Production workshop intelligent monitoring method and system based on industrial big data AI analysis

Abstract

The invention discloses a production workshop intelligent monitoring method and system based on industrial big data AI analysis, and relates to the related field of data processing.The method comprises the steps that multi-source heterogeneous data streams of a production workshop are collected in real time, timestamp alignment and format standardization processing are conducted on the multi-source heterogeneous data streams through edge computing nodes, and synchronous data streams are obtained; performing dynamic quality evaluation on the synchronous data stream, and generating a data exception label when detecting that any quality index of any data source is exceptional; according to the data exception label, a differential fault-tolerant strategy is triggered to carry out data fault-tolerant repair, and a repaired synchronous data stream is obtained; and performing real-time equipment anomaly detection on the repaired synchronous data stream to obtain an equipment health state. The technical problems of inaccurate equipment state monitoring and low production workshop intelligent monitoring level in existing production workshop intelligent monitoring are solved, and the technical effects of improving the equipment state monitoring accuracy and improving the production workshop intelligent monitoring level are achieved.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to intelligent monitoring methods and systems for production workshops based on industrial big data AI analysis. Background Technology

[0002] In modern industrial production, monitoring the operational status of equipment in production workshops is crucial for ensuring production safety, improving production efficiency, and reducing maintenance costs. Accurate monitoring and timely problem detection can prevent major production accidents and optimize production processes. Currently, the main approach to address this issue is to build a workshop monitoring system. This system uses sensors to collect equipment data and transmits it to a central server for centralized processing and analysis to monitor and assess equipment status. However, current methods suffer from several drawbacks. Production workshop data is multi-source and heterogeneous, with different data sources exhibiting differences in time synchronization, format consistency, and data quality. Furthermore, data transmission is susceptible to interference, leading to issues such as time asynchrony, inconsistent formatting, and varying data quality received by the central server. These problems ultimately affect the accuracy and timeliness of equipment status monitoring.

[0003] At present, the relevant technologies for intelligent monitoring of production workshops have technical problems such as inaccurate equipment status monitoring and low level of intelligent monitoring of production workshops. Summary of the Invention

[0004] This application provides a method and system for intelligent monitoring of production workshops based on industrial big data AI analysis. It utilizes edge computing nodes to perform timestamp alignment and format standardization on real-time collected multi-source heterogeneous data streams, forming a synchronous data stream. The synchronous data stream undergoes dynamic quality assessment; once an anomaly in the data source quality indicators is detected, a tag is generated, and a differentiated fault-tolerance strategy is triggered based on the tag to repair the data, resulting in a repaired synchronous data stream. Real-time equipment anomaly detection is then performed on the repaired synchronous data stream. These technical means solve the technical problems of inaccurate equipment status monitoring and low levels of intelligent monitoring in existing intelligent monitoring systems for production workshops, achieving the technical effect of improving the accuracy of equipment status monitoring and enhancing the level of intelligent monitoring in production workshops.

[0005] This application provides a method for intelligent monitoring of production workshops based on industrial big data AI analysis, including: real-time acquisition of multi-source heterogeneous data streams from the production workshop; performing timestamp alignment and format standardization on the multi-source heterogeneous data streams through edge computing nodes to obtain a synchronous data stream; performing dynamic quality assessment on the synchronous data stream; generating a data anomaly label when any quality indicator of any data source is detected to be abnormal; triggering a differentiated fault tolerance strategy to perform data fault tolerance repair based on the data anomaly label to obtain a repaired synchronous data stream; and performing real-time equipment anomaly detection on the repaired synchronous data stream to obtain the equipment health status.

[0006] In a possible implementation, dynamic quality assessment is performed on the synchronous data stream. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated, and the following processing is performed: signal-to-noise ratio (SNR) and transmission stability are assessed on the synchronous data stream to obtain SNR and transmission stability indicator values; when the SNR indicator value is less than a preset SNR threshold, it is determined to be a signal quality anomaly; when the transmission stability indicator value does not meet a preset transmission stability threshold, it is determined to be a transmission anomaly; when any data source is detected to have the signal quality anomaly and / or the transmission anomaly, a data anomaly label is generated.

[0007] In a possible implementation, when any data source is detected to have signal quality abnormality and / or transmission abnormality, a data abnormality label is generated, and the following processing is performed: when any data source is detected to have signal quality abnormality, a signal abnormality label is generated; when any data source is detected to have transmission abnormality, a transmission abnormality label is generated; when any data source is detected to have both signal quality abnormality and transmission abnormality, a composite abnormality label is generated.

[0008] In a possible implementation, based on the data anomaly label, a differentiated fault-tolerance strategy is triggered to perform data fault-tolerance repair, resulting in a repaired synchronous data stream. The following processing is then performed: If the data anomaly label is the signal anomaly label, the signal anomaly duration is extracted based on the timestamp information of the signal anomaly label; if the signal anomaly duration is greater than a preset signal duration threshold, it is marked as a persistent signal anomaly; otherwise, it is marked as short-term signal interference; for the persistent signal anomaly, a differentiated fault-tolerance strategy is triggered to switch to a backup data source, and bidirectional linear interpolation is performed to perform data fault-tolerance interpolation for the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream, wherein the order of the bidirectional linear interpolation is dynamically determined based on the signal anomaly duration; for the short-term signal interference, a differentiated fault-tolerance strategy is triggered to perform wavelet denoising repair for the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream.

[0009] In a possible implementation, based on the data anomaly label, a differentiated fault tolerance strategy is triggered to perform data fault tolerance repair, resulting in a repaired synchronous data stream. The following processing is then performed: If the data anomaly label is the transmission anomaly label, the transmission anomaly duration is extracted based on the timestamp information of the transmission anomaly label; if the transmission anomaly duration exceeds a preset transmission duration threshold, it is marked as a persistent transmission anomaly; otherwise, it is marked as short-term transmission interference. For the persistent transmission anomaly, a differentiated fault tolerance strategy is triggered to activate the backup communication link, and Kalman smoothing prediction is performed to complete the data fault tolerance during the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream. For the short-term transmission interference, a differentiated fault tolerance strategy is triggered based on a decision matrix to perform data fault tolerance repair during the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream. The decision matrix stores the correspondence between the transmission anomaly duration and the repair strategy.

[0010] In a possible implementation, based on the data anomaly label, a differentiated fault-tolerance strategy is triggered to perform data fault-tolerance repair, resulting in a repaired synchronous data stream. The following processing is then performed: If the data anomaly label is the composite anomaly label, the composite anomaly duration is extracted based on the timestamp information of the composite anomaly label; if the composite anomaly duration exceeds a preset composite duration threshold, it is marked as a persistent composite anomaly; otherwise, it is marked as a short-term composite interference. For the persistent composite anomaly, a differentiated fault-tolerance strategy is triggered to acquire virtual sensor data through a digital twin system to complete the data during the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream. For the short-term composite interference, a differentiated fault-tolerance strategy is triggered to enable an anti-interference communication mode, and simultaneously, relevant normal sensor data fusion processing is performed to correct data redundancy during the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream.

