A fault monitoring method and system for intelligent plate processing workshop
Through real-time monitoring and data fusion technology, the faults of intelligent production equipment in the intelligent board processing workshop are quickly identified, which solves the problem that equipment failure is difficult to quickly identify, and improves the fault response speed and the stability of the production line.
Patent Information
- Application Number
- CN202411133413.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-19
AI Technical Summary
When smart production equipment fails in smart board processing workshops, it is difficult to quickly identify and deal with it in time, resulting in interruption of production processes and economic losses.
By obtaining the operation status of the three-color alarm lights of the key stations, filtering out the target stations, and obtaining discrete and continuous data of different sensors from the database, data clustering and comparison are carried out, inter-class quantity constraints are introduced, abnormal data is identified, and Perform data fusion to obtain fault monitoring results.
It significantly improves fault positioning efficiency and response speed, reduces production stagnation time, provides accurate equipment status information, ensures timely handling and prevention of faults, and improves the stability and efficiency of the production line.
Smart Images

Figure CN118656762B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and in particular to a fault monitoring method and system for an intelligent plate processing workshop. Background Art
[0002] With the continuous advancement of automation technology, most of the workstations in the smart production workshops of the custom home furnishing production industry have been unmanned, and only employees are set up in specific locations to operate, such as the packaging stations in the back section. However, this production model also brings new challenges. In particular, when smart production equipment fails or is abnormal, it is often difficult for the workshop to quickly identify and promptly notify relevant personnel to handle it. This delay may cause equipment damage and interruption of the production process, resulting in production delays. What's more serious is that these problems will not only directly affect production efficiency, but may also cause significant economic losses. In order to reduce these potential risks, it is necessary to further optimize the equipment fault monitoring system and improve the response speed to ensure the stable operation of the production line and the economic benefits of the enterprise. Summary of the invention
[0003] The main purpose of the embodiments of the present invention is to provide a fault monitoring method and system for an intelligent plate processing workshop, aiming to solve the problem in the related art that it is difficult to quickly identify and promptly handle faults or abnormalities in intelligent production equipment.
[0004] In a first aspect, an embodiment of the present invention provides a fault monitoring method for an intelligent plate processing workshop, comprising:
[0005] Obtaining the operating status corresponding to the three-color alarm light of the key workstation, and performing abnormal screening on the key workstation according to the operating status to determine the target workstation;
[0006] Obtaining, from a database, first data corresponding to the target workstation under a first sensor and second data corresponding to the target workstation under a second sensor, wherein the first data is discrete data of an associated device corresponding to the target workstation, and the second data is continuous data of the associated device corresponding to the target workstation;
[0007] In the process of clustering the first data, a quantity constraint between classes is introduced to obtain first abnormal data corresponding to the first sensor;
[0008] Comparing the reference data with the second data to obtain second abnormal data corresponding to the second sensor;
[0009] The first abnormal data and the second abnormal data are data-fused to obtain a fault monitoring result corresponding to the associated device.
[0010] In a second aspect, an embodiment of the present invention provides a fault monitoring system for an intelligent plate processing workshop, comprising:
[0011] A workstation determination module is used to obtain the operating status corresponding to the three-color alarm light of the key workstation, and perform abnormal screening on the key workstation according to the operating status to determine the target workstation;
[0012] A data acquisition module, used to obtain from a database first data corresponding to the target station under the first sensor and second data corresponding to the second sensor, wherein the first data is discrete data of an associated device corresponding to the target station, and the second data is continuous data of the associated device corresponding to the target station;
[0013] a first identification module, configured to introduce a quantity constraint between classes during data clustering of the first data, and obtain first abnormal data corresponding to the first sensor;
[0014] A second identification module, configured to compare the reference data with the second data to obtain second abnormal data corresponding to the second sensor;
[0015] The fault determination module is used to fuse the first abnormal data and the second abnormal data to obtain a fault monitoring result corresponding to the associated device.
