Processing production line monitoring method and system for fan controller

Abstract

The invention relates to the field of data processing, in particular to a processing production line monitoring method and system for a fan controller. The method comprises the following steps: acquiring multi-modal time sequence data; constructing a production line topological structure diagram, processing the data by using a space-time diagram convolutional network, and extracting space-time correlation features; judging whether the production line is in an abnormal state or not in real time; calculating the gradient contribution degree of the health entropy of the production line to the input characteristics of each station, and positioning an abnormal root station with the highest contribution degree; and inputting the time-space correlation characteristics into a track prediction module based on Kalman filtering and an attention mechanism to generate an evolution track of the health entropy of the production line in a future preset time window, the evolution track including a predicted value and a confidence interval, and being used for performing early warning on potential deterioration of the state of the production line. By adopting the method provided by the invention, the systematic risk of the fan controller processing production line can be effectively predicted, and the troubleshooting efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and more particularly to a method and system for monitoring a processing line of a fan controller. Background Art

[0002] Wind turbine controllers, as the core control units of wind turbine generators, are manufactured with a high degree of automation and complex processes, placing extremely stringent demands on product quality and production line stability. Currently, production line monitoring for high-end electronic equipment like wind turbine controllers primarily relies on Manufacturing Execution Systems (MES) to record and compile data on production cycle time, material consumption, and output at each workstation, or employs traditional Statistical Process Control (SPC) methods to set static thresholds for key physical parameters such as temperature and torque at individual workstations. Some production lines also deploy industrial cameras for video surveillance, but the captured image data is typically used only for post-production manual quality tracing or isolated visual inspection of specific defects. These approaches often suffer from insufficient data utilization, creating "information silos" from disparate data sources such as sensors, MES systems, and video, failing to effectively integrate them to provide a comprehensive, three-dimensional understanding of production line status. At the same time, they analyze each processing station as an independent entity, ignoring the close spatial correlation and dynamic timing dependence between stations formed by process flow, making it difficult to capture chain anomalies caused by small deviations in upstream stations that are transmitted and accumulated to downstream stations.

[0003] Due to the aforementioned limitations, existing monitoring technologies are unable to cope with complex production line status changes. Furthermore, in terms of anomaly detection, the lack of a comprehensive metric that can quantify the overall health of the production line makes it difficult for the system to identify systemic risks caused by the coordinated effects of multiple workstations early on, even though they are still within the tolerance range of a single parameter. Consequently, the system often responds passively only after serious quality defects or equipment downtime occur, resulting in irreversible losses. Secondly, in terms of anomaly tracing, when an alarm is triggered, locating the actual root cause of the problem relies heavily on the engineer's personal experience and step-by-step investigation, a time-consuming and inefficient process. There is a lack of intelligent means to automatically and accurately locate the source of anomalies from massive amounts of data. Finally, in terms of risk warning, existing technologies mostly rely on post-event analysis and lack the ability to proactively predict future trends in production line status. This inability to anticipate potential performance deterioration risks and provide early warnings, causing companies to miss the optimal opportunity to implement preventive maintenance and avoid production interruptions.

[0004] In summary, the monitoring methods of fan controller processing production lines in the existing technology have the problems of being unable to provide a global and three-dimensional understanding of the production line status, being unable to predict the systemic risks of the fan controller processing production line, and having low efficiency and accuracy in troubleshooting. Summary of the Invention

[0005] In order to solve the problems in the existing technology of monitoring methods for fan controller processing production lines, such as the inability to provide a global and three-dimensional understanding of the production line status, the inability to predict the systemic risks of the fan controller processing production line, and the low efficiency and accuracy of troubleshooting, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for monitoring a processing line of a wind turbine controller, comprising: Real-time physical parameters, manufacturing execution system data, and video image data of each processing station on the production line are obtained to form multimodal time series data. A production line topology diagram is constructed with processing stations as nodes and process flow relationships as edges. The data is processed using a spatiotemporal graph convolutional network to extract spatiotemporal correlation features that combine the spatial dependencies and temporal evolution laws between stations. Based on the spatiotemporal correlation characteristics, the production line health entropy representing the current operational stability of the production line is calculated, and by comparing it with a preset health baseline, it is determined in real time whether the production line is in an abnormal state; When an abnormal state is detected on a production line, the reverse gradient propagation algorithm is used to locate the abnormal root cause station with the highest contribution by calculating the gradient contribution of the production line health entropy to the input features of each station. The spatiotemporal correlation features are input into a trajectory prediction module based on Kalman filtering and attention mechanism to generate an evolution trajectory of the production line health entropy within a future preset time window. The evolution trajectory includes a predicted value and a confidence interval, which is used to provide early warning of potential deterioration of the production line status.

