Automatic production line multi-parameter real-time monitoring and fault diagnosis early warning method and system

Abstract

The invention provides an automatic production line multi-parameter real-time monitoring and fault diagnosis early warning method and system, and relates to the technical field of automatic production line monitoring, and the method comprises the steps: obtaining the data of a multi-source sensor, carrying out the data quality evaluation and correction, carrying out the monitoring of the performance parameters of an equipment assembly, and calculating the degradation rate; fault types are analyzed and identified in combination with time-frequency domain features, and a fault propagation link is established based on a physical connection relation to determine a fault source. According to the invention, equipment abnormity can be found in time, the fault source component is accurately positioned, the accuracy and timeliness of fault diagnosis are improved, and the equipment maintenance cost is effectively reduced.

Description

Technical Field

[0001] The present invention relates to automated production line monitoring technology, and in particular to an automated production line multi-parameter real-time monitoring and fault diagnosis and early warning method and system. Background Art

[0002] With the continuous improvement of industrial automation level, automated production lines are widely used in manufacturing. In order to ensure the stable operation of the production line, multi-source sensors are usually used to monitor key equipment and process parameters in real time. However, due to factors such as sensor failure and abnormal data transmission, the reliability of collected data is difficult to guarantee, which affects the accuracy of fault diagnosis.

[0003] Existing fault diagnosis methods are mainly based on single equipment or local parameters for analysis, which cannot effectively identify the fault correlation and propagation characteristics between equipment components. At the same time, traditional methods have problems such as incomplete feature extraction and inaccurate diagnosis results when processing multi-source heterogeneous data, making it difficult to meet the fault diagnosis needs in complex industrial scenarios.

[0004] In order to solve the fault diagnosis problem of equipment components in automated production lines, there is an urgent need for a comprehensive solution that can perform reliability assessment and correction on multi-source sensor data, analyze the fault propagation link based on the physical connection relationship of equipment components, and accurately locate the fault source and identify the fault type. Summary of the invention

[0005] The embodiment of the present invention provides a method and system for real-time monitoring of multiple parameters and fault diagnosis and early warning of an automated production line, which can solve the problems in the prior art.

[0006] According to a first aspect of the embodiments of the present invention, A method for real-time monitoring of multiple parameters and fault diagnosis and early warning of an automated production line is provided, comprising: Acquire real-time operation data collected by multi-source sensors in the automated production line, calculate data quality characteristic values, generate data credibility, and when the data credibility is less than a first preset threshold, perform data correction based on historical data at adjacent moments to obtain corrected real-time operation data; The corrected real-time operation data is grouped according to the equipment components, an operation parameter sequence corresponding to the equipment components is generated, performance parameters of the equipment components are calculated based on the operation parameter sequence, and a performance degradation rate is calculated according to the performance parameters. When the performance degradation rate is greater than a second preset threshold, the corresponding standby component is determined as a component to be monitored; Performing time-frequency domain feature analysis on the operating parameter sequence of the component to be monitored, extracting time-domain statistical features and frequency-domain energy features as fault features, inputting the fault features into a pre-trained fault recognition model, outputting a probability distribution of fault types based on a combination pattern of the fault features, and selecting the fault type with the highest probability as the fault determination result of the component to be monitored; According to the physical connection relationship between the component to be monitored and the adjacent components, the performance degradation rate of the adjacent components is obtained, and a fault propagation link is established based on the time series change sequence of the performance degradation rate. The fault source component is determined, and a fault diagnosis report containing the location information of the fault source component and the fault judgment result is generated and sent to the production management system.

[0007] In an optional embodiment, Acquire real-time operation data collected by multi-source sensors in the automated production line, calculate data quality characteristic values, generate data credibility, and when the data credibility is less than a first preset threshold, perform data correction based on historical data at adjacent moments, and obtain corrected real-time operation data including: Acquire the real-time operation data collected by multi-source sensors in the automated production line and segment it according to the time window to obtain a segmented data sequence; The signal fluctuation, value variation and sampling integrity of the segmented data sequence are calculated respectively and combined to obtain the data quality characteristic value, the information entropy of the data quality characteristic value is calculated based on the probability distribution characteristics of the data in the segmented data sequence, a dynamic weight coefficient is generated according to the information entropy, and the dynamic weight coefficient is weightedly combined with the corresponding data quality characteristic value to obtain the data credibility; When the data credibility is less than a first preset threshold, the data to be corrected is determined, and a historical data sequence of adjacent moments of N time windows before and after the time window corresponding to the data to be corrected is extracted; Clustering the historical data sequences at adjacent moments according to the similarity of data change trends to obtain multiple subsequence groups, calculating the time correlation between each subsequence in each subsequence group and the data to be corrected, and constructing a recursive optimization model with differentiated weights based on the time correlation, wherein the weight is positively correlated with the time correlation; A recursive optimization model is used to iteratively correct the data to be corrected. In each round of iteration, the fitting error between the correction result and the subsequence whose correlation is greater than a preset correlation threshold is calculated. When the fitting error is less than a dynamic threshold, the correction result is output as the corrected real-time running data.

[0008] In an optional embodiment, Calculating the information entropy of the data quality characteristic value based on the probability distribution characteristics of the data in the segmented data sequence, and generating a dynamic weight coefficient according to the information entropy includes: Constructing a probability density estimation function, calculating the standard deviation and the interquartile range based on the segmented data sequence, dynamically adjusting the bandwidth coefficient of the probability density estimation function according to the ratio of the standard deviation and the interquartile range, and obtaining a continuous probability density curve of the segmented data sequence; Performing piecewise integration on the continuous probability density curve to obtain a cumulative probability value of each segment of data, dividing the segmented data sequence into a first distribution area and a second distribution area according to the cumulative probability value, and calculating a cumulative probability ratio of the first distribution area to the second distribution area as a data distribution characteristic ratio; The overlapping sliding window step length is set by using the data distribution characteristic ratio, and the first local information entropy is obtained by sliding calculation in the first distribution area; the non-overlapping sliding window width is set by the inverse of the data distribution characteristic ratio, and the second local information entropy is obtained by calculation in the second distribution area; Taking the cumulative probability values ​​of the first distribution area and the second distribution area as weight coefficients, weighting the first local information entropy and the second local information entropy to obtain a regional combined entropy; The global information entropy is calculated for the continuous probability density curve, the regional combined entropy and the global information entropy are normalized to obtain the final information entropy of the data quality characteristic value, and a dynamic weight coefficient is generated according to the final information entropy, and the dynamic weight coefficient is inversely proportional to the final information entropy.

[0009] In an optional embodiment, Calculating the performance parameters of the equipment components based on the operation parameter sequence, calculating the performance degradation rate according to the performance parameters, and when the performance degradation rate is greater than a second preset threshold, determining the corresponding standby component as a component to be monitored includes: Performing time-frequency analysis on the operating parameter sequence to obtain energy distribution characteristics, selecting a characteristic parameter combination based on the energy distribution characteristics, and constructing the characteristic parameter combination as a comprehensive performance status indicator; Analyze the temporal variation law of the comprehensive performance status index, establish a performance status transfer matrix, calculate the change rate of the performance parameter based on the performance status transfer matrix, and correct the change rate in combination with the performance status transfer matrix to obtain the performance degradation rate; Establish a performance baseline value for when the component is operating normally, determine the performance fluctuation tolerance range based on the performance baseline value combined with historical operating data, start the cumulative time count when it is detected that the performance degradation rate exceeds the performance fluctuation tolerance range, and when the cumulative time reaches a preset period, determine the corresponding equipment component as the component to be monitored.

[0010] In an optional embodiment, Establishing a performance benchmark value for a component when it is operating normally, and determining a performance fluctuation tolerance range based on the performance benchmark value combined with historical operating data include: Acquire historical operation data of the equipment components, identify normal operation periods and abnormal operation periods from the historical operation data, compare and analyze the normal operation periods and abnormal operation periods, extract false positive information and missed negative information, and construct a reward function; Inputting the reward function and the performance data of the normal operation period into a reinforcement learning framework, obtaining a performance benchmark value through reinforcement learning training, and generating an initial tolerance adjustment strategy based on the performance benchmark value; Collect the real-time operation status data of the components, match and analyze the real-time operation status data with the initial tolerance adjustment strategy, calculate the status warning probability under different tolerance settings, and generate risk prediction results within the tolerance range; Obtaining the position information of the component in the production process, the process parameter requirements of the process, and the historical maintenance cost data, performing multi-objective optimization calculation on the position information, process parameter requirements, and historical maintenance cost data and the risk prediction results to generate a tolerance optimization objective function; The initial tolerance adjustment strategy is optimized based on the tolerance optimization objective function to obtain a performance fluctuation tolerance range, the performance fluctuation tolerance range is combined with a performance benchmark value to form a monitoring threshold, and the state data corresponding to the monitoring threshold is fed back to the reinforcement learning framework for continuous optimization training.

