equipment has an abnormality; or, if the abnormal score is less than or equal to the preset score threshold, determining that the abnormal monitoring result indicates that the processing equipment does not have an abnormality.

[0015] According to an embodiment of this application, the method further includes: constructing a knowledge graph based on the processing data of the processing equipment and the anomaly monitoring results, wherein the knowledge graph includes the processing data of the processing equipment with anomalies, the anomaly score of the processing equipment with anomalies, the anomaly cause of the processing equipment with anomalies, and the anomaly solution of the processing equipment with anomalies.

[0016] A second aspect of this application provides an equipment anomaly monitoring device, comprising: a generation module for acquiring processing data from a processing equipment and generating a first data line graph corresponding to the processing data, the first data line graph including an initial change curve of the processing data over time; a prediction module for performing time-series prediction on the initial change curve based on the time-series characteristics of the initial change curve to obtain predicted data; an update module for updating the first data line graph based on the predicted data to obtain a second data line graph, the second data line graph including the updated initial change curve; a scoring module for determining an anomaly score of the processing equipment based on the distribution position and change parameters of the updated initial change curve; and a determination module for determining the anomaly monitoring result of the processing equipment based on the anomaly score.

[0017] A third aspect of this application provides an electronic device, including: a memory and a processor, wherein the processor executes computer-readable instructions stored in the memory to implement the device anomaly monitoring method.

[0018] The equipment anomaly monitoring method provided in this application generates an initial data line graph by acquiring processing data from the processing equipment. The initial data line graph includes an initial change curve of the processing data over time. The initial change curve is predicted using a time series prediction model, and the data line graph is updated. Based on the updated data line graph, the anomaly score of the processing equipment is determined. The anomaly monitoring result of the equipment is determined according to the score, thereby realizing real-time monitoring and anomaly warning of the operating status of the processing equipment and improving the timeliness and accuracy of equipment maintenance. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the device anomaly monitoring method provided in an embodiment of this application.

[0021] Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0022] Figure 3 This is a flowchart illustrating a method for generating a first data line graph provided in an embodiment of this application.

[0023] Figure 4 An example diagram of the first data line diagram provided for an embodiment of this application.

[0024] Figure 5 This is a flowchart illustrating the method for time series prediction based on an initial change curve provided in an embodiment of this application.

[0025] Figure 6 This is a flowchart illustrating a method for anomaly assessment based on a second data line graph, provided in an embodiment of this application.

[0026] Figure 7 Example diagrams illustrating multiple dimensions of anomaly assessment provided in embodiments of this application.

[0027] Figure 8 This is a flowchart illustrating a method for determining the anomaly score of a processing device, as provided in an embodiment of this application.

[0028] Figure 9 This is a schematic diagram of a device for monitoring equipment malfunctions, provided in an embodiment of this application. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this application clearer, the application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] It should be noted that in this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and drawings of this application are used to distinguish similar objects, not to describe a specific order or sequence.

[0031] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner. Unless otherwise specified, the following embodiments and features described herein can be combined with each other.

[0032] Against the backdrop of continuous upgrades in intelligent manufacturing, modern manufacturing industries are placing higher demands on the intelligence level of processing equipment. Equipment not only needs to achieve high-precision automated processing, but also needs to possess real-time status perception, dynamic monitoring, and intelligent decision-making capabilities to support end-to-end quality control and production efficiency optimization.

[0033] However, the manufacturing process of precision electronic components such as mobile phone casings still faces many challenges in the CNC machining stage. For example, the process chain involves multi-process collaboration and multi-parameter coupling, making it difficult to efficiently integrate and analyze the massive amounts of heterogeneous data generated during processing. Traditional quality inspection methods rely heavily on manual experience and offline sampling, and periodic inspections using measuring instruments and other tools have significant lags, failing to capture processing quality drift in real time. Existing anomaly handling mechanisms are mostly passive threshold alarms, lacking in-depth mining of historical data and trend prediction, making it difficult to identify gradual failures in advance. The debugging cycle required after anomalies occur is long, directly affecting production continuity and overall equipment efficiency.

[0034] To address the aforementioned issues, the equipment anomaly monitoring method provided in this application generates an initial data line graph by acquiring processing data from the processing equipment. The initial data line graph includes an initial change curve of the processing data over time. A time-series prediction model is used to predict the initial change curve and update the data line graph. An anomaly score for the processing equipment is determined based on the updated data line graph. The anomaly monitoring result is determined based on the score, thereby achieving real-time monitoring and anomaly warning of the processing equipment's operating status and improving the timeliness and accuracy of equipment maintenance.

[0035] Please see Figure 1 The diagram shown is a flowchart illustrating the device anomaly monitoring method provided in this application embodiment. The device anomaly monitoring method of this application embodiment is applied to electronic devices (such as...) Figure 2 The electronic device 10 shown includes the following steps. Depending on different needs, the order of the steps in the flowchart can be changed, and some steps can be omitted.

[0036] Step S101: Obtain processing data from the processing equipment and generate a first data line graph corresponding to the processing data. The first data line graph includes the initial change curve of the processing data over time.

[0037] In some embodiments of this application, the processing equipment includes at least one device, such as a device for processing precision electronic components such as mobile phone casings, and the workpiece may include, but is not limited to, precision electronic components such as mobile phone casings.

[0038] In one example, the processing data includes, but is not limited to, the workpiece number, workpiece size, and processing time of the same type of workpiece processed by the processing equipment. Each workpiece has a unique number to distinguish it from other workpieces. The processing equipment measures the size of each workpiece as the processing dimension, and the processing time includes a timestamp generated by the processing equipment's high-precision clock module for each workpiece.

[0039] For example, the processing equipment processed workpiece number Km0091 at 14:24:06 on June 25, 2025, and the measured processing dimension of workpiece Km0091 was 17mm. In one example, the size range of the processing dimension can be defined according to the size specification of the workpiece. For example, if the size specification is [Min, Max], the size range can be [0.9Min, 1.1Max], where Min represents the minimum size specification and Max represents the maximum size specification. In one example, processing data within a preset time length can be obtained according to actual needs. For example, the preset time length can be 1 hour (1h), and the processing time range can be expressed as [T, T+1h], where T represents the time when data acquisition begins.

[0040] In one example, subsequent embodiments of this application use workpiece dimensions as an example to illustrate a method for monitoring anomalies in processing equipment based on workpiece dimensions. In other examples, the workpiece dimensions can be replaced with any parameter such as pressure, temperature, or rotational speed to achieve anomaly monitoring of the processing equipment.

[0041] By acquiring processing data from the processing equipment, a data foundation can be provided for subsequent anomaly monitoring of the processing equipment.