[0011] In a possible implementation, dynamic quality assessment is performed on the synchronous data stream. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated. The following processing is also performed: based on the production process knowledge base, predefined sensor multimodal data constraint rules for monitoring points are defined; when the synchronous data stream does not trigger the signal quality anomaly and the transmission anomaly, the sensor multimodal data constraint rule detection is performed. If the constraint rules are violated, a logical conflict label is generated; based on the logical conflict label, the logical conflict type is extracted, and a targeted fault tolerance strategy is triggered based on the logical conflict type to perform data fault tolerance repair, thereby obtaining a repaired synchronous data stream.

[0012] In possible implementations, the following processing is also performed: periodically injecting simulated exception tags to test the performance of the differentiated fault tolerance strategy, and recording the fault tolerance strategy performance indicators corresponding to each tag; when any performance indicator fails to meet the preset fault tolerance performance threshold, generating a strategy optimization instruction to dynamically optimize the differentiated fault tolerance strategy.

[0013] In a possible implementation, after real-time device anomaly detection is performed on the repair synchronization data stream to obtain the device health status, the following processing is also performed: when the device health status indicates a high-risk fault, the digital twin system is linked to simulate the fault impact range; if the fault impact range is an independent production unit, the equipment operating parameters of the corresponding production unit are adjusted; if the fault impact range is a cross-unit collaborative production chain, a graded shutdown protection is triggered according to the criticality weight and an alarm is pushed to the mobile terminal.

[0014] This application also provides an intelligent monitoring system for production workshops based on industrial big data AI analysis, including: a synchronous data stream acquisition module, used to collect multi-source heterogeneous data streams from the production workshop in real time, and to perform timestamp alignment and format standardization processing on the multi-source heterogeneous data streams through edge computing nodes to obtain a synchronous data stream; a data anomaly detection module, used to perform dynamic quality assessment on the synchronous data stream, and to generate a data anomaly tag when any quality indicator of any data source is detected to be abnormal; a data fault tolerance repair module, used to trigger a differentiated fault tolerance strategy to perform data fault tolerance repair based on the data anomaly tag, and to obtain a repaired synchronous data stream; and a real-time equipment anomaly detection module, used to perform real-time equipment anomaly detection on the repaired synchronous data stream to obtain the equipment health status.

[0015] The proposed intelligent monitoring method and system for production workshops based on industrial big data AI analysis, as outlined in this application, firstly collects multi-source heterogeneous data streams from the production workshop in real time. Then, it performs timestamp alignment and format standardization on these data streams using edge computing nodes to obtain a synchronous data stream. Next, it performs dynamic quality assessment on this synchronous data stream. When any quality indicator from any data source is detected to be abnormal, a data anomaly label is generated. Based on this label, a differentiated fault-tolerance strategy is triggered to perform data fault-tolerance repair, resulting in a repaired synchronous data stream. Finally, the repaired synchronous data stream is used for real-time equipment anomaly detection to obtain the equipment health status. This achieves the technical effect of improving the accuracy of equipment status monitoring and enhancing the level of intelligent monitoring in the production workshop. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1A flowchart illustrating the intelligent monitoring method for production workshops based on industrial big data AI analysis provided in this application embodiment.

[0018] Figure 2 This is a schematic diagram of the structure of an intelligent monitoring system for a production workshop based on industrial big data AI analysis, provided in an embodiment of this application.

[0019] Explanation of reference numerals in the attached diagram: Synchronous data stream acquisition module 10, data anomaly detection module 20, data fault tolerance and repair module 30, real-time device anomaly detection module 40. Detailed Implementation

[0020] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] In the following description, references to "some embodiments" describe a subset of all possible embodiments; however, it is understood that "some embodiments" may be the same or different subsets of all possible embodiments and may be combined with each other without conflict. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0023] This application provides an intelligent monitoring method for production workshops based on industrial big data AI analysis, such as... Figure 1 As shown, the method includes: Step S100: Collect multi-source heterogeneous data streams from the production workshop in real time, and perform timestamp alignment and format standardization processing on the multi-source heterogeneous data streams through edge computing nodes to obtain synchronous data streams.

[0024] Specifically, multi-source heterogeneous data streams refer to data streams from different sensors, which have different formats and structures. Edge computing nodes are computing nodes deployed close to the sensors to perform preliminary data processing, reducing latency and bandwidth consumption when data is transmitted to the cloud or central server. Timestamp alignment refers to aligning data collected from different sensors according to timestamps to ensure data consistency over time. Format standardization refers to converting data in different formats into a unified format to facilitate subsequent processing and analysis.

[0025] Specifically, various operational data of the equipment are collected in real time through a sensor network. The sensor network consists of various sensors (such as temperature sensors, pressure sensors, vibration sensors, etc.) deployed in the production workshop to collect equipment operation data in real time. The collected data is transmitted to edge computing nodes through the industrial network. At the edge computing nodes, the data collected by different sensors are aligned according to timestamps to ensure the consistency of the data in time. The aligned data is then converted into a unified format to facilitate subsequent processing and analysis.

[0026] For example, in a machinery manufacturing workshop, temperature sensors, pressure sensors, and vibration sensors are installed on key equipment to collect real-time operational data. Multiple edge computing nodes are set up within the workshop, each responsible for processing sensor data within a certain range. Assuming sensor A collects data with a timestamp of 10:00:00.001 and sensor B collects data with a timestamp of 10:00:00.002, the edge computing nodes align these two timestamps to 10:00:00.000. The data format of sensor A is converted from CSV to JSON, and the data format of sensor B is converted from XML to JSON, ultimately resulting in a synchronized data stream with a unified format.

[0027] Step S200: Perform dynamic quality assessment on the synchronized data stream. When any quality indicator of any data source is detected to be abnormal, generate a data anomaly label.

[0028] Specifically, dynamic quality assessment refers to real-time analysis of synchronous data streams, calculating data quality indicators, and evaluating data quality. Data anomaly labels are tags generated when data quality anomalies are detected, recording the characteristics and location of the abnormal data.

[0029] Specifically, data quality assessment metrics are defined, such as signal-to-noise ratio (SNR). Statistical analysis is used to perform real-time analysis of the synchronous data stream, calculating the value of each data quality metric. Threshold detection, cluster analysis, and time series analysis are employed to detect anomalies in the data quality metrics. When a data quality anomaly is detected, a corresponding anomaly label is generated, recording the characteristics and location of the abnormal data.