[0011] In an optional embodiment, The fault characteristics are input into the pre-trained fault identification model, and the fault type probability distribution is output based on the combination mode of the fault characteristics. The fault type with the highest probability is selected as the fault judgment result of the component to be monitored, including: Inputting the fault features into a pre-trained fault recognition model, decomposing the fault features into common features and individual features through a feature decoupling network, and applying orthogonal constraints to the common features and individual features to obtain decoupled features; The decoupled features are constructed as an attribute graph, the feature dimension is set as a graph node, the feature association strength is set as a graph edge, and local dependencies and global dependencies are extracted through graph convolution and multi-head attention mechanism to obtain a combination pattern of fault features; Inputting the combination mode of fault features and decoupled features into a variational inference module, establishing a state transfer matrix based on the feature association strength and orthogonal constraints, and obtaining an initial fault type probability; The initial fault type probability is modeled in a Bayesian manner, and probability calibration is performed based on the local dependencies and global dependencies in the combination pattern of fault characteristics to generate a calibrated fault type probability distribution. The fault type with the highest probability is selected from the calibrated fault type probability distribution as the fault judgment result of the component to be monitored.

[0012] In an optional embodiment, According to the physical connection relationship between the component to be monitored and the adjacent components, the performance degradation rate of the adjacent components is obtained, and the fault propagation link is established based on the time sequence of the performance degradation rate. The fault source components are determined to include: Obtain the mechanical coupling relationship, energy transfer relationship and physical topological relationship between the component to be monitored and the adjacent components, and establish a component connection relationship matrix that characterizes the spatial connection strength between components; Acquiring performance parameters of the component to be monitored and adjacent components within a continuous time window, calculating the time derivative of the performance parameters to obtain a performance degradation rate, and performing a Hilbert transform on the performance degradation rate to obtain an instantaneous phase; According to the change trend of the spatial connection strength and instantaneous phase in the component connection relationship matrix, the phase synchronization strength between adjacent components is calculated, and a network order parameter reflecting the dynamic coupling relationship between components is constructed; Monitor the dynamic changes of the network order parameters, obtain the mutation time and mutation amplitude of the network order parameters, and construct a directed fault propagation graph based on the mutation time, mutation amplitude and phase synchronization strength, in which adjacent components are nodes, the time difference of the mutation time is the edge weight, and the phase synchronization strength is the edge direction; Calculating the downstream impact range of each component according to the edge weight and edge direction in the directed fault propagation graph, determining the cascading failure scale of each component based on the downstream impact range, and combining the cascading failure scale with the mutation moment to obtain a failure impact factor; According to the dynamic evolution law of the directed fault propagation graph, the mutation time characteristics and cascading failure scale in the failure influence factor are weighted, and the component with the largest configured failure influence factor and the earliest mutation time is determined as the fault source component.

[0013] According to a second aspect of the embodiments of the present invention, Provide an automated production line multi-parameter real-time monitoring and fault diagnosis early warning system, including: The first unit is used to obtain real-time operation data collected by multi-source sensors in the automated production line, calculate data quality characteristic values, generate data credibility, and when the data credibility is less than a first preset threshold, perform data correction based on historical data at adjacent moments to obtain corrected real-time operation data; The second unit is used to group the corrected real-time operation data according to the equipment components, generate an operation parameter sequence corresponding to the equipment components, calculate the performance parameters of the equipment components based on the operation parameter sequence, calculate the performance degradation rate according to the performance parameters, and when the performance degradation rate is greater than a second preset threshold, determine the corresponding standby component as a component to be monitored; The third unit is used to perform time-frequency domain feature analysis on the operating parameter sequence of the component to be monitored, extract time-domain statistical features and frequency-domain energy features as fault features, input the fault features into a pre-trained fault recognition model, output a fault type probability distribution based on a combination mode of the fault features, and select the fault type with the highest probability as the fault determination result of the component to be monitored; The fourth unit is used to obtain the performance degradation rate of the adjacent components according to the physical connection relationship between the component to be monitored and the adjacent components, establish a fault propagation link based on the time series change sequence of the performance degradation rate, determine the fault source component, generate a fault diagnosis report containing the location information of the fault source component and the fault judgment result, and send it to the production management system.

[0014] According to a third aspect of the embodiments of the present invention, An electronic device is provided, comprising: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0015] According to a fourth aspect of the embodiments of the present invention, A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0016] In this embodiment, by performing quality assessment and correction on the data collected by multi-source sensors of the automated production line, the reliability of the data is ensured, misjudgment due to data anomalies is avoided, and the accuracy and credibility of fault diagnosis are improved. Fault monitoring is performed based on the performance parameters and degradation rates of equipment components, and combined with the time-frequency domain feature analysis method, fault features are extracted from multiple dimensions. Accurate identification of fault types is achieved through pre-trained fault identification models, which improves the efficiency and accuracy of fault diagnosis. By analyzing the physical connection relationship between equipment components and the temporal variation law of performance degradation rate, a fault propagation link is established, the fault source is accurately located, and misjudgment due to fault propagation is effectively avoided. A reliable decision-making basis is provided for equipment maintenance and fault handling, and the operational reliability and maintenance efficiency of the production line are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of a method for real-time monitoring of multiple parameters and fault diagnosis and early warning of an automated production line according to an embodiment of the present invention; Figure 2 A schematic diagram of probability distribution area division according to an embodiment of the present invention; Figure 3 This is a comparison chart of early warning accuracy indicators of different methods in embodiments of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0020] Figure 1 FIG. 1 is a flow chart of a method for real-time monitoring of multiple parameters and fault diagnosis and early warning of an automated production line according to an embodiment of the present invention. Figure 1 As shown, the method includes: Acquire real-time operation data collected by multi-source sensors in the automated production line, calculate data quality characteristic values, generate data credibility, and when the data credibility is less than a first preset threshold, perform data correction based on historical data at adjacent moments to obtain corrected real-time operation data; The corrected real-time operation data is grouped according to the equipment components, an operation parameter sequence corresponding to the equipment components is generated, performance parameters of the equipment components are calculated based on the operation parameter sequence, and a performance degradation rate is calculated according to the performance parameters. When the performance degradation rate is greater than a second preset threshold, the corresponding standby component is determined as a component to be monitored; Performing time-frequency domain feature analysis on the operating parameter sequence of the component to be monitored, extracting time-domain statistical features and frequency-domain energy features as fault features, inputting the fault features into a pre-trained fault recognition model, outputting a probability distribution of fault types based on a combination pattern of the fault features, and selecting the fault type with the highest probability as the fault determination result of the component to be monitored; According to the physical connection relationship between the component to be monitored and the adjacent components, the performance degradation rate of the adjacent components is obtained, and a fault propagation link is established based on the time series change sequence of the performance degradation rate. The fault source component is determined, and a fault diagnosis report containing the location information of the fault source component and the fault judgment result is generated and sent to the production management system.

[0021] In an optional embodiment, Acquire real-time operation data collected by multi-source sensors in the automated production line, calculate data quality characteristic values, generate data credibility, and when the data credibility is less than a first preset threshold, perform data correction based on historical data at adjacent moments, and obtain corrected real-time operation data including: Acquire the real-time operation data collected by multi-source sensors in the automated production line and segment it according to the time window to obtain a segmented data sequence; The signal fluctuation, value variation and sampling integrity of the segmented data sequence are calculated respectively and combined to obtain the data quality characteristic value, the information entropy of the data quality characteristic value is calculated based on the probability distribution characteristics of the data in the segmented data sequence, a dynamic weight coefficient is generated according to the information entropy, and the dynamic weight coefficient is weightedly combined with the corresponding data quality characteristic value to obtain the data credibility; When the data credibility is less than a first preset threshold, the data to be corrected is determined, and a historical data sequence of adjacent moments of N time windows before and after the time window corresponding to the data to be corrected is extracted; Clustering the historical data sequences at adjacent moments according to the similarity of data change trends to obtain multiple subsequence groups, calculating the time correlation between each subsequence in each subsequence group and the data to be corrected, and constructing a recursive optimization model with differentiated weights based on the time correlation, wherein the weight is positively correlated with the time correlation; A recursive optimization model is used to iteratively correct the data to be corrected. In each round of iteration, the fitting error between the correction result and the subsequence whose correlation is greater than a preset correlation threshold is calculated. When the fitting error is less than a dynamic threshold, the correction result is output as the corrected real-time running data.