[0042] In one example, equipment information for the processing equipment can also be obtained, such as equipment parameters (e.g., pressure, temperature, speed, tool usage time and status), equipment location (e.g., building and floor where the equipment is located), and personnel associated with the equipment (e.g., operator identification). Furthermore, equipment information can also include the number of historical anomalies and their duration, such as the number of NG (not good) errors and their duration.

[0043] By acquiring equipment data from processing equipment, it is possible to locate the equipment and the corresponding personnel more quickly, thereby improving the efficiency of subsequent equipment maintenance.

[0044] In some embodiments of this application, the method further includes: determining abnormal data in the processing data based on an anomaly detection algorithm, cleaning the abnormal data, wherein the abnormal data includes outliers and noise data; sorting the cleaned processing data according to the processing time to obtain sorted processing data; and grouping the sorted processing data according to a division rule that the processing data corresponding to a preset number of workpieces are grouped together to obtain multiple data groups.

[0045] In one example, anomaly detection algorithms, such as statistical distributions or machine learning models, can be used to identify anomalous data in the processed data. Anomalous data includes, but is not limited to, outliers and noisy data. Outliers and noisy data can be removed to achieve data cleaning. In another example, anomalous data also includes missing values. Interpolation algorithms can be used to impute missing values ​​to ensure data integrity.

[0046] In one example, the timestamp generated by the clock module built into the processing equipment is used as a reference to perform time alignment and sorting on the cleaned processing data, thereby forming a processing data sequence with clear temporal logic.

[0047] In one example, based on production cycle time and quality control requirements, and according to a predefined rule that the processing data corresponding to a preset number of workpieces is grouped together, the sliding window method can be used to divide the sorted processing data into multiple data groups. For example, the sorted processing data may contain processing data for m workpieces, such as m processing dimensions and m processing times. The processing data for the m workpieces can be divided according to a preset quantity k. The data is processed in groups, where m represents an integer and k represents an integer greater than 1, such as 3, 5, 10, etc. This represents the floor function. Thus, each data set contains processing data for k workpieces, such as k processing dimensions and k processing times.

[0048] Based on the above embodiments, data quality is optimized through data cleaning. By performing time alignment and sorting on the cleaned data, time axis offsets caused by differences in sampling frequencies from multiple sensors can be effectively eliminated, constructing a data sequence with clear temporal logic. By dynamically dividing the sorted data into multiple data groups, structured data units can be formed. This provides highly available input data for subsequent anomaly monitoring, significantly improving the accuracy and efficiency of subsequent data analysis.

[0049] In some embodiments of this application, when generating the first data line graph corresponding to the processing data, an initial change curve of the workpiece size with processing time can be determined based on the sorted processing data. A first average size, a first control value, and a second control value are determined based on the workpiece sizes in multiple data groups, thereby determining the center line, the first upper control line, and the first lower control line of the first data line graph; the first data line graph is generated based on the aforementioned initial change curve and the three control lines.

[0050] In one example, the method for generating the first data line chart can be found in [reference needed]. Figure 3 The process is shown below. For example... Figure 4 The diagram shown is an example of the first data line diagram provided in an embodiment of this application. The center line, as shown by the straight line in the diagram, corresponds to the average first dimension of the workpiece. In production practice, the center line serves as a core benchmark for quality monitoring, reflecting the dimensional stability of the processing in real time. This provides a scientific basis for anomaly monitoring, early warning, and process adjustment, contributing to improved product quality and production efficiency.

[0051] For example Figure 4 As shown in the figure, the first upper control line is a straight line, and the corresponding workpiece size is the first control value. The first lower control line is a straight line, and the corresponding workpiece size is the second control value. The first upper and first lower control lines, determined through statistical calculations, ensure the objectivity and accuracy of the control lines, effectively avoiding the subjectivity of manually setting control limits. The first upper and first lower control lines can reflect the stability of the processing in real time, providing a scientific basis for anomaly monitoring, early warning, and process adjustment, thus helping to improve product quality and production efficiency.

[0052] In some embodiments of this application, the multiple control lines of the first data line graph further include: specification control lines for workpiece dimensions defined according to the workpiece's size specifications. For example, the workpiece's size specifications include a maximum size specification and a minimum size specification. The specification control lines include a second upper control line and a second lower control line, where the workpiece size value corresponding to the second upper control line is the maximum size specification, and the workpiece size value corresponding to the second lower control line is the minimum size specification.

[0053] In one example, a second upper control line can be determined based on the maximum size specification. This second upper control line is set parallel to the horizontal axis of the first data line graph, and the workpiece size corresponding to the second upper control line in the first data line graph is the maximum size specification. For example... Figure 4 As shown, the second upper control line is a straight line in the figure, and the corresponding workpiece size is the maximum size specification.

[0054] In one example, a second lower control line can be determined based on the minimum size specification. This second lower control line is set parallel to the horizontal axis of the first data line graph, and the workpiece size corresponding to the second lower control line in the first data line graph is the minimum size specification. For example... Figure 4 As shown, the second lower control line is a straight line in the figure, and the corresponding workpiece size is the minimum size specification.

[0055] Based on the above embodiments, the hard boundary for anomaly monitoring can be determined based on the second upper control line and the second lower control line set in the specifications. When the processing size exceeds the specification control line, it can be directly determined that there is an anomaly, thereby triggering a clear quality warning.

[0056] Based on the above embodiments, the first data line graph includes a continuous curve showing the change of processing dimensions over time, as well as a first upper control line and a first lower control line derived through statistical calculations, and a second upper control line and a second lower control line defined based on workpiece size specifications. These control lines enable a dual quality monitoring mechanism, improving the comprehensiveness of quality monitoring and enhancing the timeliness of quality early warnings. For example, it can simultaneously monitor process stability and product compliance, providing dual basis for quality decisions. In actual production, this design helps to promptly detect abnormal fluctuations in the processing process, preventing the generation of defective products, thereby ensuring the stability and consistency of product quality.

[0057] Step S102: Perform time-series prediction on the initial change curve based on its temporal characteristics to obtain prediction data.

[0058] In some embodiments of this application, the amount of data corresponding to the initial change curve is relatively small, and because it is historical data, it may be difficult to accurately predict future anomalies based on the initial change curve. The temporal features of the initial change curve can be extracted, and temporal prediction can be performed on the initial change curve based on these features to obtain predicted data. This achieves data augmentation of the data corresponding to the initial change curve, thereby improving the accuracy of anomaly monitoring and early warning of no anomalies.

[0059] In some embodiments of this application, the time-series prediction model may include a Time-series Generative Adversarial Networks (TimeGAN) architecture, an attention mechanism, and an Attention-LSTM mechanism of a Long Short-Term Memory (LSTM) network. The time-series prediction model can be used for optimization with small sample data to achieve time-series prediction of initial change curves.