[0007] Preferably, the acquisition of real-time physical parameters, manufacturing execution system data, and workstation video image data of each processing station on the production line to form multimodal time series data includes: The physical parameters of each workstation are collected through an industrial controller; the manufacturing data associated with the production batch and workpiece is obtained through the manufacturing execution system; the image data of the workstation operation is captured through an industrial camera; and a unified timestamp is used to time-align the physical parameters, manufacturing data and image data to form multimodal time series data.

[0008] Preferably, the spatiotemporal graph convolutional network is composed of alternately stacked graph convolution modules and temporal convolution modules; wherein the graph convolution module is used to aggregate the information of adjacent workstation nodes in the production line topology diagram at each time step to capture spatial dependencies; the temporal convolution module is used to extract the dynamic change law of its own multimodal time series data at each workstation node to capture temporal dependencies.

[0009] Preferably, the calculation of the production line health entropy representing the current operating stability of the production line includes: A clustering algorithm is used to group the spatiotemporal correlation feature vectors of all workstations within the current time window to obtain multiple clusters. The proportion of the number of feature vectors in each cluster to the total number is counted to form a discrete probability distribution. Based on the probability distribution, the Shannon entropy is calculated as the production line health entropy.

[0010] Preferably, the calculation of the production line health entropy that characterizes the current operating stability of the production line includes: inputting the extracted spatiotemporal correlation features into a multi-layer perceptron, whose output layer uses a Softmax function to calculate the probability distribution of the production line in various preset states, where the types of preset states include normal, slight deviation, and serious abnormality; then calculating the information entropy of the probability distribution according to the Shannon entropy formula to obtain a single scalar value, namely the production line health entropy.

[0011] Preferably, the health baseline is determined based on the statistical distribution of the production line health entropy calculated during multiple historical normal operation cycles of the production line; when the real-time calculated production line health entropy is higher than a certain proportion of the health baseline, it is judged that the production line is in an abnormal state.

[0012] Preferably, determining in real time whether the production line is in an abnormal state includes: During the normal operation cycle of the production line, a large number of health entropy values ​​are collected, and their mean and standard deviation are obtained through Gaussian distribution fitting. The mean plus three times the standard deviation is set as the dynamic threshold for abnormality judgment. If the real-time calculated health entropy of the production line exceeds the dynamic threshold, it is judged as abnormal.

[0013] Preferably, the abnormality root station with the highest positioning contribution includes: Taking the health entropy of the production line as the objective function, backpropagation is performed along the spatiotemporal graph convolutional network to calculate the gradient of the health entropy with respect to the input features of each workstation within a preset time window before the abnormality occurs. The absolute values ​​of the gradients of all input features of each workstation are accumulated to obtain the total gradient contribution of the workstation. The total gradient contributions of all workstations are ranked, and the workstation with the highest contribution is identified as the root cause of the abnormality.

[0014] Preferably, generating the evolution trajectory of the production line health entropy within a future preset time window includes: The attention mechanism is used to process the historical spatiotemporal correlation feature sequence to assign weights to the spatiotemporal correlation features of each time step; The feature sequence weighted by the attention mechanism is input into the Kalman filter. Through iterative prediction and update steps, a sequence of predicted values ​​of the production line health entropy within a preset future time window is generated. The state covariance matrix of the Kalman filter is used to calculate the confidence interval for each predicted value.

[0015] In a second aspect, the present invention provides a processing production line monitoring system for a fan controller, comprising a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the processing production line monitoring method for a fan controller of the present invention is implemented.