[0030] In one possible implementation, dynamic quality assessment is performed on the synchronous data stream. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated. Step S200 further includes step S210, which performs signal-to-noise ratio (SNR) and transmission stability assessment on the synchronous data stream to obtain SNR and transmission stability index values. Specifically, a signal processing algorithm (such as Fast Fourier Transform, FFT) is used to assess the SNR of each data source in the synchronous data stream, and the SNR index value is calculated. A network performance assessment algorithm (such as packet loss rate, latency jitter) is used to assess the transmission stability of each data source in the synchronous data stream, and the transmission stability index value is calculated.

[0031] For example, assuming sensor A collects signal data x(t), its spectrum is calculated using FFT to separate the signal and noise components, and the signal-to-noise ratio (SNR) is calculated. Similarly, assuming sensor B's data transmission path involves measuring packet loss rate and latency jitter using network monitoring tools, transmission stability metrics are calculated.

[0032] Step S220: When the signal-to-noise ratio (SNR) index value is less than a preset SNR threshold, it is determined that the signal quality is abnormal. Specifically, the SNR threshold is set based on actual application scenarios and experience. The SNR index value of each data source is compared with the preset SNR threshold; if it is less than the threshold, it is determined that the signal quality is abnormal.

[0033] For example, in an industrial workshop, based on historical data and experience, the signal-to-noise ratio (SNR) threshold is set at 18 dB. Sensor B has an SNR of 15 dB, which is less than the preset threshold of 18 dB, and is therefore judged as an abnormal signal quality.

[0034] Step S230: When the transmission stability index value does not meet the preset transmission stability threshold, it is determined to be a transmission anomaly. Specifically, the transmission stability threshold is set based on actual application scenarios and experience. The transmission stability index value of each data source is compared with the preset transmission stability threshold; if the threshold is not met, it is determined to be a transmission anomaly.

[0035] For example, in an industrial workshop, the packet loss rate threshold is set to 0.3% based on network performance requirements. Sensor B's packet loss rate is 0.5%, which does not meet the preset threshold of 0.3%, and is therefore judged as a transmission anomaly.

[0036] Step S240: When any signal quality abnormality and / or transmission abnormality is detected in any data source, a data abnormality tag is generated. Specifically, the timestamp, data source identifier, abnormality type, and other information of the abnormal data are recorded. A data abnormality tag is generated based on the recorded abnormal information.

[0037] For example, record the abnormal information of sensor B at 10:00:05.000, including timestamp, data source identifier, abnormality type, etc., and generate data abnormality labels, such as {"timestamp":"10:00:05.000","sensor_id":"sensor_B","abnormal_type":"signal_quality","abnormal_detail":"SNR=15 dB"}.

[0038] This approach, through signal-to-noise ratio and transmission stability assessments, can promptly identify data quality issues such as signal interference and transmission instability, allowing for appropriate fault-tolerant strategies to be implemented and repaired, thereby improving data accuracy and reliability. Using data that has undergone quality assessment and repair for equipment anomaly detection can more accurately reflect the actual operating status of the equipment, reducing false alarms and missed alarms, and improving the efficiency and effectiveness of equipment health management.

[0039] In one possible implementation, when any data source is detected to have signal quality abnormality and / or transmission abnormality, a data abnormality tag is generated. Step S240 further includes step S241, when any data source is detected to have signal quality abnormality, a signal abnormality tag is generated; step S242, when any data source is detected to have transmission abnormality, a transmission abnormality tag is generated; and step S243, when any data source is detected to have both signal quality abnormality and transmission abnormality, a composite abnormality tag is generated.

[0040] Specifically, the signal-to-noise ratio (SNR) and transmission stability metrics for each data source are monitored in real time. The monitored SNR values ​​are compared to preset SNR thresholds, and the transmission stability metrics are also compared to preset transmission stability thresholds. If the SNR value is lower than the preset SNR threshold, but the transmission stability metric meets the preset transmission stability threshold, the data source is determined to have signal quality anomalies, but normal transmission stability. A signal anomaly label is generated for this data source, containing key information such as the data source identifier and anomaly type (signal quality anomaly). If the transmission stability metric does not meet the preset transmission stability threshold, but the SNR value is greater than or equal to the preset SNR threshold, the data source is determined to have transmission anomalies, but normal signal quality. A transmission anomaly label is generated for this data source, containing key information such as the data source identifier and anomaly type (transmission anomaly). If the SNR value is lower than the preset SNR threshold, and the transmission stability metric does not meet the preset transmission stability threshold, the data source is determined to have both signal quality and transmission anomalies. A composite anomaly label is generated for this data source, containing key information such as the data source identifier and anomaly type (composite anomaly).

[0041] This implementation method, by simultaneously monitoring signal-to-noise ratio and transmission stability metrics and generating different types of labels based on different anomalies, can accurately pinpoint the specific problems in the data source, whether it is a signal quality anomaly, a transmission anomaly, or a combined anomaly of both. Different anomaly labels correspond to different fault tolerance strategies, and targeted fault tolerance processing can more effectively solve data quality problems and improve the accuracy and reliability of the data.

[0042] In one possible implementation, dynamic quality assessment is performed on the synchronized data stream. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated. Step S200 further includes step S250, which predefines sensor multimodal data constraint rules for monitoring points based on a production process knowledge base. Specifically, knowledge and experience related to the production process are collected and organized, including information such as production process flow, equipment operating parameters, and sensor data associations, to construct a production process knowledge base. Using a rule definition tool, sensor multimodal data constraint rules for each monitoring point are predefined based on the information in the production process knowledge base. These rules may include data range, logical relationships between data, etc. The defined sensor multimodal data constraint rules are stored in the system for subsequent detection.

[0043] For example, suppose that in a mechanical manufacturing workshop, the production process knowledge base records the following information: The temperature sensor and pressure sensor data of equipment A should satisfy the following logical relationship: when the temperature exceeds 80°C, the pressure should be higher than 100 kPa. Based on this information, the predefined sensor multimodal data constraint rule is: if the temperature sensor data exceeds 80°C, the pressure sensor data should be higher than 100 kPa.

[0044] Step S260: When the synchronized data stream does not trigger the signal quality anomaly or the transmission anomaly, the sensor multimodal data constraint rule detection is performed. If the constraint rules are violated, a logical conflict label is generated. Specifically, each data source in the synchronized data stream is monitored in real time to check whether the data from each data source violates the predefined sensor multimodal data constraint rules. If the data does not trigger the signal quality anomaly or the transmission anomaly, but violates the sensor multimodal data constraint rules, it is determined to be a logical conflict. A logical conflict label is generated for this data source, and the label contains key information such as the data source identifier and the violated sensor multimodal data constraint rule.