[0060] In some embodiments of this application, raw distribution data can be obtained based on the initial change curve. This data can then be augmented using the data augmentation module of a time-series prediction model to obtain augmented data. The encoder of the model then extracts features from the augmented data to obtain multi-scale time-series features. Next, the weights of the time-series features at each scale are determined based on the model's self-attention mechanism, and the multi-scale time-series features are fused to obtain fused features. Finally, the decoder of the model decodes the fused features to obtain the predicted data. In one example, the method for time-series prediction based on the initial change curve can refer to... Figure 5 The flowchart shown.

[0061] Based on the above embodiments, the initial change curve is first augmented with data, then multi-scale temporal features are extracted by the encoder, and the weights of each scale feature are reasonably allocated and fused with the help of the self-attention mechanism. Finally, the predicted data is obtained by the decoder. This process can fully explore the temporal information in the data, effectively capture the inherent laws and trends of data changes, help improve the accuracy and reliability of prediction, and provide more valuable data support for subsequent decision-making.

[0062] Step S103: Update the first data line graph based on the predicted data to obtain the second data line graph, which includes the updated initial change curve.

[0063] In some embodiments of this application, updating a first data line graph based on predicted data to obtain a second data line graph includes: updating an initial change curve based on the prediction curve corresponding to the predicted data, and updating the time range corresponding to the first data line graph; and obtaining the second data line graph based on the updated initial change curve and the updated time range.

[0064] In one example, a corresponding prediction curve can be plotted based on the measured data, and this prediction curve can be plotted after the initial change curve, thus achieving an extension effect on the initial change curve. Correspondingly, the time range corresponding to the first data line graph is updated according to the time range corresponding to the prediction curve. For example, if the time range corresponding to the first data line graph is 50 time points, and the time range corresponding to the prediction curve is 100 time points, the updated time range corresponding to the first data line graph can be a total of 150 time points.

[0065] In one example, the second data graph includes an updated initial change curve, a center line, a first upper control line, and a first lower control line, as well as updated second upper control lines and second lower control lines. The updating of the center line, first upper control line, first lower control line, second upper control line, and second lower control line can be a backward extension along the horizontal axis, and the corresponding workpiece dimension value remains unchanged after the update.

[0066] Based on the above embodiments, by updating the first data line chart based on predicted data, intelligent extension and time dimension expansion of the initial change curve are achieved. Specifically, based on the predicted data generated by the time-series prediction model, a predicted curve is appended to the end of the initial change curve, forming a continuous time-series extension effect, dynamically expanding the data visualization range from the original time point range to a larger range. This incremental update strategy not only maintains the integrity of historical data, but the updated second data line chart can also intuitively display the trend evolution over a longer time span, providing continuous decision-making basis for scenarios such as equipment anomaly prediction, and is particularly suitable for long-term trend analysis in short-cycle data scenarios.

[0067] Step S104: Based on the distribution location and change parameters of the updated initial change curve, determine the anomaly score of the processing equipment.

[0068] In some embodiments of this application, the distribution position and variation parameters of each data point in the updated initial change curve can be determined based on control lines such as the center line, the first upper control line, and the first lower control line in the second data line graph. A multi-dimensional anomaly assessment is performed on the data points based on their distribution position and variation parameters to obtain an anomaly score for each dimension. The anomaly score of the processing equipment is then determined based on the anomaly scores of all dimensions. In one example, the method for anomaly assessment based on the second data line graph can also refer to... Figure 6 The flowchart shown.

[0069] In one example, the distribution of data points can be relative to the positions of the center line, the first upper control line, and the first lower control line. For example, it could be located above the first upper control line. In another example, the distribution of data points can also be relative to the positions of the second upper control line and the second lower control line. For example, it could be located above the second upper control line.

[0070] In one example, the variation parameter of the data points can be the trend of workpiece size change corresponding to multiple consecutive data points, such as the workpiece size increasing or decreasing for s2 consecutive (e.g., 6) data points.

[0071] In some embodiments of this application, different distribution positions and variation parameters can be set as corresponding dimensions individually or in combination. For example, multiple dimensions may include, but are not limited to: data points located above or below the second upper control line, data points located above or below the first upper control line, s1 consecutive (e.g., 15) data points being outside a preset distance range from the center line, and s2 consecutive (e.g., 6) data points corresponding to workpiece dimensions increasing or decreasing. The values ​​of s1, s2, etc., can be set according to actual needs, and this application does not impose specific limitations on them.

[0072] In one example, if the updated initial change curve satisfies a condition in any dimension, it has an anomaly score for that dimension. For example... Figure 7 As shown, if the updated initial change curve is above the second upper control line, it has an anomaly score of 10 in that dimension.

[0073] Based on the above embodiments, key information from the second data line chart can be comprehensively utilized to evaluate the anomalies of the updated initial change curve data points from multiple dimensions. This allows for a comprehensive and detailed analysis of the relative relationships between data points and control lines and center lines, accurately capturing abnormal fluctuations in the data. This multi-dimensional analysis method effectively avoids the limitations of single-dimensional evaluation, making the final determined anomaly score for the processing equipment more accurate and reliable. It provides strong support for timely detection of potential problems in processing equipment and ensuring production quality and efficiency.

[0074] Step S105: Determine the abnormal monitoring results of the processing equipment based on the abnormality score.

[0075] In some embodiments of this application, determining the anomaly monitoring result of the processing equipment based on the anomaly score includes: if the anomaly score is greater than a preset score threshold, determining that the anomaly monitoring result indicates that the processing equipment has an anomaly; or, if the anomaly score is less than or equal to the preset score threshold, determining that the anomaly monitoring result indicates that the processing equipment does not have an anomaly. The preset score threshold can be set according to actual needs, and this application does not impose specific limitations on it.

[0076] In some embodiments of this application, after determining that the abnormal monitoring results indicate that the processing equipment is abnormal, various early warning methods can be used to issue an early warning. For example, early warnings can be issued through voice broadcasts, sound alerts, or by sending abnormal information to the personnel corresponding to the equipment.

[0077] In some embodiments of this application, a knowledge graph can be constructed based on the processing data of the processing equipment and the anomaly monitoring results. The knowledge graph includes the processing data of the processing equipment with anomalies, the anomaly score of the processing equipment with anomalies, the anomaly cause of the processing equipment with anomalies, and the anomaly solution of the processing equipment with anomalies.

[0078] In one example, equipment information for abnormal processing equipment (such as model, serial number, location, pressure, temperature, etc.) can be obtained, and abnormal alarm data and rules can be recorded, such as the time point and abnormal value of the abnormal value. The cause of the abnormality is investigated based on expert and field experience, and the cause is recorded accordingly. The abnormal behavior of the machine is defined according to preset rules, and the relationship between the abnormal behavior and its cause is analyzed. A knowledge graph containing the above content is constructed using the lightweight Python spectrogram analysis library NetworkX. Anomaly diagnosis reasoning and failure analysis are performed based on the knowledge graph, and dynamic updates and reinforcement learning are executed on the knowledge graph.