[0016] The beneficial effects of the present invention are as follows: compared with the existing technology, the present invention deeply integrates multi-source heterogeneous information such as physical parameters, manufacturing execution system data and video images by constructing a production line topology and utilizing a spatiotemporal graph convolutional network, thereby overcoming the limitations of data islands and being able to capture the spatial dependencies and temporal evolution laws between workstations from a global perspective, thereby obtaining a more comprehensive and profound understanding of the operating status of the production line. Based on this, the production line health entropy proposed by the present invention is used as a quantitative indicator to accurately evaluate the overall stability of the production line and effectively identify systematic and early abnormalities caused by the collaborative influence of multiple workstations that are difficult to detect with traditional methods. When an abnormality occurs, the present invention can automatically and quickly locate the root workstation, significantly improving the efficiency and accuracy of troubleshooting. Furthermore, by predicting the future trend of health entropy, the present invention realizes the transition from passive response to active prevention, provides a forward-looking warning for potential deterioration of production line status, and helps to take preventive measures to ensure production continuity and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flow chart schematically illustrating a method for monitoring a processing line of a wind turbine controller according to an embodiment of the present invention; Figure 2 is a schematic diagram schematically illustrating multimodal time series data according to an embodiment of the present invention; Figure 3 Schematic diagram of a spatiotemporal graph convolutional network according to an embodiment of the present invention; Figure 4 is a schematic diagram schematically illustrating health entropy and abnormality judgment of a production line according to an embodiment of the present invention; Figure 5 is a schematic diagram schematically illustrating an evolution trajectory of production line health entropy generated according to an embodiment of the present invention; Figure 6 FIG. 1 is a schematic diagram showing the structure of a processing line monitoring system for a wind turbine controller according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0019] Example of a production line monitoring method for a fan controller: like Figure 1As shown, the processing production line monitoring method for a fan controller of the present invention includes: S101. Use the spatiotemporal graph convolutional network to obtain spatiotemporal correlation features, specifically: Figure 2 As shown in the figure, real-time physical parameters, manufacturing execution system data, and video image data of each processing station on the production line are obtained to form multimodal time series data. A production line topology structure diagram is constructed with processing stations as nodes and process flow relationships as edges. The data is processed using a spatiotemporal graph convolutional network to extract spatiotemporal correlation features that integrate the spatial dependence and temporal evolution laws between stations. In this embodiment, the real-time physical parameters, manufacturing execution system data and station video image data of each processing station on the production line are obtained to form multimodal time series data, including: collecting the physical parameters of each station through an industrial controller; obtaining manufacturing data associated with production batches and workpieces through the manufacturing execution system; capturing image data of station operations through an industrial camera; and using a unified timestamp to time-align the physical parameters, manufacturing data and image data to form multimodal time series data.

[0020] In this embodiment, the industrial controller collects physical parameters of each workstation using Modbus or OPC UA protocols. Manufacturing data can be retrieved from the manufacturing execution system via an API interface. Image data of workstation operations captured by industrial cameras can be transmitted via the RTSP protocol.

[0021] The collected physical parameters include the placement machine torque, reflow oven temperature, etc. The collected manufacturing execution system data includes work order number, cycle time, etc.

[0022] According to the process flow, equipment such as placement machines, reflow soldering, and optical inspection are defined as graph nodes, the material flow direction is defined as directed edges, and an adjacency matrix is ​​generated to represent the production line topology.

[0023] The visual features of the video data are extracted through a pre-trained ResNet network, and then input together with other data as node attributes into a composite spatiotemporal graph convolutional network model consisting of a spatial graph convolutional layer and a gated recurrent unit. The spatiotemporal graph convolutional network model learns and outputs a feature vector that integrates global spatiotemporal dependencies. This feature vector is used to represent spatiotemporal correlation features.

[0024] like Figure 3 As shown, in this embodiment, the spatiotemporal graph convolution network is composed of alternately stacked graph convolution modules and time convolution modules; wherein, the graph convolution module is used to aggregate the information of adjacent workstation nodes in the production line topology diagram at each time step to capture spatial dependencies; the time convolution module is used to extract the dynamic change law of its own multimodal time series data at each workstation node to capture time dependencies.

[0025] Specifically, assume that a production line consists of three serial workstations: loading, processing, and unloading. At time T, the graph convolution module processes the processing station's own data while simultaneously integrating information about product arrival at the upstream loading station and the idle status of the downstream unloading station, reflecting their spatial interaction. Subsequently, the temporal convolution module focuses on analyzing the changing trends of a series of data such as temperature, pressure, and vibration at the processing station from time T minus 10 seconds to time T, such as whether the temperature continues to rise, thereby capturing dynamic characteristics in the temporal dimension.

[0026] S102: Determine in real time whether the production line is in an abnormal state, specifically by calculating the production line health entropy representing the current operating stability of the production line based on the spatiotemporal correlation characteristics, and comparing it with a preset health baseline to determine in real time whether the production line is in an abnormal state; In this embodiment, the calculation of the production line health entropy that characterizes the current operating stability of the production line includes: using a clustering algorithm to group the spatiotemporal correlation feature vectors of all workstations in the current time window to obtain multiple clusters; counting the proportion of the number of feature vectors in each cluster to the total number to form a discrete probability distribution; and calculating the Shannon entropy as the production line health entropy based on the probability distribution.