[0045] For example, suppose that at a certain moment, the temperature sensor data of device M is 85°C and the pressure sensor data is 90 kPa. According to the predefined sensor multimodal data constraint rule (when the temperature exceeds 80°C, the pressure should be higher than 100 kPa), the pressure data violates the sensor multimodal data constraint rule at this time.

[0046] Step S270: Based on the logical conflict label, extract the logical conflict type, and trigger a targeted fault tolerance strategy to perform data fault tolerance repair based on the logical conflict type, obtaining a repaired synchronous data stream. Specifically, extract the specific type of logical conflict from the logical conflict label, select the corresponding targeted fault tolerance strategy based on the extracted logical conflict type, and perform fault tolerance repair on the data that violates the sensor multimodal data constraint rules. The repaired data is then reassembled into a synchronous data stream for subsequent equipment anomaly detection.

[0047] For example, assuming the logical conflict type is "temperature and pressure data are logically inconsistent," the directional fault-tolerance strategy is: if the temperature data is normal, but the pressure data violates the sensor's multimodal data constraint rules, then the temperature data is used to predict the pressure data. Based on this strategy, the pressure data is predicted using the temperature data of 85°C, and the pressure value calculated by the prediction model is assumed to be 105 kPa.

[0048] This approach, by introducing multimodal data constraint rule detection based on a production process knowledge base, can more comprehensively monitor data quality, including not only signal quality and transmission stability, but also the logical relationships between data. This allows for more accurate detection of potential data quality issues and reduces misjudgments and erroneous decisions caused by data quality problems.

[0049] Step S300: Based on the data anomaly label, trigger the differentiated fault tolerance strategy to perform data fault tolerance repair and obtain the repaired synchronous data stream.

[0050] Specifically, differentiated fault-tolerance strategies are designed based on the type of data anomaly and sensor characteristics. Different fault-tolerance strategies are employed, including data imputation, interpolation, and prediction, to repair the abnormal data. The repaired data is then reassembled into a synchronous data stream, known as the repaired synchronous data stream (the fault-tolerant repaired synchronous data stream), for use in subsequent equipment anomaly detection.

[0051] For example, for sensor data missing anomalies, linear interpolation is used to fill in the missing data; for data error anomalies, a prediction model based on historical data is used to correct the erroneous data. Assuming sensor A has missing data at 10:00:05.000, the missing data at 10:00:05.000 is calculated using linear interpolation based on data from 10:00:04.000 and 10:00:06.000. The repaired data is then reassembled into a synchronized data stream, for example, the repaired synchronized data stream is {"timestamp":"10:00:05.000", "sensor_A":"repaired data value", "sensor_B":"original data value"}.

[0052] In one possible implementation, based on the data anomaly label, a differentiated fault-tolerance strategy is triggered to perform data fault-tolerance repair, resulting in a repaired synchronous data stream. Step S300 further includes step S310: if the data anomaly label is the signal anomaly label, the signal anomaly duration is extracted based on the timestamp information of the signal anomaly label. Specifically, after detecting the signal anomaly label, the start and end timestamps in the label are extracted, and the time difference between the two timestamps is calculated to obtain the signal anomaly duration.

[0053] Step S320: If the duration of the abnormal signal exceeds a preset signal duration threshold, it is marked as a continuous signal abnormality; otherwise, it is marked as short-term signal interference. Specifically, a preset signal duration threshold is set, for example, 5 seconds. The duration of the abnormal signal is compared with the preset signal duration threshold. If the duration of the abnormal signal exceeds the preset signal duration threshold, it is marked as a continuous signal abnormality; otherwise, it is marked as short-term signal interference.

[0054] Step S330: For the persistent signal anomaly, a differentiated fault-tolerant strategy is triggered to switch to the backup data source. Simultaneously, bidirectional linear interpolation is performed to perform data fault-tolerant interpolation for the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream. The order of the bidirectional linear interpolation is dynamically determined based on the duration of the signal anomaly. Specifically, when a persistent signal anomaly occurs in the primary data source, the system automatically switches to the backup data source. This backup data source switching ensures that the system can continue to provide reliable data even when the primary data source experiences a persistent failure. The backup data source can be calibrated redundant sensors or historical data, which can serve as a temporary replacement when the primary data source fails, thereby reducing data interruption time and ensuring the continuity of production monitoring.

[0055] Based on the timestamp information of the signal anomaly tags, the start and end times of the abnormal period are determined. The order of bidirectional linear interpolation is dynamically determined based on the duration of the signal anomaly; the longer the signal anomaly duration, the higher the interpolation order, to improve interpolation accuracy. For example, if the signal anomaly duration is 5 seconds, the interpolation order can be set to 3; if the signal anomaly duration is 10 seconds, the interpolation order can be set to 5. The bidirectional linear interpolation algorithm is used to interpolate the data during the abnormal period. Bidirectional linear interpolation considers the sequential relationship of time series data on the time axis, filling data gaps and restoring data continuity through linear interpolation. The interpolated data is combined with normal data to obtain a repaired synchronized data stream.

[0056] Step S340: For the short-term signal interference, a differentiated fault-tolerant strategy is triggered to perform wavelet denoising repair on the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream. Specifically, when the duration of the signal abnormality is less than or equal to a preset signal duration threshold, it is determined to be short-term signal interference. Data for the abnormal period is extracted based on the timestamp information of the signal abnormality label. Wavelet denoising is used to process the data for the abnormal period. Wavelet denoising is a signal processing method based on wavelet transform. By decomposing the signal into sub-bands of different frequencies, it removes high-frequency noise and retains low-frequency signals, effectively removing sudden noise while preserving the main characteristics of the signal. The denoised data can more accurately reflect the actual operating status of the equipment. The denoised data is combined with normal data to obtain the repaired synchronous data stream.

[0057] This implementation distinguishes between persistent signal anomalies and short-term signal interference, enabling precise selection of different fault-tolerance strategies. Through differentiated fault-tolerance strategies, various signal anomalies can be effectively handled, reducing the impact of abnormal data on subsequent equipment anomaly detection and other processes, thereby improving data reliability and accuracy.

[0058] In one possible implementation, based on the data anomaly label, a differentiated fault-tolerance strategy is triggered to perform data fault-tolerance repair, resulting in a repaired synchronous data stream. Step S300 further includes step S350: if the data anomaly label is the transmission anomaly label, the transmission anomaly duration is extracted based on the timestamp information of the transmission anomaly label. Specifically, after detecting the transmission anomaly label, the start and end timestamps in the label are extracted, and the time difference between the two timestamps is calculated to obtain the transmission anomaly duration.