[0079] Based on the above embodiments, a knowledge graph can be constructed using Python's networkX library by integrating data such as abnormal equipment information, abnormal scores, abnormal causes, and solutions, based on processing equipment data and abnormal monitoring results. This enables abnormal diagnosis reasoning and failure analysis, and the knowledge graph can be continuously optimized through dynamic updates and reinforcement learning, providing efficient references for subsequent abnormal early warning and handling, and improving the efficiency of abnormal response.

[0080] The equipment anomaly monitoring method provided in this application generates an initial data line graph and change curve by acquiring processing data of the processing equipment, predicts the curve using a time series prediction model and updates the data line graph, performs anomaly assessment on the processing equipment based on the updated data line graph to generate anomaly score, and determines the anomaly monitoring result of the equipment according to the score. This method can realize real-time monitoring and anomaly early warning of the operating status of the processing equipment, and improve the timeliness and accuracy of equipment maintenance.

[0081] In one example, reference Figure 3 As shown, the method for generating the first data line chart may include the following process.

[0082] Step S301: Based on the sorted processing data, determine the change curve of workpiece size with processing time to obtain the initial change curve.

[0083] In some embodiments of this application, curve fitting algorithms can be used to fit the sorted processing data, using the continuous curve of workpiece size changing with processing time as the initial curve. (See reference...) Figure 4 The diagram shown is an example of a first data line graph provided in an embodiment of this application. The first data line graph has a first dimension and a second dimension; for example, the first dimension is the processing time shown on the horizontal axis, and the second dimension is the processing dimension shown on the vertical axis. The initial change curve, as shown in the figure, represents the change curve of the workpiece dimension with processing time.

[0084] Step S302: Determine the first average size based on the workpiece size in multiple data groups, and determine the first control value and the second control value based on the workpiece size in each of the multiple data groups.

[0085] In some embodiments of this application, the average value of all workpiece dimensions in multiple data sets can be determined to obtain a first dimension average. In one example, the formula used to determine the first dimension average can be exemplarily expressed as: , in, This represents the average of the first dimension. Indicates the first The first in the group processing data The workpiece size is denoted by m, where m represents the number of workpieces corresponding to all processing data.

[0086] Based on the above embodiments, the first direct calculation method for the mean of the first size is suitable for situations where the amount of data is small or computing resources are sufficient, and can directly reflect the average level of the overall data.

[0087] In another example, determining the first dimension mean may further include: determining the average value of the workpiece dimensions for each data group based on the workpiece dimensions in each data group, obtaining a second dimension mean, and determining the first dimension mean based on the second dimension mean. For example, the formula used to determine the second dimension mean can be exemplarily expressed as: , in, Indicates the first The second dimension mean of the group processing data. Indicates the first The first in the group processing data The size of the workpiece, k represents the preset quantity, for example, the size of the first workpiece. The number of workpieces corresponding to the group processing data. The formula used to determine the average of the first dimension based on the average of the second dimension can be expressed as follows: , in, This represents the average of the first dimension. Indicates the first The second dimension mean of the group processing data. Indicates shared ownership Group processing data.

[0088] Based on the above embodiments, the second grouping calculation method for the first size mean, by first grouping and then summarizing, can better balance the differences within and between groups while ensuring calculation efficiency, and is suitable for scenarios with large amounts of data or those requiring hierarchical analysis.

[0089] In some embodiments of this application, determining the first control value and the second control value includes: determining the difference between the maximum and minimum dimensions in each data group to obtain a size range; determining the average of the size ranges of multiple data groups based on the size ranges of each data group to obtain a mean range; and determining the first control value and the second control value based on the mean size and the mean range, wherein the first control value is greater than the second control value.

[0090] In one example, the formula used to determine the dimensional range can be exemplarily expressed as: =max{ }-min{ }, in, Indicates the first Dimensional range of the group processing data, { } indicates the first In the group processing data Each processing dimension, max{ } indicates the first The maximum dimension in the group processing data, min{ } indicates the first The smallest dimension in the group processing data.

[0091] In one example, the formula used to determine the mean of the range can be represented as follows: = , in, This represents the mean of the range. Indicates the first Dimensional range of the group processing data Indicates shared ownership Group processing data.

[0092] In one example, the formula used to determine the first control value and the second control value can be represented as follows: UCL = , LCL = , in, This represents the average of the first dimension. The range represents the mean, UCL represents the first control value, and LCL represents the second control value. Indicates the sample size per group The relevant statistical constants can be obtained by querying the preset standard coefficient table.

[0093] Step S303: Determine the center line of the first data line graph based on the first size average value, determine the first upper control line based on the first control value, and determine the first lower control line based on the second control value.

[0094] In one example, when determining the centerline based on the first average dimension, the centerline can be set to be parallel to the horizontal axis of the first data line graph, and the workpiece dimension corresponding to the centerline in the first data line graph is the first average dimension. For example... Figure 4 As shown in the figure, the centerline, represented by the straight line, corresponds to the average first dimension of the workpiece. In production practice, the centerline serves as a core benchmark for quality monitoring, reflecting the dimensional stability of the machining process in real time. This provides a scientific basis for anomaly monitoring, early warning, and process adjustment, ultimately contributing to improved product quality and production efficiency.

[0095] In one example, when determining the first upper control line based on the first control value, the first upper control line can be set to be parallel to the horizontal axis of the first data line graph, and the workpiece dimension corresponding to the first upper control line in the first data line graph is the first control value. For example... Figure 4 As shown in the figure, the first upper control line is a straight line, and the corresponding workpiece size is the first control value.

[0096] In one example, when determining the first lower control line based on the second control value, the first lower control line can be set to be parallel to the horizontal axis of the first data line graph, and the workpiece dimension corresponding to the first lower control line in the first data line graph is the second control value. For example... Figure 4 As shown in the figure, the first lower control line is a straight line, and the corresponding workpiece size is the second control value.

[0097] Based on the above embodiments, the first upper control line and the first lower control line determined through statistical calculation can ensure the objectivity and accuracy of the control lines, effectively avoiding the subjectivity of manually setting control limits. The first upper control line and the first lower control line can reflect the stability status of the processing process in real time, providing a scientific basis for anomaly monitoring, early warning, and process adjustment, which helps to improve product quality and production efficiency.

[0098] Step S304: Generate a first data line graph based on the initial change curve, center line, first upper control line and first lower control line.

[0099] In some embodiments of this application, the initial change curve, center line, first upper control line, and first lower control line can be plotted in the same image to obtain a first data line graph. For example, Figure 4 As shown, the horizontal axis of the first data line graph corresponds to the processing time, and the vertical axis corresponds to the processing dimensions.