[0027] Specifically, in a production line with 10 workstations, the model generated a feature vector for each workstation in the past 5 minutes, for a total of 10 vectors. These 10 vectors were divided into 3 clusters using the K-means clustering algorithm. If 8 of the vectors are clustered in cluster A, representing stable operation, 1 vector is in cluster B, representing slight fluctuations, and 1 vector is in cluster C, representing anomalies, then the resulting probability distribution is P(A) equal to 0.8, P(B) equal to 0.1, and P(C) equal to 0.1. The Shannon entropy value calculated based on this distribution is low, for example, 0.92, which indicates that the operating status of the production line is highly consistent and stable. Conversely, if the vectors are evenly distributed in the three clusters, the entropy value will increase significantly, for example, close to 1.58, indicating that the production line is in a chaotic or unstable state.

[0028] In another embodiment, calculating the production line health entropy, which characterizes the current operational stability of the production line, includes: inputting the extracted spatiotemporal correlation features into a multi-layer perceptron, whose output layer uses a Softmax function to calculate the probability distribution of the production line being in various preset states, including normal, slightly deviated, and severely abnormal; then calculating the information entropy of the probability distribution based on the Shannon entropy formula to obtain a single scalar value, the production line health entropy. A smaller single scalar value indicates a higher certainty of the production line state and more stable operation.

[0029] In this embodiment, the health baseline is determined based on the statistical distribution of the production line health entropy calculated during multiple historical normal operation cycles of the production line; when the real-time calculated production line health entropy is higher than a certain proportion of the health baseline, it is judged that the production line is in an abnormal state.

[0030] In this embodiment, the method for calculating the health baseline includes: collecting a large number of health entropy values ​​during the normal operation cycle of the production line, and using the average of the collected health entropy values ​​as the health baseline.

[0031] For example, by analyzing the data of the production line during normal production over the past month, the health entropy of the production line was calculated every minute. The average of these entropy values ​​was 0.9. Therefore, 0.9 was set as the health baseline.

[0032] In another embodiment, real-time determination of whether a production line is in an abnormal state includes: collecting a large number of health entropy values ​​during the normal operation cycle of the production line, obtaining their mean and standard deviation through Gaussian distribution fitting, and setting the mean plus three times the standard deviation as a dynamic threshold for abnormality determination; if the real-time calculated health entropy of the production line exceeds the dynamic threshold, it is determined to be abnormal.

[0033] If the health baseline is 1 and the dynamic threshold is 2, when the health entropy of the production line exceeds 2, it is determined that the production line is abnormal.

[0034] S103. Locating the abnormal root cause workstation with the highest contribution. Specifically, when it is determined that the production line is in an abnormal state, the reverse gradient propagation algorithm is used to calculate the gradient contribution of the production line health entropy to the input features of each workstation to locate the abnormal root cause workstation with the highest contribution. In this embodiment, if Figure 4 As shown, locating the abnormality root cause station with the highest contribution includes: using the production line health entropy as the objective function, performing backpropagation along the spatiotemporal graph convolutional network, and calculating the gradient of the health entropy with respect to the input features of each station within a preset time window before the abnormality occurred; accumulating the absolute values ​​of the gradients of all input features of each station to obtain the total gradient contribution of that station; and sorting the total gradient contributions of all stations. The station with the highest contribution is identified as the abnormality root cause station. For example, if a fan controller processing line has three stations—the placement machine, reflow soldering, and optical inspection—and the placement machine has the highest total gradient contribution, then the placement machine is determined to be the abnormality root cause station.

[0035] For example, if a production line's health entropy is detected to have suddenly changed from a normal 0.9 to an abnormal 1.8, the target entropy value of 1.8 is set. The algorithm then traces back to calculate the impact of this entropy change on all input data in the minute before the anomaly, i.e., the gradient. The calculation results may show that the sum of the absolute values ​​of the gradient of the pressure sensor data and motor current data for stamping station 3 to the entropy change is 5.6, while the combined gradient contributions of other stations, such as loading station 1 and cutting station 2, are only 0.3 and 0.5 respectively. Because station 3's total gradient contribution of 5.6 is significantly higher than that of other stations, it is identified as the root cause of the production line's abnormal status.