[0016] The embodiment of the present invention provides a fault monitoring method and system for an intelligent plate processing workshop, the method comprising obtaining the operating status corresponding to the three-color alarm light of the key station, and being able to quickly screen out the target station with abnormalities. This real-time monitoring method can greatly improve the efficiency of fault location, so that the problem is discovered in time, and the time of production stagnation is reduced. The discrete data and continuous data of the target station under different sensors are extracted from the database, so that the operating status of the equipment can be fully understood. This detailed data support not only helps technicians to conduct in-depth analysis, but also provides accurate equipment operating status information, which is helpful for formulating effective maintenance plans and optimization measures. When clustering the data of the first sensor, the number constraint between classes is introduced, and the first abnormal data of the first sensor can be effectively identified. This method improves the sensitivity and accuracy of detection, reduces the false alarm rate, and ensures the reliability of abnormal data. By comparing the reference data with the data of the second sensor, the second abnormal data of the second sensor can be accurately identified. This comparative analysis further improves the accuracy of fault diagnosis and ensures comprehensive monitoring of the equipment status. The first abnormal data and the second abnormal data are fused to comprehensively analyze the overall status of the equipment. This data fusion method provides a more comprehensive fault monitoring result, makes problem identification more reliable, and helps to carry out maintenance and optimization in time to prevent the spread of equipment failure. This method effectively solves the problem in related technologies that it is difficult to quickly identify and promptly handle faults or anomalies in intelligent production equipment. Through real-time data monitoring, anomaly detection and data fusion, the speed of fault response and processing effect are significantly improved, thereby improving the stability and efficiency of the production line. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0018] Figure 1 A schematic flow chart of a fault monitoring method for an intelligent plate processing workshop provided by an embodiment of the present invention;
[0019] Figure 2 A schematic diagram of the module structure of a fault monitoring system for an intelligent plate processing workshop provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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.
[0021] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.
[0022] It should be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.
[0023] The embodiment of the present invention provides a fault monitoring method and system for a smart plate processing workshop. The fault monitoring method for a smart plate processing workshop can be applied to a terminal device, which can be an electronic device such as a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device. The terminal device can be a server or a server cluster.
[0024] Some embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.
[0025] Please refer to Figure 1 , Figure 1 A schematic flow chart of a fault monitoring method for an intelligent plate processing workshop provided in an embodiment of the present invention.
[0026] like Figure 1 As shown, the intelligent plate processing workshop fault monitoring method includes steps S101 to S105.
[0027] Step S101: obtaining the operating status corresponding to the three-color alarm light of the key workstation, and performing abnormal screening on the key workstation according to the operating status to determine the target workstation.
[0028] Exemplarily, ensure that each key station is equipped with a three-color alarm light (such as red, yellow, and green), and that the alarm light system is properly connected to the control system of the associated station, and configure the triggering conditions of the alarm light so that it can send a signal of the corresponding color when a failure or abnormality occurs at the key station. Collect the operating status corresponding to the three-color alarm light of each key station in real time. This can be done by automatically reading the current color of the three-color alarm light through a sensor or a control system, and then determining the operating status corresponding to the key station based on the current color. For example, when the current color of the three-color alarm light is red, it indicates that the operating status is a serious failure, when the current color of the three-color alarm light is yellow, it indicates that the operating status is a warning state, and when the current color of the three-color alarm light is green, it indicates that the operating status is normal operation.
[0029] For example, when the operation status is normal operation, the corresponding key stations are eliminated, so that the remaining key stations are determined as target stations, that is, the key stations when the operation status is a warning state or a serious fault are determined as target stations. In addition, when the operation status of the key stations includes a warning state and a serious fault, the key stations with a serious fault operation status are preferentially determined as target stations.
[0030] Step S102, obtaining first data corresponding to the target workstation under the first sensor and second data corresponding to the second sensor from a database, wherein the first data is discrete data of an associated device corresponding to the target workstation, and the second data is continuous data of the associated device corresponding to the target workstation.