[0100] In some embodiments of this application, the second upper control line and the second lower control line can also be drawn in the first data line diagram, for example... Figure 4 As shown.

[0101] In some embodiments of this application, multiple dividing lines may be drawn in the first data line graph. These dividing lines indicate the division of the area between the first control line and the second control line. For example, dividing the area between the first control line and the second control line into 6 equal parts can determine sigma = (first control value - second control value) / 6, and the dividing lines can be set at positions ±2 sigma and ±1 sigma of the center line. Figure 4 As shown, the dividing lines are represented by the dashed lines in the figure.

[0102] Based on the above embodiments, the first data line graph includes a continuous curve showing the change of processing dimensions over time, as well as a first upper control line and a first lower control line derived through statistical calculations, and a second upper control line and a second lower control line defined based on workpiece size specifications. These control lines enable a dual quality monitoring mechanism, improving the comprehensiveness of quality monitoring and enhancing the timeliness of quality early warnings. For example, it can simultaneously monitor process stability and product compliance, providing dual basis for quality decisions. In actual production, this design helps to promptly detect abnormal fluctuations in the processing process, preventing the generation of defective products, thereby ensuring the stability and consistency of product quality.

[0103] In one example, methods for time series forecasting based on initial change curves include, for example: Figure 5 The process is shown below.

[0104] Step S501: Obtain the original distribution data based on the initial change curve, and use the data augmentation module of the time series prediction model to augment the original distribution data to obtain augmented data.

[0105] In some embodiments of this application, the original distribution data of the initial change curve may include, but is not limited to, one or more of the following characteristics: slope characteristics, curvature characteristics, peak characteristics, and fluctuation characteristics of the initial change curve.

[0106] In one example, the least squares method can be used to fit the overall trend line of the initial change curve, and the slope sequence for each time period can be calculated to obtain the slope characteristics. In another example, a cubic spline interpolation algorithm can be used to calculate the curvature sequence of the initial change curve based on its first and second numerical derivatives, thus obtaining the curvature characteristics. In yet another example, a continuous wavelet transform algorithm can be used to identify the set of local maxima of the initial change curve, and the significance index of the initial change curve can be calculated based on the mean and standard deviation peak of its peak neighborhood, thus obtaining the peak characteristics. Finally, a sliding window algorithm can be used to divide the initial change curve into equal-length time windows, and the fluctuation amplitude within each window can be calculated to obtain the fluctuation characteristics.

[0107] In one example, a multidimensional feature matrix can be constructed based on the original distribution data described above. For instance, the multidimensional feature matrix can be represented as follows: D raw =[ S T , K T , P max T , A T ].in, S T Represents the transpose matrix of the slope sequence. K T . represents the transpose matrix of the curvature sequence. P max T Represents the peak characteristic matrix, A T This represents the transpose of the fluctuation amplitude sequence.

[0108] In some embodiments of this application, the data augmentation module of a time-series prediction model is used to augment the original distributed data to obtain augmented data. This includes: normalizing the original distributed data to obtain normalized distributed data; augmenting the normalized distributed data to obtain synthetic data, wherein the data augmentation objective includes minimizing the distribution difference between the synthetic data and the normalized distributed data; and obtaining the augmented data based on the set of the normalized distributed data and the synthetic data.

[0109] In one example, normalization can include, but is not limited to, min-max normalization and Z-score normalization. For instance, min-max normalization can be performed on slope features and curvature features, while Z-score normalization can be applied to peak amplitudes. Normalization makes different data points in a normalized distribution comparable.

[0110] In one example, the data augmentation module may include a generator built on an LSTM-Self-Attention architecture. G When performing data augmentation on normalized distributed data to obtain synthetic data, random noise vectors can be used. z The feature matrix corresponding to the normalized distribution data is used to construct the input data. The temporal dependency features of the input data are captured by the self-attention mechanism of the data augmentation module, and synthetic data is generated based on these temporal dependency features.

[0111] In one example, the data augmentation module uses a discriminator to determine the generator's loss value using a predefined loss function. This loss value indicates the distributional difference between the normalized and synthetic data. For example, the loss function could include the sum of adversarial loss, dynamic time warping loss (e.g., Soft-DTW), and feature matching loss (e.g., higher-order moment loss).

[0112] By minimizing the loss function, data augmentation goals can be achieved, such as minimizing distributional differences so that the distribution of the synthetic data is as close as possible to the distribution of the original data.

[0113] Based on the above embodiments, the generator of the data augmentation module can be trained until the discriminator can no longer effectively distinguish between normalized distributed data and synthetic data. The generated synthetic data and normalized distributed data are then combined in a certain proportion to obtain the augmented data set.

[0114] Step S502: Use the encoder of the time series prediction model to extract features from the augmented data to obtain multi-scale time series features.

[0115] In some embodiments of this application, the encoder may employ an Inception-Time module, which can extract multi-scale temporal features of the augmented data in parallel during feature extraction. For example, the Inception-Time module contains one-dimensional convolutional kernels of different sizes, and the captured multi-scale temporal features include local and global features of different sizes. In one example, the accuracy of the temporal features extracted by the encoder can also be improved by using batch normalization layers and a preset activation function (e.g., ELU).

[0116] Step S503: Determine the weight of the temporal features at each scale based on the self-attention mechanism of the temporal prediction model, and fuse the multi-scale temporal features according to the weights to obtain the fused features.

[0117] In some embodiments of this application, a multi-head self-attention mechanism can be applied to determine the attention weight of each temporal point in the temporal features. For example, the query matrix of the temporal features at each scale can be calculated. Q Key matrix KSum matrix V Based on query matrix Q Key matrix K Sum matrix V， Attention scores for temporal features at each scale are calculated by scaling dot product attention; softmax normalization is then applied to the attention scores to obtain the weights of the temporal features at each scale.

[0118] In some embodiments of this application, multi-scale temporal features are weighted and fused according to the calculated attention weights to obtain fused features. In one example, a linear layer can be used after weighted fusion to avoid the dimensionality explosion problem. In this way, the fused features can indicate information of important temporal points while suppressing the influence of noise and irrelevant information.

[0119] Step S504: Use the decoder of the time-series prediction model to decode the fused features to obtain the prediction data.

[0120] In some embodiments of this application, the decoder may employ an LSTM-Attention hybrid decoder, using the fused features output by the encoder as initial hidden data. The temporal information in the initial hidden data is decoded progressively by LSTM units, while an attention mechanism focuses on the temporal features relevant to the current decoding step. During decoding, the attention mechanism dynamically adjusts the degree of attention to different temporal points by calculating the attention weight between the decoder's current state and the fused features output by the encoder. The decoder's output is mapped to predicted values ​​through fully connected layers, generating multi-step prediction results to obtain the predicted data. In one example, a teacher-forced strategy can be used to train the decoder to improve its prediction accuracy and stability.