[0022] For example, in the process of fault diagnosis of an automated production line, the real-time operation data collected by multi-source sensors must first be preprocessed and quality evaluated. The data is segmented by setting a fixed-length time window, and the length of the time window is determined according to the dynamic response characteristics of the equipment components. For example, for a robot welding station, considering the integrity of the welding process, the time window can be set to a welding cycle to ensure that the segmented data sequence can fully reflect the changes in the operating status of the equipment.

[0023] The calculation of data quality characteristic values ​​includes three aspects: signal volatility reflects the stability of the data sequence, which is specifically characterized by calculating the degree of discreteness of all values ​​in the data sequence, that is, first calculating the average value of the entire sequence, then calculating the deviation of each data point from the average value, and finally obtaining the ratio of the standard deviation to the mean; numerical variability represents the dynamic change characteristics of the data, which is measured by calculating the numerical difference between adjacent sampling points, specifically calculating the difference between two adjacent data points, and then calculating the root mean square value of the difference sequence; sampling completeness reflects the integrity of the data, which is measured by the ratio of the actual number of sampling points to the number of sampling points that should be obtained according to the sampling frequency theory. The closer the value is to one, the more complete the data is.

[0024] In order to achieve a reasonable assessment of data quality characteristics, a dynamic weight allocation mechanism based on information entropy is introduced. First, the segmented data sequence is statistically analyzed, the data value range is divided into several equally spaced intervals, the frequency of occurrence of data points in each interval is counted, and the probability distribution is calculated. On this basis, the information entropy of the data distribution is calculated. The information entropy reflects the uncertainty of the data distribution. The calculation process is to substitute the probability value of each interval into the information entropy calculation formula. A larger information entropy value indicates that the data distribution is more dispersed and the uncertainty is high, and the weight of the corresponding feature should be reduced.

[0025] In the process of generating dynamic weight coefficients based on information entropy values, first calculate the information entropy values ​​corresponding to each quality feature, normalize the information entropy values, then subtract the normalized information entropy values ​​from 1 to get the initial weights, and finally normalize the initial weights to get the final dynamic weight coefficients. Multiply the dynamic weight coefficients with the corresponding data quality feature values ​​and sum them to get a comprehensive data credibility index.

[0026] When the data credibility is lower than the preset threshold, the data needs to be corrected. First, determine the data segment to be corrected, and extract the historical data of the time window before and after the data segment. The amount of historical data to be extracted is determined according to the periodic characteristics of the equipment operation, and usually a time range that can cover the complete operation cycle of the equipment is selected.

[0027] When clustering the extracted historical data series, it is necessary to extract the trend characteristics of the data. The trend characteristics include the distribution of the rising, falling and stable segments of the data, as well as the slope characteristics of each segment. Based on these characteristics, the similarity between different data series is calculated. The similarity calculation takes into account the matching degree and amplitude difference of the trend characteristics. Then, a density-based clustering algorithm is used to divide the sequences with high similarity into the same group.

[0028] When calculating the time correlation between the subsequence and the data to be corrected, a weighted method based on time decay is used. Specifically, the time interval between the subsequence and the data to be corrected is calculated, and the time interval is substituted into the decreasing function to obtain the time weight. The larger the time interval, the smaller the weight. At the same time, the trend similarity between the subsequence and the data to be corrected is calculated, and the final time correlation is obtained by multiplying the time weight and the trend similarity.

[0029] When constructing a recursive optimization model, the time correlation is used as the weight coefficient of each subsequence reference value. The optimization goal of the model is to ensure that the corrected data not only maintains a consistent trend with the highly correlated historical data, but also meets the physical constraints of the data. In each round of iteration, the weighted error is calculated based on the current corrected value and each reference sequence, and at the same time, it is checked whether the corrected value meets the constraints such as the value range and the rate of change.

[0030] The determination of the dynamic threshold takes into account the fluctuation characteristics of the data and is set by calculating the data fluctuation range of the highly correlated subsequence group. When the fitting error between the correction result and the highly correlated subsequence is less than the dynamic threshold, and the correction value changes in multiple consecutive iterations are small, the correction process is considered to have converged, and the current correction result is output as the final correction data.

[0031] In this embodiment, real-time quality assessment and dynamic correction of multi-source sensor data are realized, which improves the accuracy and reliability of automated production line data. Data quality is comprehensively evaluated by signal volatility, value change and sampling integrity, and information entropy is introduced to dynamically adjust the weights of each feature to ensure that the evaluation results are adapted to different data distribution characteristics. For low-credibility data, the trend characteristics of historical data sequences are used for clustering, and a recursive optimization model is constructed in combination with time correlation to ensure that the correction results fit the actual operating status, reduce errors and noise interference, and effectively improve the stability and accuracy of fault diagnosis and process control.

[0032] In an optional embodiment, Calculating the information entropy of the data quality characteristic value based on the probability distribution characteristics of the data in the segmented data sequence, and generating a dynamic weight coefficient according to the information entropy includes: Constructing a probability density estimation function, calculating the standard deviation and the interquartile range based on the segmented data sequence, dynamically adjusting the bandwidth coefficient of the probability density estimation function according to the ratio of the standard deviation and the interquartile range, and obtaining a continuous probability density curve of the segmented data sequence; Performing piecewise integration on the continuous probability density curve to obtain a cumulative probability value of each segment of data, dividing the segmented data sequence into a first distribution area and a second distribution area according to the cumulative probability value, and calculating a cumulative probability ratio of the first distribution area to the second distribution area as a data distribution characteristic ratio; The overlapping sliding window step length is set by using the data distribution characteristic ratio, and the first local information entropy is obtained by sliding calculation in the first distribution area; the non-overlapping sliding window width is set by the inverse of the data distribution characteristic ratio, and the second local information entropy is obtained by calculation in the second distribution area; Taking the cumulative probability values ​​of the first distribution area and the second distribution area as weight coefficients, weighting the first local information entropy and the second local information entropy to obtain a regional combined entropy; The global information entropy is calculated for the continuous probability density curve, the regional combined entropy and the global information entropy are normalized to obtain the final information entropy of the data quality characteristic value, and a dynamic weight coefficient is generated according to the final information entropy, and the dynamic weight coefficient is inversely proportional to the final information entropy.

[0033] Exemplarily, a kernel density estimation function is first constructed to evaluate the data distribution characteristics. For each segmented data sequence collected by each sensor, such as the speed data of the robot joint, the current data of the motor, etc., its statistical characteristics are calculated. The standard deviation is calculated by summing the square root of the difference between each value in the data sequence and the mean value; the interquartile range is to sort the data sequence by size and find the difference between the 75th percentile and the 25th percentile. The standard deviation is divided by the interquartile range to obtain a ratio, which is used to adjust the bandwidth coefficient of the kernel density estimation function. When the ratio is greater than two, it indicates that the overall fluctuation of the data is significantly greater than the fluctuation in the middle area, and the bandwidth coefficient needs to be increased by 50%; when the ratio is less than 0.5, it indicates that the data distribution is concentrated, and the bandwidth coefficient is reduced by 30%.

[0034] The adjusted bandwidth coefficient is used to perform kernel density estimation on the data sequence to obtain a continuous probability density curve that reflects the data distribution. The probability density curve represents the density of data at different value locations. The higher the curve, the denser the data. The entire data value range is divided into multiple small intervals, and the probability density curve is definite integrated in each interval to obtain the cumulative probability value of the interval.

[0035] The data sequence is divided into regions based on the cumulative probability value. The cumulative probability values ​​are sorted from large to small, and the interval set with a sum of probability values ​​reaching 70% is divided into the first distribution area, indicating the main distribution area of ​​the data; the remaining intervals are divided into the second distribution area, indicating the sparse distribution area of ​​the data. The ratio of the cumulative probability value of the first distribution area to the cumulative probability value of the second distribution area is calculated to obtain the data distribution characteristic ratio.

[0036] The local information entropy is calculated using overlapping sliding windows in the first distribution area. The window step size is set to the product of the data distribution feature ratio and the basic step size, and the basic step size is 5% of the data sequence length. For example, when the data distribution feature ratio is 2.5, the sliding step size is 2.5 times the basic step size. In each sliding window, the probability distribution of the data points is counted and the first local information entropy is calculated.