[0059] Step S360: If the duration of the transmission abnormality exceeds a preset transmission duration threshold, it is marked as a continuous transmission abnormality; otherwise, it is marked as short-term transmission interference. Specifically, a preset transmission duration threshold is set, for example, 5 seconds. The duration of the transmission abnormality is compared with the preset transmission duration threshold. If the duration of the transmission abnormality exceeds the preset transmission duration threshold, it is marked as a continuous transmission abnormality; otherwise, it is marked as short-term transmission interference.

[0060] Step S370: For the persistent transmission anomaly, a differentiated fault-tolerance strategy is triggered to activate the backup communication link. Simultaneously, Kalman smoothing prediction is performed to complete the data fault tolerance during the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream. Specifically, when a persistent transmission anomaly occurs on the primary communication link, the system automatically switches to the backup communication link. By activating the backup communication link, it is ensured that the system can continue to provide reliable data transmission even when the primary communication link experiences a persistent failure. The backup communication link can be a calibrated redundant network path or communication device. These links can serve as a temporary replacement when the primary communication link fails, thereby reducing the time of data transmission interruption and ensuring the continuity of production monitoring.

[0061] Based on the timestamp information of the transmission anomaly tags, the start and end times of the anomaly period are determined. The Kalman smoothing algorithm is then used to predict the data during the anomaly period. Kalman smoothing is a state-space model-based prediction method that can predict future data based on historical data and the current state. By minimizing the variance of the prediction error, it provides optimal prediction results to fill data gaps. The predicted data is then combined with normal data to obtain the repaired synchronized data stream.

[0062] Step S380: For the short-term transmission interference, a differentiated fault-tolerance strategy is triggered based on the decision matrix to perform data fault-tolerance repair for the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream. The decision matrix stores the correspondence between the transmission anomaly duration and the repair strategy. Specifically, when the transmission anomaly duration is less than or equal to a preset transmission duration threshold, it is determined to be short-term transmission interference. Data for the abnormal period is extracted based on the timestamp information of the transmission anomaly tag. A corresponding fault-tolerance strategy is selected from the decision matrix based on the transmission anomaly duration. The data for the abnormal period is repaired according to the selected fault-tolerance strategy. The repaired data is combined with normal data to obtain the repaired synchronous data stream. An example of the decision matrix is ​​shown in Table 1.

[0063] Table 1: Example of a decision matrix Transmission error duration (seconds) Repair strategy 0-2 Linear interpolation 2-5 Wavelet denoising >5 Kalman smoothing prediction This implementation distinguishes between persistent transmission anomalies and short-term transmission interference, enabling precise selection of different fault-tolerance strategies. Through differentiated fault-tolerance strategies, various transmission anomalies can be effectively handled, reducing the impact of abnormal data on subsequent equipment anomaly detection and other processes, thereby improving data reliability and accuracy.

[0064] In one possible implementation, based on the data anomaly label, a differentiated fault-tolerance strategy is triggered to perform data fault-tolerance repair, resulting in a repaired synchronous data stream. Step S300 further includes step S390: if the data anomaly label is the composite anomaly label, the composite anomaly duration is extracted based on the timestamp information of the composite anomaly label. Specifically, after detecting the composite anomaly label, the start and end timestamps in the label are extracted, and the time difference between the two timestamps is calculated to obtain the composite anomaly duration.

[0065] Step S3100: If the duration of the composite anomaly is greater than a preset composite duration threshold, it is marked as a persistent composite anomaly; otherwise, it is marked as a short-term composite interference. Specifically, a preset composite duration threshold is set, for example, 5 seconds. The duration of the composite anomaly is compared with the preset composite duration threshold. If the duration of the composite anomaly is greater than the preset composite duration threshold, it is marked as a persistent composite anomaly; otherwise, it is marked as a short-term composite interference.

[0066] Step S3110: For the persistent composite anomaly, a differentiated fault-tolerance strategy is triggered to acquire virtual sensing data through the digital twin system, and to complete the data during the abnormal period of the synchronous data stream, thereby obtaining a repaired synchronous data stream. Specifically, when the duration of the composite anomaly exceeds a preset composite duration threshold, it is determined to be a persistent composite anomaly. Virtual sensing data corresponding to the abnormal period is generated through the digital twin system. The digital twin system is a virtual model that can simulate and predict the behavior of physical devices by combining historical and real-time data. In the case of persistent composite anomalies, the main data source is unreliable, and the digital twin system can generate virtual sensing data based on historical data and model predictions to fill the data gaps. The virtual data generated by the digital twin system can ensure the continuity of the data stream, reduce data loss caused by anomalies, and provide complete data support for subsequent device anomaly detection. The virtual sensing data is combined with normal data to complete the data stream.

[0067] Step S3120: For the short-term composite interference, a differentiated fault-tolerant strategy is triggered to enable the anti-interference communication mode. Simultaneously, relevant normal sensor data fusion processing is performed to correct data redundancy during the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream. Specifically, when the composite abnormal duration is less than or equal to a preset composite duration threshold, it is determined to be short-term composite interference. The anti-interference communication mode is enabled to optimize the data transmission process and reduce the impact of interference on data transmission. The anti-interference communication mode improves the reliability of data transmission and reduces data errors caused by short-term interference by optimizing the communication protocol and adding redundancy checks.

[0068] The system extracts sensor data from normal data sources related to the abnormal data source. A data fusion algorithm is then used to combine this data with the relevant normal data source sensor data to correct the data from the abnormal period. Data fusion processing combines sensor data from multiple normal data sources, utilizing redundant information to correct the abnormal data. This method improves data accuracy and reliability, reducing data errors caused by a single data source failure. The corrected data is then combined with normal data to obtain a patched and synchronized data stream.

[0069] This implementation distinguishes between persistent composite anomalies and short-term composite disturbances, enabling precise selection of different fault-tolerance strategies. Through differentiated fault-tolerance strategies, various composite anomaly situations can be effectively handled, reducing the impact of abnormal data on subsequent equipment anomaly detection and other processes, thereby improving data reliability and accuracy.

[0070] In one possible implementation, the method further includes: periodically injecting simulated anomaly labels to test the performance of the differentiated fault tolerance strategy, and recording the fault tolerance strategy performance indicators corresponding to each label; when any performance indicator does not meet the preset fault tolerance performance threshold, generating a strategy optimization instruction to dynamically optimize the differentiated fault tolerance strategy.