[0121] Based on the above embodiments, a prediction system optimized for small-sample time-series data was constructed by integrating an improved TimeGAN architecture and a multi-scale attention mechanism, achieving high-precision trend extrapolation of the initial change curve. In the data augmentation stage, an LSTM-Self-Attention generator and a TCN discriminator are used for adversarial training, combined with the Soft-DTW loss function and high-order moment feature matching to reduce the dynamic time warping distance between the synthetic data distribution and the original data, thereby improving feature coverage. The encoder extracts multi-scale time-series features in parallel through the Inception-Time module, combining batch normalization and the ELU activation function to improve local feature sensitivity and global pattern capture capability. In the feature fusion stage, a multi-head self-attention mechanism dynamically allocates weights, increasing the contribution of key time-series information and suppressing noise influence. The decoder adopts an LSTM-Attention hybrid structure, effectively improving prediction accuracy compared to the traditional TimeGAN model, making it particularly suitable for short-period time-series prediction scenarios.

[0122] In other embodiments of this application, the time series prediction model can be dynamically and periodically updated based on the updated processing data obtained during practical application. For example, the prediction accuracy of the time series prediction model can be trained and tested using the updated processing data at preset time intervals to ensure the continuous performance of the time series prediction model.

[0123] In one example, methods for anomaly assessment based on a second data line graph include, for example... Figure 6 The process is shown below.

[0124] Step S601: Based on the updated initial change curve, center line, first upper control line and first lower control line, determine the distribution position and change parameters of each data point in the updated initial change curve.

[0125] In some embodiments of this application, the distribution position of the data points can be relative to the center line, the first upper control line, and the first lower control line. For example, they can be located above the first upper control line, below the first lower control line, or outside a preset distance range from the center line. The preset distance range can be a preset range of the workpiece size and can be set according to actual needs; this application does not impose specific limitations on this.

[0126] In one example, the distribution of data points can also be relative to the positions of the second upper control line and the second lower control line. For example, above the second upper control line, below the second lower control line, etc.

[0127] In some embodiments of this application, the variation parameter of the data points can be the variation trend of the workpiece size corresponding to multiple consecutive data points, such as the workpiece size increasing or decreasing corresponding to s2 consecutive (e.g., 6) data points.

[0128] Step S602: Based on the distribution location and change parameters, perform anomaly assessment on the data points of the updated initial change curve in multiple dimensions to obtain anomaly scores for each dimension.

[0129] In some embodiments of this application, different distribution positions and variation parameters can be set as corresponding dimensions individually or in combination. For example, multiple dimensions may include, but are not limited to: data points located above the second upper control line or below the second lower control line; data points located above the first upper control line or below the first lower control line; s1 consecutive (e.g., 15) data points being outside a preset distance range from the center line; and s2 consecutive (e.g., 6) data points corresponding to workpiece dimensions that increase or decrease. The values ​​of s1, s2, etc., can be set according to actual needs, and this application does not impose specific limitations on them.

[0130] In one example, the anomaly assessment of the data points of the updated initial change curve can also be performed by combining the historical number of anomalies of the processing equipment (e.g., number of NGs) and the historical anomaly duration (e.g., NG duration). For example, multiple dimensions can include the number of NGs and the NG duration.

[0131] In one example, reference Figure 7 The diagram shown is an example of multiple dimensions for anomaly assessment provided in an embodiment of this application. It can receive dimensions and corresponding scoring criteria set by expert users to determine multiple dimensions of the anomaly assessment and the anomaly score value for each dimension. For example, various values ​​and regions (e.g., regions A, B, C, etc.) can be set according to actual needs, and this application does not impose specific limitations on them. For example, region A represents the region ±3 sigma of the center line, region B represents the region ±2 sigma of the center line, and region C represents the region ±1 sigma of the center line, where sigma represents a preset value, such as (first control value - second control value) / 6.

[0132] In one example, if the updated initial change curve satisfies a condition in any dimension, it has an anomaly score for that dimension. For example... Figure 7 As shown, if the updated initial change curve is above the second upper control line or below the second lower control line, it has an anomaly score of 10 in this dimension.

[0133] Based on the above embodiments, a multi-dimensional anomaly assessment mechanism can be used to combine statistical control lines, trend patterns, and historical anomaly data to perform comprehensive anomaly detection on the updated time series curve, significantly improving the accuracy and comprehensiveness of anomaly identification.

[0134] Step S603: Determine the anomaly score of the processing equipment based on the anomaly scores of all dimensions.

[0135] In some embodiments of this application, the anomaly score of the processing equipment can be obtained based on the sum of the anomaly scores of all dimensions. For example... Figure 7 As shown, if the updated initial change curve belongs to any dimension, it has an anomaly score for that dimension. The anomaly score of the processing equipment can be obtained by summing the anomaly scores of all dimensions.

[0136] Based on the above embodiments, the overall abnormality level of the equipment can be quantitatively assessed by comprehensively evaluating the abnormality scores of various dimensions, providing data support for precise maintenance.

[0137] In some embodiments of this application, if the processing equipment includes multiple processing devices, the anomaly score of the processing device is determined based on the anomaly scores of all dimensions, including such as Figure 8 The following is the flowchart of the method for determining the anomaly score of the processing equipment.

[0138] Step S801: Based on the anomaly scores of all processing equipment, determine the extreme significance level value of each processing equipment among the multiple processing equipment.

[0139] In some embodiments of this application, ANOVA analysis of variance can be performed on all processing equipment based on the abnormal scores of all processing equipment to calculate the extremely significant level (e.g., P-value) of the average measurement of each processing equipment, in order to assess the statistical significance difference of abnormal score results between different machines.

[0140] Step S802: Normalize the anomaly score of each processing equipment in each dimension to obtain the normalized anomaly score of each processing equipment in each dimension.

[0141] In some embodiments of this application, the outlier scores for each processing device in each dimension are normalized to eliminate the influence of dimensions. For example, the min-max normalization method is used to normalize the outlier scores of all dimensions to the range [0, 1].

[0142] Step S803: Determine the weight of each dimension based on the highly significant level value and the normalized outlier score.

[0143] In some embodiments of this application, the initial weights of each dimension can be determined using the entropy weighting method based on the normalized anomaly scores. For example, an index matrix Y can be constructed. ,in, Indicates the number of processing equipment. Indicates the number of dimensions. Indicates the first The processing equipment is in the first Normalized anomaly scores for each dimension.

[0144] The formula used to calculate information entropy based on the indicator matrix is ​​as follows: , in, Indicates the first Information entropy in one dimension is used to indicate the degree of dispersion of the data. The greater the dispersion, the greater the information entropy.