[0037] For the second distribution area, a non-overlapping sliding window is used. The window width is set to the length of the data sequence divided by the data distribution feature ratio to ensure that the sparse area is reasonably divided. For example, when the data distribution feature ratio is two, the window width is half of the data sequence length. The probability distribution of the data is calculated in each window to obtain the second local information entropy.

[0038] The first local information entropy and the second local information entropy are weighted and combined. The weight coefficient adopts the cumulative probability value of each region. For example, if the cumulative probability value of the first distribution region is 70%, the weight of the first local information entropy is 0.7. The two weighted entropy values ​​are added together to obtain the regional combined entropy.

[0039] The global information entropy is calculated for the entire probability density curve, which covers the entire range of values ​​of the data sequence. The regional combination entropy and the global information entropy are normalized by dividing them by their maximum possible values, ensuring that the two entropy values ​​are in the range of zero to one. The normalized regional combination entropy and the global information entropy are combined in a ratio of six to four to obtain the final information entropy.

[0040] Finally, the final information entropy is converted into a dynamic weight coefficient. The weight coefficient is inversely proportional to the information entropy by subtracting the final information entropy from one and then normalizing. For example, when the final information entropy of a data quality feature is 0.8, its corresponding initial weight is 0.2. Then the initial weights of all features are normalized to obtain the final dynamic weight coefficient. These weight coefficients are used to calculate the credibility of the data in the subsequent calculation, providing an objective basis for data quality assessment.

[0041] Figure 2 Schematic diagram of probability distribution area division according to an embodiment of the present invention, as shown in FIG. Figure 2 As shown in the figure, the regional division and information entropy calculation strategy for bimodal distribution data are shown. The horizontal axis in the figure represents the data value (0-125), and the vertical axis represents the probability density (0-0.04). It can be observed that the data presents obvious bimodal distribution characteristics, with probability density peaks around 25 and 75 respectively. According to the regional division principle in the proposed method, the system divides the data into five regions: two first distribution regions (located around data values ​​25 and 75) and three second distribution regions (located in the range of data values ​​0-15, 35-65, and 90-125). Through the piecewise integration of the probability density curve, it is calculated that the cumulative probabilities of the first region are 30% and 40%, respectively, totaling 70%; the cumulative probabilities of the second region are 5%, 15% and 10%, respectively, totaling 30%. Different information entropy calculation strategies are adopted for different distribution regions: the overlapping sliding window method is used in the first distribution region (data dense area) to realize refined local information entropy calculation; the non-overlapping window method is used in the second distribution region (data sparse area) to ensure that effective probability distribution characteristics can be obtained even when the data is sparse. This adaptive region division and calculation strategy based on data distribution characteristics fully considers the characteristics of complex distribution data and can accurately capture the information entropy characteristics of each region. Compared with the traditional global information entropy calculation method, it has obvious advantages, especially for unconventional distribution data such as multi-peak distribution and skewed distribution.

[0042] In the prior art, the quality assessment of multi-source sensor data usually combines multiple feature indicators in a fixed weighted manner, without considering the impact of data distribution characteristics on quality assessment, resulting in inaccurate assessment results. At the same time, in the process of data correction, historical data at adjacent moments are often directly used for interpolation or fitting, which does not fully utilize the time correlation characteristics of the data, and the correction effect is greatly affected by abnormal data. This application introduces a dynamic weight allocation mechanism based on data distribution characteristics, adaptively adjusts the bandwidth coefficient of kernel density estimation according to the probability distribution characteristics of the data, and combines regional information entropy and global information entropy to make data quality assessment more objective and reasonable. When correcting data, a recursive optimization model considering time correlation is adopted, and the reliability of the correction result is improved by clustering and weight optimization of historical data sequences. Through the above technical solution, anomalies in multi-source sensor data can be accurately identified and effectively corrected, thereby improving the reliability of data collection for automated production lines. At the same time, the combination of the dynamic weight allocation mechanism and the recursive optimization model enhances the adaptability of the solution to different types of abnormal data, and provides a more reliable data basis for state monitoring and fault diagnosis of production lines.

[0043] In an optional embodiment, Calculating the performance parameters of the equipment components based on the operation parameter sequence, calculating the performance degradation rate according to the performance parameters, and when the performance degradation rate is greater than a second preset threshold, determining the corresponding standby component as a component to be monitored includes: Performing time-frequency analysis on the operating parameter sequence to obtain energy distribution characteristics, selecting a characteristic parameter combination based on the energy distribution characteristics, and constructing the characteristic parameter combination as a comprehensive performance status indicator; Analyze the temporal variation law of the comprehensive performance status index, establish a performance status transfer matrix, calculate the change rate of the performance parameter based on the performance status transfer matrix, and correct the change rate in combination with the performance status transfer matrix to obtain the performance degradation rate; Establish a performance baseline value for when the component is operating normally, determine the performance fluctuation tolerance range based on the performance baseline value combined with historical operating data, start the cumulative time count when it is detected that the performance degradation rate exceeds the performance fluctuation tolerance range, and when the cumulative time reaches a preset period, determine the corresponding equipment component as the component to be monitored.

[0044] In the performance monitoring process of automated production line equipment components, it is first necessary to perform time-frequency analysis on the collected operating parameter sequence. Taking the vibration signal as an example, the time domain signal is converted to the time-frequency domain through short-time Fourier transform to obtain the frequency characteristics that reflect the change of the signal over time. The energy statistics of the time-frequency graph are performed, the energy distribution ratio of different frequency bands is calculated, and the energy distribution characteristics are obtained. The energy distribution characteristics include information such as the energy proportion of each frequency band and the energy concentration of the main frequency band.

[0045] When selecting a combination of characteristic parameters based on energy distribution characteristics, first analyze the correlation between the energy of each frequency band and the performance status of the device. For frequency bands with a high energy ratio and a significant correlation with the performance status, extract characteristic parameters such as the average energy value and energy fluctuation. At the same time, consider the energy transfer characteristics between frequency bands and calculate the energy ratio of adjacent frequency bands as a supplementary feature. The selected characteristic parameters are weighted and combined according to their importance to construct a comprehensive indicator that reflects the overall performance status of the device components.

[0046] When analyzing the time series changes of the comprehensive performance status indicators, the indicator values ​​are first divided into multiple status levels according to their size. By counting the transitions of the status levels at adjacent moments, a transfer matrix describing the probability of status transition is established. Each element in the transfer matrix represents the probability of transitioning from one status level to another, and the eigenvalue of the matrix reflects the stability of the status transition.

[0047] When calculating the rate of change of performance parameters, the direction of state transition needs to be considered. The greater the probability of transition to the direction of performance degradation, the faster the performance degradation rate. The initial rate of change is obtained by calculating the weighted sum of the elements reflecting performance degradation in the transfer matrix. At the same time, the eigenvalues ​​of the transfer matrix are analyzed, and the rate of change is corrected according to the distribution characteristics of the eigenvalues ​​to obtain a more accurate performance degradation rate.

[0048] When establishing a performance benchmark, select historical data of normal operation of equipment components and calculate the statistical characteristics of comprehensive performance status indicators. By analyzing a large amount of normal operation data, obtain the average level of the indicator as the benchmark value. Combined with the fluctuations in historical operation data, calculate the standard deviation of the indicator, and use the multiple range of the standard deviation as the tolerance range of performance fluctuations.

[0049] During real-time monitoring, when the performance degradation rate is detected to be beyond the tolerance range, the cumulative time counter is started. The cumulative time count is used to determine whether the performance degradation persists to avoid misjudgment due to instantaneous fluctuations. When the cumulative time reaches the preset monitoring period, it is confirmed that the performance of the device component is indeed abnormal, and it is marked as a component to be monitored and enters the key monitoring state.

[0050] Taking the bearing assembly on the automated production line as an example, the vibration signal sequence of the bearing is collected by sensors. Time-frequency analysis of the vibration signal shows that there are significant differences in the energy distribution in the low-frequency band, the medium-frequency band, and the high-frequency band. During normal operation, the energy in the medium-frequency band accounts for the highest proportion, about half of the total energy; when the bearing performance begins to deteriorate, the energy proportion in the high-frequency band gradually increases.

[0051] The energy proportion of the mid-frequency band, the energy proportion of the high-frequency band, and the ratio of the two are selected as characteristic parameters to construct a comprehensive performance status index. The index value is divided into five status levels, and the state transition matrix is ​​established through continuous monitoring data. Analysis shows that when the bearing performance begins to deteriorate, the probability of the state level shifting to a worse direction increases significantly. Combined with the eigenvalue distribution of the transfer matrix, the corrected performance degradation rate is obtained.