[0031] Exemplarily, it is ensured that the sensors on the associated equipment corresponding to the target station can upload data to the database in real time or regularly. Then, when the target station detects an abnormal operation state, the first data corresponding to the first sensor and the second data corresponding to the second sensor of the associated equipment in the operation state are obtained from the database.
[0032] Exemplarily, the first data is the snapshot data collected by the first sensor at a specified time point, that is, the first data is the discrete data of the associated equipment corresponding to the target station. The second data is the stream data collected by the second sensor in real time or at intervals, that is, the second data is the continuous data of the associated equipment corresponding to the target station.
[0033] Step S103: introducing a quantity constraint between classes during data clustering of the first data to obtain first abnormal data corresponding to the first sensor.
[0034] Exemplarily, determine the goal of clustering, such as identifying abnormal data points or discovering patterns in the data. Clarify the number of classes that need to be identified or the characteristics of abnormal data. Set the number constraints between classes, that is, the quantitative relationship between each category in the cluster analysis. This may include limiting the amount of data in each category, or requiring the proportional relationship between categories to meet specific requirements. Then choose a suitable clustering algorithm to process the data. Common algorithms include K-means, DBSCAN, hierarchical clustering, etc. Different algorithms have different applicable scenarios, and the characteristics of the data and the number constraints between classes need to be considered when choosing.
[0035] Exemplarily, when clustering, the number constraints between classes are applied to the algorithm. For example, if K-means is used, constraints can be added during the optimization process to control the size of each cluster or the proportional relationship between classes. The operation mechanism of the algorithm is adjusted according to the number constraints between classes to ensure that the final clustering results meet the constraint requirements. It may be necessary to modify the standard operation of the algorithm or design a custom constraint handling strategy.
[0036] Exemplarily, the clustering results are analyzed to identify clusters that are significantly different from most data points. These clusters usually contain abnormal data points, that is, data points that are significantly different from the normal pattern. The identified abnormal data are compared with the preset standard to confirm whether they truly represent abnormal situations, thereby obtaining the first abnormal data corresponding to the first sensor. When the cluster analysis is performed on the first data, the number constraints between classes are introduced, and the accuracy and stability of anomaly detection can be improved when the abnormal data is identified.
[0037] In some embodiments, the method of introducing an inter-class quantity constraint in the process of clustering the first data to obtain the first abnormal data corresponding to the first sensor includes: determining an initial cluster center and determining the membership between the first data and the initial cluster center; calculating the first distance information between the first data and the initial cluster center; determining a first loss value corresponding to the initial cluster center according to the first distance information and the membership; setting the inter-class quantity constraint according to the membership to determine a second loss value corresponding to the initial cluster center; performing loss fusion on the first loss value and the second loss value to obtain a target loss value corresponding to the initial cluster center; updating the initial cluster center according to the target loss value to obtain a target cluster center corresponding to the first data and cluster data corresponding to the target cluster center; performing abnormal classification on the cluster data to obtain the first abnormal data corresponding to the first sensor; wherein the first loss value is obtained according to the following formula:
[0038]
[0039] represents the first loss value, N represents the number of the initial cluster centers, sum represents the number of the first data, represents the degree of membership between the nth first data and the mth initial clustering center, represents the degree of membership between the i-th first data and the m-th initial cluster center, represents the first distance information between the nth first data and the mth initial clustering center, and t represents a constant.
[0040] Exemplarily, the initial cluster centers are determined by random selection, K-means++ initialization or other heuristic methods. These initial cluster centers serve as references for data points for subsequent clustering adjustments.
[0041] Exemplarily, the distance from each data point to all initial cluster centers is calculated. Commonly used distance metrics include Euclidean distance, Manhattan distance, etc., so as to calculate the membership of each data point to each initial cluster center based on the distance. The membership indicates the degree to which a data point belongs to a cluster, which can be represented by a weight value, and is usually calculated by the membership function in the fuzzy clustering algorithm. The distance information from each data point to the initial cluster center is recorded to obtain the first distance information. These first distance information are used for subsequent loss calculation and cluster center update.