[0071] Specifically, a test cycle is set, and within each test cycle, simulated anomaly tags are injected into the system, including signal anomaly tags, transmission anomaly tags, and composite anomaly tags. Corresponding fault-tolerance strategies are triggered, and the execution metrics of these strategies are recorded, such as fault tolerance time and data accuracy. Preset fault-tolerance performance thresholds are set, for example, fault tolerance time not exceeding 3 seconds and data accuracy not less than 95%. The execution metrics are compared with the preset fault-tolerance performance thresholds. If the execution metrics do not meet the preset fault-tolerance performance thresholds, a strategy optimization instruction is generated to dynamically optimize the fault-tolerance strategy.

[0072] This implementation method, through periodic injection of simulated exception labels for testing, can promptly identify shortcomings in the fault tolerance strategy, generate strategy optimization instructions in a timely manner, and dynamically optimize the fault tolerance strategy, thereby improving the adaptability and effectiveness of the fault tolerance strategy.

[0073] Step S400: Perform real-time device anomaly detection on the repair synchronization data stream to obtain the device health status.

[0074] Specifically, features related to equipment health, such as temperature, pressure, and vibration, are extracted from the repair synchronization data stream. Equipment anomaly detection algorithms are used to analyze these extracted features in real time to detect any anomalies. Based on the detected anomaly information and in conjunction with an equipment health assessment model, the equipment's health status is evaluated, categorized as normal, warning, or faulty. This equipment health status information is then output to the monitoring system for production management personnel to view and make decisions.

[0075] For example, characteristic data such as temperature, pressure, and vibration of the equipment can be extracted from the repair synchronization data stream. A Long Short-Term Memory (LSTM) network is used to analyze the equipment vibration data in real time to detect abnormal changes. Based on the degree and duration of the vibration anomaly, combined with an equipment health assessment model, the health status of the equipment is evaluated. The equipment health status information is then output to the monitoring system; for example, displaying that the health status of equipment M is "warning" and the health status of equipment N is "normal."

[0076] In one possible implementation, after real-time device anomaly detection is performed on the repair synchronization data stream to obtain the device health status, the method further includes: when the device health status indicates a high-risk fault, the digital twin system is linked to simulate the fault impact range; if the fault impact range is an independent production unit, the equipment operating parameters of the corresponding production unit are adjusted; if the fault impact range is a cross-unit collaborative production chain, a graded shutdown protection is triggered according to the criticality weight and an alarm is pushed to the mobile terminal.

[0077] Specifically, when the equipment health status indicates a high-risk failure, the fault simulation function of the digital twin system is triggered. Based on the parameters of the faulty equipment and the production process model, the digital twin system simulates the production process after the failure occurs to determine the specific scope of the failure's impact.

[0078] If the fault affects an independent production unit, the operating parameters of the affected unit's equipment are adjusted based on the fault type and production process, such as reducing production rate, to minimize the impact. Simultaneously, the adjusted equipment operating status is monitored in real-time to ensure the parameter adjustments are effective. If the fault affects a cross-unit collaborative production chain, a tiered shutdown protection mechanism is triggered based on criticality weights. Specifically, key factors affecting production are selected as evaluation indicators, such as the importance of equipment to the production process, the impact of equipment failure on product quality, and equipment redundancy. Criticality weights are assigned to each production unit or piece of equipment based on these evaluation indicators. These weights can be adjusted according to actual conditions; for example, critical equipment receives higher weights, while non-critical equipment receives lower weights. A shutdown strategy is determined based on the criticality weights. High-critical equipment is immediately shut down to ensure safety, while low-critical equipment enters a safe operating mode to reduce production interruption losses. In both shutdown and safe operating modes, equipment status is monitored in real-time to ensure operational safety and effectiveness. Simultaneously, alarm information, including fault details, impact scope, and recommended measures, is pushed to the mobile terminals of relevant personnel.

[0079] This approach uses a digital twin system to quickly simulate the impact of a failure, enabling timely assessment of its effects on the production process and providing a basis for rapid response. Based on criticality weights, critical equipment is immediately shut down to ensure safety, while non-critical equipment enters a safe operating mode to minimize production interruption losses.

[0080] This application utilizes edge computing nodes to perform timestamp alignment and format standardization on real-time acquired multi-source heterogeneous data streams, forming a synchronous data stream. Dynamic quality assessment is performed on the synchronous data stream; once an anomaly in the data source quality indicators is detected, a tag is generated. Based on the tag, a differentiated fault-tolerance strategy is triggered to repair the data, resulting in a repaired synchronous data stream. Real-time equipment anomaly detection and other technical means are then applied to the repaired synchronous data stream. This solves the technical problems of inaccurate equipment status monitoring and low levels of intelligent monitoring in existing production workshops, achieving the technical effect of improving the accuracy of equipment status monitoring and enhancing the level of intelligent monitoring in production workshops.

[0081] In the above text, refer to Figure 1 This paper describes in detail an intelligent monitoring method for production workshops based on industrial big data AI analysis according to embodiments of the present invention. Next, we will refer to... Figure 2 This invention describes an intelligent monitoring system for production workshops based on industrial big data AI analysis, according to an embodiment of the present invention.

[0082] The intelligent monitoring system for production workshops based on industrial big data AI analysis, according to embodiments of the present invention, addresses the technical problems of inaccurate equipment status monitoring and low levels of intelligent monitoring in existing production workshop intelligent monitoring systems. It aims to improve the accuracy of equipment status monitoring and enhance the level of intelligent monitoring in production workshops. The intelligent monitoring system for production workshops based on industrial big data AI analysis includes: a synchronous data stream acquisition module 10, a data anomaly detection module 20, a data fault tolerance and repair module 30, and a real-time equipment anomaly detection module 40.

[0083] The synchronous data stream acquisition module 10 is used to collect multi-source heterogeneous data streams from the production workshop in real time. It performs timestamp alignment and format standardization on the multi-source heterogeneous data streams through edge computing nodes to obtain a synchronous data stream. The data anomaly detection module 20 is used to perform dynamic quality assessment on the synchronous data stream. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated. The data fault tolerance and repair module 30 is used to trigger a differentiated fault tolerance strategy to perform data fault tolerance and repair based on the data anomaly label, obtaining a repaired synchronous data stream. The real-time equipment anomaly detection module 40 is used to perform real-time equipment anomaly detection on the repaired synchronous data stream to obtain the equipment health status.

[0084] The specific configuration of the data anomaly detection module 20 will be described in detail below. As mentioned above, the synchronous data stream undergoes dynamic quality assessment. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated. The data anomaly detection module 20 may further include: a dynamic quality assessment unit for performing signal-to-noise ratio (SNR) and transmission stability assessments on the synchronous data stream to obtain SNR and transmission stability index values; a signal quality anomaly determination unit for determining a signal quality anomaly when the SNR index value is less than a preset SNR threshold; a transmission anomaly determination unit for determining a transmission anomaly when the transmission stability index value does not meet a preset transmission stability threshold; and a data anomaly label generation unit for generating a data anomaly label when any data source is detected to have the signal quality anomaly and / or the transmission anomaly.