[0052] Based on a large amount of historical operating data, a performance benchmark value for normal bearing operation is established. The fluctuation range of the comprehensive performance status index is usually within plus or minus 15% of the benchmark value. When the performance degradation rate is continuously monitored to exceed this range, and the cumulative time exceeds the set two operating cycles, the bearing component is listed in the monitoring state and timely inspection and maintenance are carried out.

[0053] In this embodiment, the energy distribution characteristics are extracted through time-frequency analysis to establish a dynamic characteristic parameter combination, thereby realizing a comprehensive evaluation of the performance status of the equipment. The analysis method based on the performance state transfer matrix can accurately capture the gradual characteristics of the performance parameters and improve the prediction accuracy of the performance degradation trend. By establishing an adaptive performance fluctuation tolerance range based on historical data, the adaptability of performance monitoring is improved. The cumulative time counting mechanism effectively filters the impact of instantaneous fluctuations and ensures the stability and reliability of performance degradation judgment. Accurate evaluation and timely warning of the performance status of equipment components are realized, providing a reliable basis for preventive maintenance decisions and improving the operation reliability and maintenance efficiency of the automated production line.

[0054] In an optional embodiment, Establishing a performance benchmark value for a component when it is operating normally, and determining a performance fluctuation tolerance range based on the performance benchmark value combined with historical operating data include: Acquire historical operation data of the equipment components, identify normal operation periods and abnormal operation periods from the historical operation data, compare and analyze the normal operation periods and abnormal operation periods, extract false positive information and missed negative information, and construct a reward function; Inputting the reward function and the performance data of the normal operation period into a reinforcement learning framework, obtaining a performance benchmark value through reinforcement learning training, and generating an initial tolerance adjustment strategy based on the performance benchmark value; Collect the real-time operation status data of the components, match and analyze the real-time operation status data with the initial tolerance adjustment strategy, calculate the status warning probability under different tolerance settings, and generate risk prediction results within the tolerance range; Obtaining the position information of the component in the production process, the process parameter requirements of the process, and the historical maintenance cost data, performing multi-objective optimization calculation on the position information, process parameter requirements, and historical maintenance cost data and the risk prediction results to generate a tolerance optimization objective function; The initial tolerance adjustment strategy is optimized based on the tolerance optimization objective function to obtain a performance fluctuation tolerance range, the performance fluctuation tolerance range is combined with a performance benchmark value to form a monitoring threshold, and the state data corresponding to the monitoring threshold is fed back to the reinforcement learning framework for continuous optimization training.

[0055] For example, when obtaining historical operation data of equipment components, it mainly includes performance parameter data, operating condition parameter data and maintenance record data. Performance parameter data includes time series of physical quantities such as vibration, temperature, and pressure; operating condition parameter data includes operating status information such as equipment operating speed and load; maintenance record data includes historical fault information, maintenance records, and replacement parts records.

[0056] A multi-layer data analysis method is used to identify normal operation periods and abnormal operation periods. First, the performance parameter data is statistically analyzed to calculate the mean, standard deviation, skewness, kurtosis and other statistics. Then, the data is stratified based on the operating parameters, and the data under similar operating conditions are clustered for analysis. Known abnormal periods are marked based on expert experience and historical maintenance records. The density clustering algorithm is used for anomaly detection on unlabeled data. The density clustering algorithm identifies abnormal points by calculating the local density of data points and the distance to high-density points.

[0057] During the comparative analysis, the test results are compared with the maintenance records to count the false positives and missed positives. A false positive refers to a situation where the test result is abnormal but actually normal, and a missed positive refers to a situation where an actual abnormality is not detected. The causes of the misjudgment are analyzed based on the time distribution of false positives and missed positives, working condition characteristics, and other information.

[0058] The construction of the reward function takes into account multiple evaluation dimensions. Accurate warnings receive positive rewards, and the reward value is proportional to the advance amount; false alarms and missed alarms receive negative rewards, and the negative reward value is related to the degree of impact caused. The degree of impact is quantified by factors such as equipment downtime and maintenance costs. At the same time, the impact of operating conditions is taken into account, and the penalty for false alarms is appropriately reduced during the operating condition switching period.

[0059] The reinforcement learning framework adopts a deep Q network structure. The state space contains the performance parameter values ​​at the current moment, the recent change trend characteristics and the working condition parameters. The action space is the optional range of the performance benchmark value, and the continuous value is converted into a finite number of discrete actions through discretization. The Q network uses a multi-layer neural network structure. The input layer receives the state information, the hidden layer uses a fully connected layer with batch normalization, and the output layer corresponds to the expected reward value of each action. The training process adopts an experience replay mechanism to store historical experience in the experience pool. A batch of experience samples are randomly selected for each training, including the state transition sequence and the corresponding reward value. The target Q value is calculated by the temporal difference algorithm, and the network parameters are updated by minimizing the mean square error between the predicted Q value and the target Q value. In order to balance exploration and utilization, the attenuated ε-greedy strategy is used to select actions.

[0060] In the matching and analysis process of real-time operation status data, the data is first preprocessed, including denoising, standardization and other operations. Then the change characteristics of performance parameters on different time scales are calculated, including short-term fluctuation characteristics and long-term trend characteristics. For each candidate tolerance range, the warning trigger probability is calculated through sliding window analysis. When calculating the status warning probability, multiple indicators are considered comprehensively. The short-term warning probability is calculated based on the frequency and amplitude of the parameter exceeding the tolerance range; the long-term warning probability considers the change trend and cumulative effect of the parameter. The short-term and long-term warning probabilities are weighted and combined, and the weights are dynamically adjusted according to different working conditions.

[0061] The analysis of process flow location information includes the analysis of the upstream and downstream correlation of equipment. The impact of equipment failure on related processes is statistically analyzed, including direct impact and chain reaction. The process parameter requirements are converted into specific performance indicator constraints, such as the tolerance limit corresponding to the accuracy level. Maintenance cost data includes spare parts cost, labor cost, downtime loss, etc., and a mapping relationship between cost and failure type and maintenance time is established. The multi-objective optimization calculation adopts the Pareto optimization method. The objectives such as early warning timeliness, reliability, and maintenance cost are constructed as optimization objective functions. The candidate solution set of the tolerance adjustment strategy is generated by genetic algorithm, and the performance of each solution on each objective is evaluated. The Pareto optimal solution set is selected by non-dominated sorting, and the final solution is selected based on actual needs.

[0062] The continuous optimization of monitoring thresholds uses an online learning mechanism. Regularly collect new monitoring data and maintenance records, and update samples in the experience pool. Use new samples to incrementally train the reinforcement learning model and adjust network parameters. At the same time, regularly evaluate the monitoring effect and dynamically adjust the parameter configuration of the reward function based on the evaluation results.

[0063] This adaptive threshold optimization method based on reinforcement learning can adapt to the dynamic changes of equipment operating status through continuous learning and dynamic adjustment, and improve the accuracy and reliability of monitoring. At the same time, multi-objective optimization ensures the balance between practicality and economy of the monitoring scheme, and provides effective decision support for equipment preventive maintenance.

[0064] Figure 3 This is a comparison chart of early warning accuracy indicators of different methods in the embodiments of the present invention. Figure 3 As shown in the figure, the comparison results of five different equipment status monitoring methods on five key performance indicators are shown. It can be clearly seen from the data in the table that this technical solution is significantly better than the existing technology in all indicators. In terms of warning accuracy, this technical solution reaches 94.2%, which is 14.7 percentage points higher than the machine learning method and 51.9 percentage points higher than the fixed threshold method; in terms of false alarm rate, this technical solution is only 4.2%, which is 60.0% lower than the machine learning method and 89.0% lower than the fixed threshold method; in terms of missed alarm rate, this technical solution is only 3.7%, which is 76.6% lower than the machine learning method and 88.6% lower than the fixed threshold method. It is particularly noteworthy that the warning lead time indicator, this technical solution can warn 46 hours in advance, which is 1.92 times that of the machine learning method (24 hours) and 5.75 times that of the fixed threshold method (8 hours), which has great practical value for maintenance decision-making.

[0065] In terms of the working condition adaptability index, this technical solution reached 92.3%, which is much higher than other methods, indicating that it has excellent adaptability in complex and changing industrial environments. This series of experimental results fully proves that the adaptive threshold optimization method based on reinforcement learning has obvious comprehensive performance advantages over traditional technologies, especially in reducing false alarms and missed alarms and improving the advance warning amount, providing more reliable technical support for industrial equipment status monitoring and preventive maintenance.