[0145] The formula used to calculate information utility value based on information entropy is as follows: , in, Indicates the first The information utility value of each dimension represents the degree to which that dimension contributes to the overall evaluation.

[0146] The initial weights are calculated based on information utility values, using the following formula: , in, Indicates the first Initial weights for each dimension.

[0147] Based on the above embodiments, the initial weights of each dimension can be dynamically determined based on the normalized anomaly score using the entropy weight method, effectively quantifying the contribution of each dimension to the overall anomaly assessment and improving the objectivity and accuracy of the assessment.

[0148] In some embodiments of this application, the initial weights can be updated based on the highly significant level value to obtain the final weights. For example, if the P-value of a certain processing equipment is less than a preset P-value threshold (e.g., 0.03), it indicates that the processing equipment has a large systematic deviation. The dimension corresponding to the maximum abnormal score of the processing equipment can be determined, and the weight of the determined dimension can be increased by a preset proportion, such as 10%. For example, if the initial weight of the dimension corresponding to the maximum abnormal score is 0.3, it can be increased by 10%, updating the weight to 0.33.

[0149] Based on the above embodiments, the dimension weights can be dynamically adjusted based on the extremely significant level value (P-value). For processing equipment with significant systematic deviations, the weight of the dimension with the largest abnormal score can be increased, thereby enhancing the sensitivity to key anomalies.

[0150] Step S804: Based on the weight of each dimension and the normalized anomaly score of each processing equipment in each dimension, determine the weighted sum of all normalized anomaly scores corresponding to all dimensions to obtain the anomaly score of each processing equipment.

[0151] In some embodiments of this application, after determining the weight of each dimension, the weighted sum of the normalized anomaly scores of the processing equipment in each dimension can be used as the anomaly score of the processing equipment.

[0152] Based on the above embodiments, after determining the weights of each dimension, the normalized anomaly score of each processing equipment in all dimensions is calculated and weighted to comprehensively and quantitatively assess the degree of anomaly of the processing equipment, generating a score that reflects the overall anomaly risk of the equipment, and providing a basis for subsequent graded alarms and risk response.

[0153] Figure 9 This is a structural diagram of an equipment anomaly monitoring device provided in an embodiment of this application.

[0154] In some embodiments, the device anomaly monitoring device 90 may include multiple functional modules composed of computer program segments. The computer programs for each program segment in the device anomaly monitoring device 90 may be stored in the memory of the electronic device and executed by at least one processor to perform (see details). Figure 2 (Description) The function of equipment anomaly monitoring.

[0155] In this embodiment, the equipment anomaly monitoring device 90 can be divided into multiple functional modules according to its functions. These functional modules may include: a generation module 901, a prediction module 902, an update module 903, a scoring module 904, and a determination module 905. The term "module" in this application refers to a series of computer program segments that can be executed by at least one processor and perform a fixed function, stored in memory. In this embodiment, the functional implementation of each module in the equipment anomaly monitoring device 90 can be found in the above description of the equipment anomaly monitoring method, and will not be repeated here.

[0156] The generation module 901 is used to acquire processing data from the processing equipment and generate a first data line graph corresponding to the processing data. The first data line graph includes the initial change curve of the processing data over time.

[0157] The prediction module 902 is used to perform time-series prediction on the initial change curve based on the time-series characteristics of the initial change curve to obtain prediction data.

[0158] The update module 903 is used to update the first data line graph based on the predicted data to obtain a second data line graph, wherein the second data line graph includes the updated initial change curve.

[0159] The scoring module 904 is used to determine the abnormal score of the processing equipment based on the distribution position and change parameters of the updated initial change curve.

[0160] The determination module 905 is used to determine the abnormal monitoring result of the processing equipment based on the abnormal score.

[0161] Please see Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 2 As shown, the device anomaly monitoring method provided in this application embodiment can be applied to electronic device 10. Electronic device 10 can be mobile phone, tablet computer, smart wearable device, augmented reality (AR) / virtual reality (VR) device, laptop computer, netbook, energy storage device, power distribution equipment, vehicle equipment, self-moving device, scanner and other electronic devices. This application embodiment does not impose any restrictions on the specific type of electronic device.

[0162] like Figure 2 As shown, the electronic device 10 may include a communication module 101, a memory 102, a processor 103, an input / output (I / O) interface 104, and a bus 105. The processor 103 is coupled to the communication module 101, the memory 102, and the I / O interface 104 via the bus 105.

[0163] Communication module 101 may include a wired communication module and / or a wireless communication module. The wired communication module may provide one or more wired communication solutions such as Universal Serial Bus (USB) and Controller Area Network (CAN). The wireless communication module may provide one or more wireless communication solutions such as Wireless Fidelity (Wi-Fi), Bluetooth (BT), mobile communication networks, Frequency Modulation (FM), Near Field Communication (NFC), and Infrared (IR).

[0164] Memory 102 may include one or more random access memory (RAM) and one or more non-volatile memory (NVM). The RAM can be directly read and written by the processor 103, and can be used to store executable programs (such as machine instructions) of the operating system or other running programs, as well as user and application data. The RAM may include static random-access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), etc.

[0165] Non-volatile memory can also store executable programs and user and application data, and can be pre-loaded into random access memory for direct reading and writing by the processor 103. Non-volatile memory can include disk storage devices and flash memory.

[0166] The memory 102 is used to store one or more computer programs. The one or more computer programs are configured to be executed by the processor 103. The one or more computer programs include multiple instructions that, when executed by the processor 103, enable a device anomaly monitoring method to be executed on the electronic device 10.

[0167] In other embodiments, the electronic device 10 further includes an external memory interface for connecting to an external memory to expand the storage capacity of the electronic device 10.

[0168] Processor 103 may include one or more processing units, such as application processor (AP), modem processor, graphics processing unit (GPU), image signal processor (ISP), controller, video codec, digital signal processor (DSP), baseband processor, and / or neural network processing unit (NPU). These different processing units may be independent devices or integrated into one or more processors.

[0169] The processor 103 provides computing and control capabilities. For example, the processor 103 is used to execute computer programs stored in the memory 102 to implement the device anomaly monitoring method described above.

[0170] I / O interface 104 is used to provide a channel for user input or output. For example, I / O interface 104 can be used to connect various input / output devices, such as a mouse, keyboard, touch device, display screen, etc., so that users can enter information or visualize information. In addition, I / O interface 104 can also be used to connect processing equipment, such as processing equipment 2, to obtain the required processing data and other data.

[0171] Bus 105 is used at least to provide a channel for communication between communication modules 101, memory 102, processor 103, and I / O interface 104 in electronic device 10.

[0172] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the electronic device 10. In other embodiments of this application, the electronic device 10 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0173] This application also provides a computer-readable storage medium storing a computer program, which includes program instructions. When the program instructions are executed, the method implemented can refer to the methods in the above embodiments of this application.