[0036] S104. Generate an evolution trajectory of the health entropy of the production line within a future preset time window, specifically: input the spatiotemporal correlation features into a trajectory prediction module based on Kalman filtering and attention mechanism to generate an evolution trajectory of the health entropy of the production line within a future preset time window. The evolution trajectory includes a predicted value and a confidence interval, which is used to provide an early warning of potential deterioration of the production line status.

[0037] In this embodiment, the trajectory prediction module adopts an encoder-decoder architecture based on a long short-term memory network and introduces an attention mechanism; the spatiotemporal correlation feature sequence of the past period of time is input into the encoder, and the decoder generates the predicted value of future health entropy step by time; during the decoding process, the attention mechanism enables the decoder to focus on the most relevant spatiotemporal features in history when generating each prediction point; the output sequence of the decoder is used as the observation value of the Kalman filter, and through the prediction and update steps of the filter, the predicted trajectory is smoothed and the covariance of the prediction error is estimated. The covariance is used to calculate the confidence interval of each prediction point, thereby forming an evolution trajectory containing a central prediction line and upper and lower boundaries. In this embodiment, the generated evolution trajectory is as follows Figure 5 shown.

[0038] In this embodiment, generating the evolution trajectory of the production line health entropy within a future preset time window includes: S201, using the attention mechanism to process the historical spatiotemporal correlation feature sequence to assign weights to the spatiotemporal correlation features of each time step; The attention mechanism can automatically learn the importance of spatiotemporal correlation features. It can dynamically calculate weights to assign higher importance to spatiotemporal correlation features that are more relevant to the current production line status prediction. Spatiotemporal correlation features that are more relevant to the current production line status prediction refer to spatiotemporal correlation features that have a greater impact on the current production line status prediction.

[0039] S202. Input the feature sequence weighted by the attention mechanism into the Kalman filter, and generate a sequence of predicted values ​​of the production line health entropy within a future preset time window through iterative prediction and update steps; and use the state covariance matrix of the Kalman filter to calculate the confidence interval for each predicted value.

[0040] For a certain predicted value y, constructing a confidence interval involves: (1) Setting the confidence level ; In this embodiment, Can be set to 95%.

[0041] (2) Obtain the corresponding standard normal distribution quantile by looking up the table .

[0042] (3) Calculate the confidence interval. The confidence interval expression is ; in, is the variance corresponding to the predicted value y.

[0043] For example, to predict the health entropy of a production line 10 minutes into the future, the feature data from the past hour is first analyzed. The attention mechanism might discover that a brief material supply delay that occurred 15 minutes ago had the greatest impact on the current state and therefore assign it a weight of 0.6, while the feature weight of the stable operating state 30 minutes ago was only 0.1. This weighted feature sequence is fed into the Kalman filter, which predicts a health entropy of 0.95 for the first minute, 0.98 for the second minute, and so on, to generate a ten-minute prediction sequence. The filter also provides a confidence interval for each prediction value. For example, the 95 percent confidence interval for the first-minute prediction of 0.95 is between 0.92 and 0.98, providing decision makers with a quantitative assessment of the reliability of the prediction.

[0044] Example of a production line monitoring system for a fan controller: The present invention also provides a production line monitoring system for a fan controller. Figure 6 As shown, the processing production line monitoring system for a fan controller includes a processor and a memory, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, a processing production line monitoring method for a fan controller described in the above embodiment is implemented.

[0045] The processing production line monitoring system for the fan controller also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface. Their settings and functions are known in the art and will not be described in detail here.

[0046] In the description of this specification, "multiple" and "several" mean at least two, such as two, three or more, etc., unless otherwise clearly defined.

[0047] While several embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous modifications, variations, and alternatives will occur to those skilled in the art without departing from the concept and spirit of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be employed in practicing the present invention.

Claims

1. A method for monitoring a production line of a fan controller, characterized in that: include: Real-time physical parameters, manufacturing execution system data, and video image data of each processing station on the production line are obtained to form multimodal time series data. A production line topology diagram is constructed with processing stations as nodes and process flow relationships as edges. The data is processed using a spatiotemporal graph convolutional network to extract spatiotemporal correlation features that combine the spatial dependencies and temporal evolution laws between stations. Based on the spatiotemporal correlation characteristics, the production line health entropy representing the current operational stability of the production line is calculated, and by comparing it with a preset health baseline, it is determined in real time whether the production line is in an abnormal state; When an abnormal state is detected on a production line, the reverse gradient propagation algorithm is used to locate the abnormal root cause station with the highest contribution by calculating the gradient contribution of the production line health entropy to the input features of each station. The spatiotemporal correlation features are input into a trajectory prediction module based on Kalman filtering and attention mechanism to generate an evolution trajectory of the production line health entropy within a future preset time window. The evolution trajectory includes a predicted value and a confidence interval, which is used to provide early warning of potential deterioration of the production line status.