[0066] Existing equipment performance monitoring methods mostly use fixed thresholds or simple statistical models for judgment, which cannot adapt to the performance fluctuation characteristics of equipment under different working conditions, and do not fully consider the impact of false alarms and missed alarms on the monitoring system, resulting in insufficient reliability of monitoring results. At the same time, traditional methods do not take into account comprehensive factors such as equipment location, process requirements and maintenance costs, making it difficult to achieve economic optimization of monitoring strategies. This application improves the accuracy of the monitoring system in identifying equipment performance changes by constructing a reward function based on false alarm and missed alarm information, combining a reinforcement learning framework to achieve adaptive optimization of performance benchmark values. A multi-objective optimization method is used to comprehensively consider equipment location, process requirements and maintenance costs, and achieve dynamic adjustment of the tolerance range, while ensuring monitoring reliability and reducing maintenance costs. By continuously optimizing the training mechanism, the monitoring system can continuously learn and adapt to the dynamic changes in equipment status. This solution significantly improves the accuracy and reliability of equipment performance monitoring, reduces the occurrence of false alarms and missed alarms, achieves the optimization of preventive maintenance decisions, and improves the economic benefits of equipment management. At the same time, the adaptive ability of the monitoring system is enhanced, which can better adapt to the monitoring needs under different working conditions.

[0067] In an optional embodiment, The fault characteristics are input into the pre-trained fault identification model, and the fault type probability distribution is output based on the combination mode of the fault characteristics. The fault type with the highest probability is selected as the fault judgment result of the component to be monitored, including: Inputting the fault features into a pre-trained fault recognition model, decomposing the fault features into common features and individual features through a feature decoupling network, and applying orthogonal constraints to the common features and individual features to obtain decoupled features; The decoupled features are constructed as an attribute graph, the feature dimension is set as a graph node, the feature association strength is set as a graph edge, and local dependencies and global dependencies are extracted through graph convolution and multi-head attention mechanism to obtain a combination pattern of fault features; Inputting the combination mode of fault features and decoupled features into a variational inference module, establishing a state transfer matrix based on the feature association strength and orthogonal constraints, and obtaining an initial fault type probability; The initial fault type probability is modeled in a Bayesian manner, and probability calibration is performed based on the local dependencies and global dependencies in the combination pattern of fault characteristics to generate a calibrated fault type probability distribution. The fault type with the highest probability is selected from the calibrated fault type probability distribution as the fault judgment result of the component to be monitored.

[0068] For example, in the process of diagnosing equipment faults in automated production lines, the fault features need to be identified and analyzed first. Fault features include multi-source sensor data such as vibration signals, temperature, pressure, current, and equipment operating parameters and process parameters. These features are pre-processed and input into a pre-trained fault recognition model, which uses a deep learning architecture and includes core components such as feature decoupling networks, graph convolutional networks, and variational inference modules.

[0069] The feature decoupling network adopts an encoder-decoder structure, and maps the input fault features to the latent space through a multi-layer neural network. In the latent space, the features are decomposed into two subspaces: common features and individual features. Common features reflect the common feature patterns between different fault types, such as the basic characteristics of equipment vibration; individual features characterize the unique manifestations of various faults, such as the impact characteristics unique to bearing faults. To ensure the effectiveness of feature decomposition, an orthogonal constraint is imposed between the two subspaces to make them independent of each other. The orthogonal constraint is achieved by minimizing the inner product of the feature vectors of the two subspaces.

[0070] The decoupled features are constructed as an attribute graph structure, where the nodes represent features of different dimensions, such as vibration amplitude, frequency features, etc. The edges between nodes represent the strength of association between features, which is calculated through feature correlation analysis. The graph convolution operation is applied to the attribute graph, and the information of adjacent nodes is aggregated through the message passing mechanism to capture the local dependencies between features. At the same time, a multi-head attention mechanism is introduced to calculate the attention weights between different feature nodes and extract the global dependencies between features. Through the extraction of this dual dependency, a more complete fault feature combination pattern is obtained.

[0071] The variational inference module receives the combination pattern of fault features and decoupled features as input. The state transition matrix is ​​constructed based on the feature correlation strength, and the matrix elements represent the transition probability between feature states. At the same time, the orthogonal constraints of the decoupled features are considered to ensure the rationality of the state transition. The initial probability distribution of each fault type is obtained through variational inference calculation.

[0072] The initial fault type probability is modeled in Bayesian fashion, and prior knowledge is introduced. Prior knowledge includes the distribution of fault types in historical fault data, the rules of fault occurrence under different working conditions, etc. The initial probability is calibrated by combining the local dependencies and global dependencies extracted from the feature combination pattern. The local dependencies are used to adjust the fault probability distribution between related features, while the global dependencies consider the synergistic effect of the overall features. Finally, the calibrated fault type probability distribution is obtained, and the type with the highest probability is selected as the fault judgment result.

[0073] Taking bearing fault diagnosis as an example, we first collect multi-source data such as bearing vibration signals and temperature as fault features. These features are decomposed into common features reflecting the basic vibration characteristics of the bearing and individual features representing specific fault modes through a feature decoupling network. In the attribute graph, frequency domain features, time domain features, etc. are set as nodes, and the correlation coefficients between features are set as edge weights.

[0074] The graph convolution operation extracts the local correlation patterns between vibration frequency features, such as the mutual influence of energy in different frequency bands. The multi-head attention mechanism captures the long-range dependencies between vibration features and temperature features. These dependencies together constitute the combined pattern of fault features, reflecting the overall characteristics of fault development.

[0075] The variational inference module calculates the initial fault probability based on the feature combination pattern, such as the probability distribution of inner race fault, outer race fault, etc. Historical fault statistics are introduced through Bayesian modeling, and probability calibration is performed in combination with feature dependencies. Finally, the calibrated fault type probability distribution is output to accurately determine the specific fault type of the bearing.

[0076] In this embodiment, by introducing a feature decoupling network, the fault features are decomposed into common features and individual features, which improves the accuracy and discrimination of feature expression and realizes the effective identification of fault features. The attribute graph structure constructed based on graph convolution and multi-head attention mechanism can simultaneously capture the local dependency and global dependency between features, and enhance the expression ability of the fault feature combination pattern. The probability inference method combining variational inference and Bayesian modeling is adopted, and the probability calibration is performed in combination with the feature correlation strength and orthogonal constraints, which improves the accuracy and reliability of fault type determination. It can accurately identify various types of faults of equipment components, reduce the misjudgment rate and missed judgment rate, and provide reliable decision support for equipment preventive maintenance. At the same time, it has strong generalization ability, can adapt to the fault diagnosis needs under different working conditions, and improves the practicality and reliability of the fault diagnosis system.

[0077] In an optional embodiment, According to the physical connection relationship between the component to be monitored and the adjacent components, the performance degradation rate of the adjacent components is obtained, and the fault propagation link is established based on the time sequence of the performance degradation rate. The fault source components are determined to include: Obtain the mechanical coupling relationship, energy transfer relationship and physical topological relationship between the component to be monitored and the adjacent components, and establish a component connection relationship matrix that characterizes the spatial connection strength between components; Acquiring performance parameters of the component to be monitored and adjacent components within a continuous time window, calculating the time derivative of the performance parameters to obtain a performance degradation rate, and performing a Hilbert transform on the performance degradation rate to obtain an instantaneous phase; According to the change trend of the spatial connection strength and instantaneous phase in the component connection relationship matrix, the phase synchronization strength between adjacent components is calculated, and a network order parameter reflecting the dynamic coupling relationship between components is constructed; Monitor the dynamic changes of the network order parameters, obtain the mutation time and mutation amplitude of the network order parameters, and construct a directed fault propagation graph based on the mutation time, mutation amplitude and phase synchronization strength, in which adjacent components are nodes, the time difference of the mutation time is the edge weight, and the phase synchronization strength is the edge direction; Calculating the downstream impact range of each component according to the edge weight and edge direction in the directed fault propagation graph, determining the cascading failure scale of each component based on the downstream impact range, and combining the cascading failure scale with the mutation moment to obtain a failure impact factor; According to the dynamic evolution law of the directed fault propagation graph, the mutation time characteristics and cascading failure scale in the failure influence factor are weighted, and the component with the largest configured failure influence factor and the earliest mutation time is determined as the fault source component.

[0078] In the process of fault propagation analysis of automated production lines, the connection relationship between components must be obtained first. The mechanical coupling relationship describes the physical connection between components, such as the fit between bearings and shafts, the meshing of gears, etc. The energy transfer relationship represents the energy flow path between components, including the transmission of mechanical energy, thermal energy, etc. The physical topological relationship reflects the relative relationship between components in space. Based on these relationships, a component connection relationship matrix is ​​constructed. The matrix elements represent the spatial connection strength between components. The connection strength is calculated through physical parameters such as the contact area and connection stiffness between components.