[0174] The computer-readable storage medium can be the internal memory of the electronic device described in the above embodiments, such as the hard disk or memory of the electronic device. Alternatively, the computer-readable storage medium can be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., provided on the electronic device.

[0175] In some embodiments, a computer-readable storage medium may include a stored program area and a stored data area, wherein the stored program area may store an operating system, an application program required for at least one function, etc.; and the stored data area may store data created based on the use of the electronic device, etc.

[0176] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0177] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0178] In the embodiments provided in this application, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0179] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0180] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for monitoring equipment anomalies, characterized in that, The method includes: Acquire processing data from the processing equipment and generate a first data line graph corresponding to the processing data. The first data line graph includes an initial change curve of the processing data over time. Based on the temporal characteristics of the initial change curve, a temporal prediction is performed on the initial change curve to obtain predicted data; The first data line graph is updated based on the predicted data to obtain a second data line graph, the second data line graph including the updated initial change curve; Based on the distribution location and change parameters of the updated initial change curve, the anomaly score of the processing equipment is determined; Based on the anomaly score, the anomaly monitoring results of the processing equipment are determined.

2. The equipment anomaly monitoring method according to claim 1, characterized in that, The processing data includes the workpiece dimensions and processing time of the same type of workpiece processed by the processing equipment.

3. The equipment anomaly monitoring method according to claim 2, characterized in that, The method further includes: preprocessing the processing data, the preprocessing including: Anomaly detection algorithms are used to identify anomalous data in the processing data, and the anomalous data is then cleaned. The anomalous data includes outliers and noisy data. The cleaned processing data is sorted according to the processing time to obtain the sorted processing data; According to the rule of grouping the processing data of a preset number of workpieces into groups, the sorted processing data is grouped to obtain multiple data groups.

4. The equipment anomaly monitoring method according to claim 3, characterized in that, The generation of the first data line chart corresponding to the processing data includes: Based on the sorted processing data, determine the change curve of the workpiece size with the processing time, and obtain the initial change curve; A first average size is determined based on the workpiece size in the plurality of data groups, and a first control value and a second control value are determined based on the workpiece size in each of the plurality of data groups. The center line of the first data line graph is determined based on the first average size, the first upper control line is determined based on the first control value, and the first lower control line is determined based on the second control value. The first data line graph is generated based on the initial change curve, the center line, the first upper control line, and the first lower control line.

5. The equipment anomaly monitoring method according to claim 4, characterized in that, The step of determining the first average size, the first control value, and the second control value based on the workpiece size in each of the plurality of data groups includes: The average value of all workpiece dimensions in the plurality of data groups is determined to obtain the first average dimension value; The difference between the maximum and minimum workpiece size in each data group is determined to obtain the size range; Based on the size range of each data group, the average size range of the multiple data groups is determined to obtain the mean range. The first control value is determined based on the sum of the first size mean and the range mean, and the second control value is determined based on the difference between the first size mean and the range mean.

6. The equipment anomaly monitoring method according to claim 1, characterized in that, The step of performing time-series prediction on the initial change curve based on its time-series characteristics to obtain prediction data includes: The original distribution data is obtained based on the initial change curve. The data augmentation module of the time series prediction model is then used to augment the original distribution data to obtain augmented data. The encoder of the time series prediction model is used to extract features from the enhanced data to obtain multi-scale time series features; The weights of the temporal features at each scale are determined based on the self-attention mechanism of the time series prediction model, and the multi-scale temporal features are fused according to the weights to obtain fused features. The fused features are decoded using the decoder of the time-series prediction model to obtain the prediction data.

7. The equipment anomaly monitoring method according to claim 6, characterized in that, The data augmentation module using the time-series prediction model performs data augmentation on the original distributed data to obtain augmented data, including: The original distribution data is normalized to obtain normalized distribution data; Data augmentation is performed on the normalized distribution data to obtain synthetic data, wherein the objective of the data augmentation includes minimizing the distribution difference between the synthetic data and the normalized distribution data; The enhanced data is obtained by combining the normalized distribution data and the synthetic data.

8. The equipment anomaly monitoring method according to claim 1, characterized in that, The step of updating the first data line chart based on the predicted data to obtain the second data line chart includes: Based on the prediction curve corresponding to the prediction data, update the initial change curve and update the time range corresponding to the first data line graph; The second data line graph is obtained based on the updated initial change curve and the updated time range.

9. The equipment anomaly monitoring method according to claim 1, characterized in that, The second data graph includes an updated initial change curve, a center line, a first upper control line, and a first lower control line. Determining the anomaly score of the processing equipment based on the distribution position and change parameters of the updated initial change curve includes: Based on the updated initial change curve, the center line, the first upper control line, and the first lower control line, the distribution position and change parameters of each data point in the updated initial change curve are determined. Based on the distribution location and the change parameters, the data points of the updated initial change curve are evaluated for anomalies in multiple dimensions to obtain an anomaly score for each dimension. The anomaly score of the processing equipment is determined based on the anomaly scores of all dimensions.

10. The equipment anomaly monitoring method according to claim 9, characterized in that, If the processing equipment includes multiple processing devices, determining the anomaly score of the processing device based on the anomaly scores of all dimensions includes: Based on the anomaly scores of all processing equipment, determine the extreme significance level value of each of the multiple processing equipment; The anomaly scores of each processing device in each dimension are normalized to obtain the normalized anomaly scores of each processing device in each dimension. The weight of each dimension is determined based on the highly significant level value and the normalized outlier score. Based on the weight of each dimension and the normalized anomaly score of each processing device in each dimension, the weighted sum of all normalized anomaly scores corresponding to all dimensions is determined to obtain the anomaly score of each processing device.

11. The equipment anomaly monitoring method according to claim 1, characterized in that, The step of determining the anomaly monitoring result of the processing equipment based on the anomaly score includes: If the anomaly score is greater than a preset score threshold, the anomaly monitoring result indicates that the processing equipment has an anomaly; or... If the anomaly score is less than or equal to the preset score threshold, the anomaly monitoring result indicates that the processing equipment does not have an anomaly.

12. The equipment anomaly monitoring method according to claim 1, characterized in that, The method further includes: Based on the processing data of the processing equipment and the anomaly monitoring results, a knowledge graph is constructed. The knowledge graph includes the processing data of the processing equipment with anomalies, the anomaly score of the processing equipment with anomalies, the anomaly cause of the processing equipment with anomalies, and the anomaly solution of the processing equipment with anomalies.

13. An electronic device, characterized in that, The electronic device includes a processor and a memory, wherein the processor is configured to implement the device anomaly monitoring method as described in any one of claims 1 to 12 when executing a computer program stored in the memory.

Images