[0085] Specifically, when any data source is detected to have signal quality abnormalities and / or transmission abnormalities, a data abnormality tag is generated. The data abnormality tag generation unit may further include: a signal abnormality tag generation subunit for generating a signal abnormality tag when any data source is detected to have signal quality abnormalities; a transmission abnormality tag generation subunit for generating a transmission abnormality tag when any data source is detected to have transmission abnormalities; and a composite abnormality tag generation subunit for generating a composite abnormality tag when any data source is detected to have both signal quality abnormalities and transmission abnormalities.

[0086] The specific configuration of the data fault tolerance and repair module 30 will be described in detail below. As mentioned above, based on the data anomaly label, a differentiated fault tolerance strategy is triggered to perform data fault tolerance and repair, resulting in a repaired synchronous data stream. The data fault tolerance and repair module 30 may further include: a signal anomaly duration extraction unit, used to extract the signal anomaly duration based on the timestamp information of the signal anomaly label if the data anomaly label is the signal anomaly label; a signal anomaly marking unit, used to mark the signal anomaly as a persistent signal anomaly if the signal anomaly duration is greater than a preset signal duration threshold, otherwise, marked as short-term signal interference; a bidirectional linear interpolation unit, used to trigger a differentiated fault tolerance strategy to switch to a backup data source for the persistent signal anomaly, and simultaneously perform bidirectional linear interpolation to perform data fault tolerance interpolation for the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream, wherein the order of the bidirectional linear interpolation is dynamically determined based on the signal anomaly duration; and a wavelet denoising unit, used to trigger a differentiated fault tolerance strategy to perform wavelet denoising repair for the abnormal period of the synchronous data stream for the short-term signal interference, resulting in a repaired synchronous data stream.

[0087] The data fault tolerance repair module 30 further includes: a transmission anomaly duration extraction unit for extracting the transmission anomaly duration based on the timestamp information of the transmission anomaly label if the data anomaly label is the transmission anomaly label; a transmission anomaly marking unit for marking the transmission anomaly as a continuous transmission anomaly if the transmission anomaly duration exceeds a preset transmission duration threshold, otherwise marking it as a short-term transmission interference; a Kalman smoothing prediction unit for triggering the differentiated fault tolerance strategy to enable the backup communication link for the continuous transmission anomaly, and simultaneously performing Kalman smoothing prediction to complete the data fault tolerance during the abnormal period of the synchronous data stream, thus obtaining a repaired synchronous data stream; and a decision repair unit for triggering the differentiated fault tolerance strategy based on a decision matrix to repair the data fault tolerance during the abnormal period of the synchronous data stream for the short-term transmission interference, thus obtaining a repaired synchronous data stream, wherein the decision matrix stores the correspondence between the transmission anomaly duration and the repair strategy.

[0088] The data fault tolerance and repair module 30 further includes: a composite anomaly duration extraction unit for extracting the composite anomaly duration based on the timestamp information of the composite anomaly label if the data anomaly label is the composite anomaly label; a composite anomaly marking unit for marking the composite anomaly as a persistent composite anomaly if the composite anomaly duration exceeds a preset composite duration threshold, otherwise marking it as a short-term composite interference; a virtual sensor data acquisition unit for triggering a differentiated fault tolerance strategy to acquire virtual sensor data through a digital twin system for the persistent composite anomaly, performing data completion for the abnormal period of the synchronous data stream, and obtaining a repaired synchronous data stream; and a data redundancy correction unit for triggering a differentiated fault tolerance strategy to enable an anti-interference communication mode for the short-term composite interference, while simultaneously performing relevant normal sensor data fusion processing to correct data redundancy for the abnormal period of the synchronous data stream, and obtaining a repaired synchronous data stream.

[0089] The synchronous data stream undergoes dynamic quality assessment. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated. The data anomaly detection module 20 may further include: a sensor multimodal data constraint rule definition unit for predefining sensor multimodal data constraint rules for monitoring points based on a production process knowledge base; a constraint rule detection unit for performing sensor multimodal data constraint rule detection when the synchronous data stream does not trigger the signal quality anomaly and the transmission anomaly, and generating a logical conflict label if the constraint rules are violated; and a directional fault tolerance strategy triggering unit for extracting the logical conflict type based on the logical conflict label, and triggering a directional fault tolerance strategy to perform data fault tolerance repair based on the logical conflict type, thereby obtaining a repaired synchronous data stream.

[0090] The system may further include: a differentiated fault tolerance strategy execution effect testing module for periodically injecting simulated anomaly labels to test the execution effect of the differentiated fault tolerance strategy and recording the fault tolerance strategy execution indicators corresponding to each label; and a strategy optimization module for generating strategy optimization instructions to dynamically optimize the differentiated fault tolerance strategy when any execution indicator does not meet the preset fault tolerance performance threshold.

[0091] The system, after performing real-time equipment anomaly detection on the repair synchronization data stream and obtaining the equipment health status, may further include: a fault impact range simulation module for simulating the fault impact range in conjunction with a digital twin system when the equipment health status indicates a high-risk fault; an equipment operating parameter adjustment module for adjusting the equipment operating parameters of the corresponding production unit if the fault impact range is an independent production unit; and a graded shutdown protection triggering module for triggering graded shutdown protection and pushing alarms to mobile terminals according to criticality weights if the fault impact range is a cross-unit collaborative production chain.