[0079] In the set continuous time window, the performance parameters of the components to be monitored and the adjacent components are collected. Performance parameters include physical quantities such as vibration, temperature, and pressure. The performance degradation rate is obtained by calculating the rate of change of these parameters over time, which reflects the speed of component performance degradation. The performance degradation rate is subjected to Hilbert transform, which is a signal processing method that can extract the instantaneous phase information of the signal. The instantaneous phase represents the phase angle of the signal at each moment, reflecting the dynamic change characteristics of performance degradation.

[0080] According to the change law of spatial connection strength and instantaneous phase in the component connection relationship matrix, the phase synchronization strength between adjacent components is calculated. The phase synchronization strength indicates the degree of coordination of the performance degradation process of two components. The higher the synchronization strength, the more correlated the performance degradation of the two components is. The network order parameter is constructed based on the phase synchronization strength. The network order parameter is a parameter that describes the overall dynamic characteristics of a complex network and is used to characterize the co-evolution characteristics of a component group.

[0081] Dynamically monitor the network order parameters to identify their mutation moments and mutation amplitudes. The mutation moment refers to the time point when the network order parameters change significantly, and the mutation amplitude indicates the degree of change. Combine this information with the phase synchronization strength to construct a directed fault propagation graph. In this graph, components are nodes, and the fault propagation relationship between components is represented by directed edges. The weight of the edge is determined by the order of the mutation moments, and the component that mutates first points to the component that mutates later. The direction of the edge is determined by the phase synchronization strength, and the direction with greater synchronization strength is the main direction of fault propagation.

[0082] The downstream impact range of each component is analyzed based on the directed fault propagation graph. The downstream impact range refers to the number of other components that may be affected after a component fails. By analyzing the connection relationship of the directed edges, the number of nodes reachable from each component is calculated to obtain the scale of cascading failures that may be caused by the component. The cascading failure scale is combined with the mutation time information to construct the failure impact factor. The failure impact factor comprehensively considers the timing characteristics and impact range of component failures.

[0083] The dynamic evolution law of the directed fault propagation graph is analyzed, including the changing characteristics of the network topology and the evolution pattern of the fault propagation path. Based on these laws, the failure influence factors are weighted and the weight coefficients are assigned according to the importance of the mutation moment characteristics and the cascading failure scale. Finally, the component with the largest configured failure influence factor and the earliest mutation moment is identified as the fault source component.

[0084] Take the production line transmission system as an example, which includes key components such as motors, reducers, and spindles. First, establish the connection relationship matrix of these components. For example, the motor and reducer are connected through a coupling, and the connection strength is determined by the stiffness of the coupling. Within a one-hour time window, monitor the vibration signals of each component, calculate the change rate of the vibration amplitude, and obtain the performance degradation rate.

[0085] By extracting the instantaneous phase of the vibration signal of each component through Hilbert transform, it was found that the phase changes of the reducer and the main shaft have strong synchronization, while the phase synchronization between the motor and the reducer is weak. The constructed network order parameter shows that at a certain moment, the reducer first mutates, followed by abnormalities in the main shaft and the motor.

[0086] In the directed fault propagation graph, the edge weight from the reducer to the spindle is large, indicating that the fault is likely to propagate from the reducer to the spindle. Analysis shows that the reducer has the largest downstream impact range, and its failure may cause a chain reaction of the spindle and the motor. Taking into account the mutation timing and cascading failure scale, the reducer is finally determined to be the fault source component.

[0087] Existing fault tracing methods mainly rely on the fault feature analysis of a single component or simple correlation analysis, and fail to fully consider the complex physical connection relationship between components and the dynamic characteristics of fault propagation, which makes it difficult to accurately identify the fault source component. At the same time, the traditional method lacks quantitative analysis of the cascade effect during fault propagation, and cannot effectively evaluate the impact range and propagation path of the fault. This application comprehensively describes the mechanical coupling, energy transfer and physical topological relationship between components by constructing a component connection relationship matrix, providing a reliable physical basis for fault propagation analysis. The introduction of Hilbert transform and phase synchronization analysis methods realizes the accurate characterization of the dynamic characteristics of fault propagation. Based on the analysis framework of directed fault propagation graphs and failure influencing factors, a quantitative evaluation of fault propagation paths and cascade effects is realized, which significantly improves the accuracy of fault source identification, can effectively predict the propagation range and impact of faults, and provides a reliable basis for equipment maintenance decisions. At the same time, it has strong adaptability, can handle fault tracing problems of different types of equipment systems, and enhances the practical value of fault diagnosis systems.

[0088] A second aspect of the present invention provides a multi-parameter real-time monitoring and fault diagnosis and early warning system for an automated production line, the system comprising: The first unit is used to obtain real-time operation data collected by multi-source sensors in the automated production line, calculate data quality characteristic values, generate data credibility, and when the data credibility is less than a first preset threshold, perform data correction based on historical data at adjacent moments to obtain corrected real-time operation data; The second unit is used to group the corrected real-time operation data according to the equipment components, generate an operation parameter sequence corresponding to the equipment components, calculate the performance parameters of the equipment components based on the operation parameter sequence, calculate the performance degradation rate according to the performance parameters, and when the performance degradation rate is greater than a second preset threshold, determine the corresponding standby component as a component to be monitored; The third unit is used to perform time-frequency domain feature analysis on the operating parameter sequence of the component to be monitored, extract time-domain statistical features and frequency-domain energy features as fault features, input the fault features into a pre-trained fault recognition model, output a fault type probability distribution based on a combination mode of the fault features, and select the fault type with the highest probability as the fault determination result of the component to be monitored; The fourth unit is used to obtain the performance degradation rate of the adjacent components according to the physical connection relationship between the component to be monitored and the adjacent components, establish a fault propagation link based on the time series change sequence of the performance degradation rate, determine the fault source component, generate a fault diagnosis report containing the location information of the fault source component and the fault judgment result, and send it to the production management system.

[0089] According to a third aspect of the embodiments of the present invention, An electronic device is provided, comprising: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0090] According to a fourth aspect of the embodiments of the present invention, A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0091] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for real-time monitoring and fault diagnosis and early warning of multiple parameters in an automated production line, characterized in that: include: Acquire real-time operation data collected by multi-source sensors in the automated production line, calculate data quality characteristic values, generate data credibility, and when the data credibility is less than a first preset threshold, perform data correction based on historical data at adjacent moments to obtain corrected real-time operation data; The corrected real-time operation data is grouped according to the equipment components, an operation parameter sequence corresponding to the equipment components is generated, performance parameters of the equipment components are calculated based on the operation parameter sequence, and a performance degradation rate is calculated according to the performance parameters. When the performance degradation rate is greater than a second preset threshold, the corresponding standby component is determined as a component to be monitored; Performing time-frequency domain feature analysis on the operating parameter sequence of the component to be monitored, extracting time-domain statistical features and frequency-domain energy features as fault features, inputting the fault features into a pre-trained fault recognition model, outputting a probability distribution of fault types based on a combination pattern of the fault features, and selecting the fault type with the highest probability as the fault determination result of the component to be monitored; According to the physical connection relationship between the component to be monitored and the adjacent components, the performance degradation rate of the adjacent components is obtained, and a fault propagation link is established based on the time series change sequence of the performance degradation rate. The fault source component is determined, and a fault diagnosis report containing the location information of the fault source component and the fault judgment result is generated and sent to the production management system.

2. The method according to claim 1, characterized in that: Acquire real-time operation data collected by multi-source sensors in the automated production line, calculate data quality characteristic values, generate data credibility, and when the data credibility is less than a first preset threshold, perform data correction based on historical data at adjacent moments, and obtain corrected real-time operation data including: Acquire the real-time operation data collected by multi-source sensors in the automated production line and segment it according to the time window to obtain a segmented data sequence; The signal fluctuation, value variation and sampling integrity of the segmented data sequence are calculated respectively and combined to obtain the data quality characteristic value, the information entropy of the data quality characteristic value is calculated based on the probability distribution characteristics of the data in the segmented data sequence, a dynamic weight coefficient is generated according to the information entropy, and the dynamic weight coefficient is weightedly combined with the corresponding data quality characteristic value to obtain the data credibility; When the data credibility is less than a first preset threshold, the data to be corrected is determined, and a historical data sequence of adjacent moments of N time windows before and after the time window corresponding to the data to be corrected is extracted; Clustering the historical data sequences at adjacent moments according to the similarity of data change trends to obtain multiple subsequence groups, calculating the time correlation between each subsequence in each subsequence group and the data to be corrected, and constructing a recursive optimization model with differentiated weights based on the time correlation, wherein the weight is positively correlated with the time correlation; A recursive optimization model is used to iteratively correct the data to be corrected. In each round of iteration, the fitting error between the correction result and the subsequence whose correlation is greater than a preset correlation threshold is calculated. When the fitting error is less than a dynamic threshold, the correction result is output as the corrected real-time running data.