2. The method for monitoring a fan controller processing line according to claim 1, wherein: The acquisition of real-time physical parameters, manufacturing execution system data, and video image data of each processing station on the production line to form multimodal time series data includes: The physical parameters of each workstation are collected through an industrial controller; the manufacturing data associated with the production batch and workpiece is obtained through the manufacturing execution system; the image data of the workstation operation is captured through an industrial camera; and a unified timestamp is used to time-align the physical parameters, manufacturing data and image data to form multimodal time series data.

3. The method for monitoring a fan controller processing line according to claim 1, wherein: The spatiotemporal graph convolutional network consists of alternately stacked graph convolution modules and temporal convolution modules. The graph convolution module is used to aggregate the information of adjacent workstation nodes in the production line topology diagram at each time step to capture spatial dependencies. The temporal convolution module is used to extract the dynamic change rules of each workstation node's own multimodal time series data to capture temporal dependencies.

4. The method for monitoring a fan controller processing line according to claim 1, wherein: The calculation of the production line health entropy representing the current operational stability of the production line includes: A clustering algorithm is used to group the spatiotemporal correlation feature vectors of all workstations within the current time window to obtain multiple clusters. The proportion of the number of feature vectors in each cluster to the total number is counted to form a discrete probability distribution. Based on the probability distribution, the Shannon entropy is calculated as the production line health entropy.

5. The method for monitoring a production line of a fan controller according to claim 1, wherein: The calculation of the production line health entropy, which characterizes the current operating stability of the production line, includes: inputting the extracted spatiotemporal correlation features into a multi-layer perceptron, whose output layer uses a Softmax function to calculate the probability distribution of the production line in various preset states, where the types of preset states include normal, slightly deviated, and seriously abnormal; then calculating the information entropy of the probability distribution according to the Shannon entropy formula to obtain a single scalar value, namely the production line health entropy.

6. The method for monitoring a processing line of a fan controller according to claim 1, wherein: The health baseline is determined based on the statistical distribution of the production line health entropy calculated during multiple historical normal operation cycles of the production line; when the real-time calculated production line health entropy is higher than a certain proportion of the health baseline, it is judged that the production line is in an abnormal state.

7. The method for monitoring a production line of a fan controller according to claim 1, wherein: Real-time judgment of whether the production line is in an abnormal state includes: During the normal operation cycle of the production line, a large number of health entropy values ​​are collected, and their mean and standard deviation are obtained through Gaussian distribution fitting. The mean plus three times the standard deviation is set as the dynamic threshold for abnormality judgment. If the real-time calculated health entropy of the production line exceeds the dynamic threshold, it is judged as abnormal.

8. The method for monitoring a production line of a fan controller according to claim 1, wherein: The abnormal root cause stations with the highest contribution to positioning include: Taking the health entropy of the production line as the objective function, backpropagation is performed along the spatiotemporal graph convolutional network to calculate the gradient of the health entropy with respect to the input features of each workstation within a preset time window before the abnormality occurs. The absolute values ​​of the gradients of all input features of each workstation are accumulated to obtain the total gradient contribution of the workstation. The total gradient contributions of all workstations are ranked, and the workstation with the highest contribution is identified as the root cause of the abnormality.

9. The method for monitoring a fan controller processing line according to any one of claims 1 to 8, wherein: The evolution trajectory of the production line health entropy within a preset future time window is generated by: The attention mechanism is used to process the historical spatiotemporal correlation feature sequence to assign weights to the spatiotemporal correlation features of each time step; The feature sequence weighted by the attention mechanism is input into the Kalman filter. Through iterative prediction and update steps, a sequence of predicted values ​​of the production line health entropy within a preset future time window is generated. The state covariance matrix of the Kalman filter is used to calculate the confidence interval for each predicted value.

10. A production line monitoring system for a fan controller, comprising a memory and a processor, wherein the memory stores computer program instructions, characterized in that: When the computer program instructions are executed by a processor, the processing production line monitoring method for a fan controller according to any one of claims 1 to 9 is implemented.

Images