[0092] The intelligent monitoring system for production workshops based on industrial big data AI analysis provided in this invention can execute the intelligent monitoring method for production workshops based on industrial big data AI analysis provided in any embodiment of this invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0093] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0094] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A production plant intelligent monitoring method based on industrial big data AI analysis, characterized in that, The method includes: Real-time acquisition of multi-source heterogeneous data streams from the production workshop; and time-stamp alignment and format standardization of the multi-source heterogeneous data streams through edge computing nodes to obtain synchronous data streams. The synchronous data stream is dynamically evaluated for quality. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated. Based on the data anomaly label, a differentiated fault tolerance strategy is triggered to perform data fault tolerance repair, resulting in a repaired synchronous data stream; Real-time device anomaly detection is performed on the repair synchronization data stream to obtain the device health status. 2.The production plant intelligent monitoring method based on industrial big data AI analysis according to claim 1, wherein, The synchronized data stream is dynamically quality-assessed. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated, including: The synchronous data stream is evaluated for signal-to-noise ratio and transmission stability to obtain signal-to-noise ratio and transmission stability index values. When the signal-to-noise ratio (SNR) value is less than the preset SNR threshold, it is determined that the signal quality is abnormal. When the transmission stability index value does not meet the preset transmission stability threshold, it is determined to be a transmission anomaly; When any data source is detected to have signal quality abnormalities and / or transmission abnormalities, a data abnormality tag is generated. 3.The production plant intelligent monitoring method based on industrial big data AI analysis according to claim 2, wherein, When any signal quality anomaly and / or transmission anomaly is detected in any data source, a data anomaly tag is generated, including: When any data source is detected to have an abnormal signal quality, a signal abnormality label is generated. When a transmission anomaly is detected in any data source, a transmission anomaly tag is generated; When both the signal quality abnormality and the transmission abnormality are detected simultaneously from any data source, a composite abnormality label is generated. 4.The production plant intelligent monitoring method based on industrial big data AI analysis according to claim 3, wherein, Based on the aforementioned data anomaly labels, a differentiated fault tolerance strategy is triggered to perform data fault tolerance repair, resulting in a repaired synchronous data stream, including: If the data anomaly label is the signal anomaly label, extract the signal anomaly duration based on the timestamp information of the signal anomaly label; If the duration of the abnormal signal exceeds a preset signal duration threshold, it is marked as a continuous signal abnormality; otherwise, it is marked as short-term signal interference. In response to the persistent signal anomaly, a differentiated fault-tolerance strategy is triggered to switch to a backup data source. At the same time, bidirectional linear interpolation is performed to perform data fault-tolerance interpolation during the abnormal period of the synchronous data stream, thereby obtaining a repaired synchronous data stream. The order of the bidirectional linear interpolation is dynamically determined according to the duration of the signal anomaly. To address the short-term signal interference, a differentiated fault-tolerance strategy is triggered to perform wavelet denoising and repair of the synchronous data stream during the abnormal period, thereby obtaining a repaired synchronous data stream. 5.The production plant intelligent monitoring method based on industrial big data AI analysis according to claim 3, wherein, Based on the aforementioned data anomaly labels, a differentiated fault tolerance strategy is triggered to perform data fault tolerance repair, resulting in a repaired synchronous data stream, including: If the data anomaly label is the transmission anomaly label, extract the transmission anomaly duration based on the timestamp information of the transmission anomaly label; If the duration of the transmission anomaly exceeds a preset transmission duration threshold, it is marked as a continuous transmission anomaly; otherwise, it is marked as a short-term transmission interference. In response to the persistent transmission anomaly, a differentiated fault-tolerance strategy is triggered to activate the backup communication link. At the same time, Kalman smoothing prediction is performed to perform data fault-tolerance completion during the abnormal period of the synchronous data stream, thereby obtaining a repaired synchronous data stream. To address the short-term transmission interference, a differentiated fault-tolerance strategy is triggered based on a decision matrix to perform data fault-tolerance repair during the abnormal period of the synchronous data stream, resulting in a repaired synchronous data stream. The decision matrix stores the correspondence between the duration of the transmission anomaly and the repair strategy. 6.The production plant intelligent monitoring method based on industrial big data AI analysis according to claim 3, wherein, Based on the aforementioned data anomaly labels, a differentiated fault tolerance strategy is triggered to perform data fault tolerance repair, resulting in a repaired synchronous data stream, including: If the data anomaly label is the composite anomaly label, extract the composite anomaly duration based on the timestamp information of the composite anomaly label; If the duration of the composite anomaly exceeds a preset composite duration threshold, it is marked as a persistent composite anomaly; otherwise, it is marked as a short-term composite interference. In response to the persistent composite anomaly, a differentiated fault-tolerance strategy is triggered to acquire virtual sensing data through a digital twin system, and to complete the data during the abnormal period of the synchronous data stream to obtain a repaired synchronous data stream. In response to the short-term composite interference, a differentiated fault-tolerant strategy is triggered to enable the anti-interference communication mode. At the same time, relevant normal sensor data fusion processing is performed to correct the data redundancy during the abnormal period of the synchronous data stream, thereby obtaining a repaired synchronous data stream.

7. The intelligent monitoring method for production workshops based on industrial big data AI analysis as described in claim 2, characterized in that, The synchronous data stream undergoes dynamic quality assessment. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated. The process also includes: Based on the production process knowledge base, predefined sensor multimodal data constraint rules for monitoring points are established. When the synchronous data stream does not trigger the signal quality abnormality and the transmission abnormality, the sensor multimodal data constraint rule detection is performed. If the constraint rule is violated, a logical conflict label is generated. Based on the logical conflict label, the logical conflict type is extracted, and a targeted fault tolerance strategy is triggered based on the logical conflict type to perform data fault tolerance repair, thereby obtaining a repaired synchronous data stream.

8. The intelligent monitoring method for production workshops based on industrial big data AI analysis as described in claim 1, characterized in that, Also includes: Periodically inject simulated anomaly tags to test the effectiveness of differentiated fault tolerance strategies, and record the corresponding fault tolerance strategy execution metrics for each tag; When any execution metric fails to meet the preset fault tolerance performance threshold, a strategy optimization instruction is generated to dynamically optimize the differentiated fault tolerance strategy.

9. The intelligent monitoring method for production workshops based on industrial big data AI analysis as described in claim 1, characterized in that, After performing real-time device anomaly detection on the repair synchronization data stream and obtaining the device health status, the process further includes: When the health status of the device indicates a high-risk fault, the linked digital twin system simulates the scope of the fault's impact. If the scope of the fault affects an independent production unit, then adjust the equipment operating parameters of the corresponding production unit; If the fault affects a cross-unit collaborative production chain, a graded shutdown protection will be triggered based on the criticality weight, and an alarm will be pushed to the mobile terminal.

10. A production workshop intelligent monitoring system based on industrial big data AI analysis, characterized in that, The system is used to implement the intelligent monitoring method for production workshops based on industrial big data AI analysis as described in any one of claims 1-9, and the system includes: The synchronous data stream acquisition module is used to collect multi-source heterogeneous data streams from the production workshop in real time. The multi-source heterogeneous data streams are timestamped and format standardized by edge computing nodes to obtain synchronous data streams. The data anomaly detection module is used to perform dynamic quality assessment on the synchronous data stream. When any quality indicator of any data source is detected to be abnormal, a data anomaly label is generated. The data fault tolerance and repair module is used to trigger a differentiated fault tolerance strategy to perform data fault tolerance and repair based on the data anomaly label, and obtain a repaired synchronous data stream; The real-time device anomaly detection module is used to perform real-time device anomaly detection on the repair synchronization data stream and obtain the device health status.

Images