3. The method according to claim 2, characterized in that Calculating the information entropy of the data quality characteristic value based on the probability distribution characteristics of the data in the segmented data sequence, and generating a dynamic weight coefficient according to the information entropy includes: Constructing a probability density estimation function, calculating the standard deviation and the interquartile range based on the segmented data sequence, dynamically adjusting the bandwidth coefficient of the probability density estimation function according to the ratio of the standard deviation and the interquartile range, and obtaining a continuous probability density curve of the segmented data sequence; Performing piecewise integration on the continuous probability density curve to obtain a cumulative probability value of each segment of data, dividing the segmented data sequence into a first distribution area and a second distribution area according to the cumulative probability value, and calculating a cumulative probability ratio of the first distribution area to the second distribution area as a data distribution characteristic ratio; The overlapping sliding window step length is set by using the data distribution characteristic ratio, and the first local information entropy is obtained by sliding calculation in the first distribution area; the non-overlapping sliding window width is set by the inverse of the data distribution characteristic ratio, and the second local information entropy is obtained by calculation in the second distribution area; Taking the cumulative probability values ​​of the first distribution area and the second distribution area as weight coefficients, weighting the first local information entropy and the second local information entropy to obtain a regional combined entropy; The global information entropy is calculated for the continuous probability density curve, the regional combined entropy and the global information entropy are normalized to obtain the final information entropy of the data quality characteristic value, and a dynamic weight coefficient is generated according to the final information entropy, and the dynamic weight coefficient is inversely proportional to the final information entropy.

4. The method according to claim 1, characterized in that: Calculating the performance parameters of the equipment components based on the operation parameter sequence, calculating the performance degradation rate according to the performance parameters, and when the performance degradation rate is greater than a second preset threshold, determining the corresponding standby component as a component to be monitored includes: Performing time-frequency analysis on the operating parameter sequence to obtain energy distribution characteristics, selecting a characteristic parameter combination based on the energy distribution characteristics, and constructing the characteristic parameter combination as a comprehensive performance status indicator; Analyze the temporal variation law of the comprehensive performance status index, establish a performance status transfer matrix, calculate the change rate of the performance parameter based on the performance status transfer matrix, and correct the change rate in combination with the performance status transfer matrix to obtain the performance degradation rate; Establish a performance baseline value for when the component is operating normally, determine the performance fluctuation tolerance range based on the performance baseline value combined with historical operating data, start the cumulative time count when it is detected that the performance degradation rate exceeds the performance fluctuation tolerance range, and when the cumulative time reaches a preset period, determine the corresponding equipment component as the component to be monitored.

5. The method according to claim 4, characterized in that Establishing a performance benchmark value for a component when it is operating normally, and determining a performance fluctuation tolerance range based on the performance benchmark value combined with historical operating data include: Acquire historical operation data of the equipment components, identify normal operation periods and abnormal operation periods from the historical operation data, compare and analyze the normal operation periods and abnormal operation periods, extract false positive information and missed negative information, and construct a reward function; Inputting the reward function and the performance data of the normal operation period into a reinforcement learning framework, obtaining a performance benchmark value through reinforcement learning training, and generating an initial tolerance adjustment strategy based on the performance benchmark value; Collect the real-time operation status data of the components, match and analyze the real-time operation status data with the initial tolerance adjustment strategy, calculate the status warning probability under different tolerance settings, and generate risk prediction results within the tolerance range; Obtaining the position information of the component in the production process, the process parameter requirements of the process, and the historical maintenance cost data, performing multi-objective optimization calculation on the position information, process parameter requirements, and historical maintenance cost data and the risk prediction results to generate a tolerance optimization objective function; The initial tolerance adjustment strategy is optimized based on the tolerance optimization objective function to obtain a performance fluctuation tolerance range, the performance fluctuation tolerance range is combined with a performance benchmark value to form a monitoring threshold, and the state data corresponding to the monitoring threshold is fed back to the reinforcement learning framework for continuous optimization training.

6. The method according to claim 1, characterized in that The fault characteristics are input into the pre-trained fault identification model, and the fault type probability distribution is output based on the combination mode of the fault characteristics. The fault type with the highest probability is selected as the fault judgment result of the component to be monitored, including: Inputting the fault features into a pre-trained fault recognition model, decomposing the fault features into common features and individual features through a feature decoupling network, and applying orthogonal constraints to the common features and individual features to obtain decoupled features; The decoupled features are constructed as an attribute graph, the feature dimension is set as a graph node, the feature association strength is set as a graph edge, and local dependencies and global dependencies are extracted through graph convolution and multi-head attention mechanism to obtain a combination pattern of fault features; Inputting the combination mode of fault features and decoupled features into a variational inference module, establishing a state transfer matrix based on the feature association strength and orthogonal constraints, and obtaining an initial fault type probability; The initial fault type probability is modeled in a Bayesian manner, and probability calibration is performed based on the local dependencies and global dependencies in the combination pattern of fault characteristics to generate a calibrated fault type probability distribution. The fault type with the highest probability is selected from the calibrated fault type probability distribution as the fault judgment result of the component to be monitored.

7. The method according to claim 1, characterized in that According to the physical connection relationship between the component to be monitored and the adjacent components, the performance degradation rate of the adjacent components is obtained, and the fault propagation link is established based on the time sequence of the performance degradation rate. The fault source components are determined to include: Obtain the mechanical coupling relationship, energy transfer relationship and physical topological relationship between the component to be monitored and the adjacent components, and establish a component connection relationship matrix that characterizes the spatial connection strength between components; Acquiring performance parameters of the component to be monitored and adjacent components within a continuous time window, calculating the time derivative of the performance parameters to obtain a performance degradation rate, and performing a Hilbert transform on the performance degradation rate to obtain an instantaneous phase; According to the change trend of the spatial connection strength and instantaneous phase in the component connection relationship matrix, the phase synchronization strength between adjacent components is calculated, and a network order parameter reflecting the dynamic coupling relationship between components is constructed; Monitor the dynamic changes of the network order parameters, obtain the mutation time and mutation amplitude of the network order parameters, and construct a directed fault propagation graph based on the mutation time, mutation amplitude and phase synchronization strength, in which adjacent components are nodes, the time difference of the mutation time is the edge weight, and the phase synchronization strength is the edge direction; Calculating the downstream impact range of each component according to the edge weight and edge direction in the directed fault propagation graph, determining the cascading failure scale of each component based on the downstream impact range, and combining the cascading failure scale with the mutation moment to obtain a failure impact factor; According to the dynamic evolution law of the directed fault propagation graph, the mutation time characteristics and cascading failure scale in the failure influence factor are weighted, and the component with the largest configured failure influence factor and the earliest mutation time is determined as the fault source component.

8. An automated production line multi-parameter real-time monitoring and fault diagnosis and early warning system, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to obtain real-time operation data collected by multi-source sensors in the automated production line, calculate data quality characteristic values, generate data credibility, and when the data credibility is less than a first preset threshold, perform data correction based on historical data at adjacent moments to obtain corrected real-time operation data; The second unit is used to group the corrected real-time operation data according to the equipment components, generate an operation parameter sequence corresponding to the equipment components, calculate the performance parameters of the equipment components based on the operation parameter sequence, calculate the performance degradation rate according to the performance parameters, and when the performance degradation rate is greater than a second preset threshold, determine the corresponding standby component as a component to be monitored; The third unit is used to perform time-frequency domain feature analysis on the operating parameter sequence of the component to be monitored, extract time-domain statistical features and frequency-domain energy features as fault features, input the fault features into a pre-trained fault recognition model, output a fault type probability distribution based on a combination mode of the fault features, and select the fault type with the highest probability as the fault determination result of the component to be monitored; The fourth unit is used to obtain the performance degradation rate of the adjacent components according to the physical connection relationship between the component to be monitored and the adjacent components, establish a fault propagation link based on the time series change sequence of the performance degradation rate, determine the fault source component, generate a fault diagnosis report containing the location information of the fault source component and the fault judgment result, and send it to the production management system.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images