[0042] Exemplarily, the first loss value is calculated using the defined loss function. The first loss value generally represents the distance error between the data point and the initial cluster center, and is used to evaluate the initial quality of the clustering, and then the first loss value is obtained according to the following formula:
[0043]
[0044] represents the first loss value, N represents the number of initial cluster centers, sum represents the number of first data, represents the degree of membership between the nth first data and the mth initial cluster center, represents the degree of membership between the i-th first data and the m-th initial cluster center, It represents the first distance information between the nth first data and the mth initial cluster center, and t represents a constant. In general, t takes a positive integer, which can be 1 or 2.
[0045] The above formula takes into account the membership of each data point to all initial cluster centers, which can improve the robustness of the model. By comprehensively considering the membership and the first distance information, the model can be more adaptable to changes and anomalies of data points, thereby reducing sensitivity to noise and outliers. By calculating the first loss value, the quality of clustering can be quantified. The size of the loss value directly reflects the error between the data point and the cluster center, helps to judge the quality of the current clustering result, and provides a basis for further optimization.
[0046] Exemplarily, the inter-class quantity constraint is set according to the specific application, such as the range of the number of data points in each cluster, the density of the cluster, etc. The inter-class quantity constraint is included in the calculation to evaluate the constraint compliance of the initial cluster center. The second loss value reflects the satisfaction degree of the inter-class quantity constraint.
[0047] Exemplarily, the first loss value and the second loss value are fused into a target loss value. The fusion method can be a simple weighting, a linear combination, or other suitable strategy for comprehensively evaluating the optimization effect of the cluster center. The position of the cluster center is optimized according to the target loss value. This usually involves iteratively adjusting the initial cluster center to minimize the target loss value. During the update process, the new cluster center positions are calculated to ensure that they can better represent the cluster structure of the data points, and then the data points are assigned to the new clusters according to the updated cluster centers. Each data point will be assigned to the cluster corresponding to the cluster center closest to it, so as to obtain the target cluster center and the cluster data corresponding to each target cluster center.
[0048] Exemplarily, the number of data points, mean, standard deviation, etc. of each cluster are counted according to the clustering results. These statistical information helps to understand the internal structure and distribution of the cluster. According to actual needs, a suitable anomaly detection method is selected, such as distance-based anomaly detection, density-based anomaly detection, statistics-based anomaly detection, etc. Therefore, according to the selected anomaly detection method, the anomaly threshold is determined. For example, points whose distance exceeds a certain threshold can be regarded as anomalies; or in a statistical distribution, data points falling in the tail area can be regarded as anomalies. Therefore, according to the calculated anomaly score and the set threshold, those data points that are considered to be abnormal are marked, and then the first abnormal data corresponding to the first sensor is obtained.
[0049] Specifically, by setting the number constraints between classes, we can ensure that the size or number ratio of each cluster meets the preset business or technical requirements. This helps to meet specific requirements in practical applications, such as processing capacity limitations or business rules. This strategy enables the model to adjust the loss function and constraints according to different data sets and business needs, thereby improving the flexibility and adaptability of the model. Furthermore, by conducting a detailed analysis of clustering data and anomaly classification, we can provide decision makers with more accurate information support and help them make more informed decisions.
[0050] In some embodiments, the first loss value and the second loss value are subjected to loss fusion to obtain a target loss value corresponding to the initial cluster center, including: performing a ratio operation on the first loss value and the second loss value to obtain a ratio result; obtaining the number of clustering times corresponding to the initial cluster center, and determining an adjustment parameter corresponding to the initial cluster center based on the clustering number; determining a fusion weight corresponding to the second loss value based on the ratio result and the adjustment parameter; and performing loss fusion on the first loss value and the second loss value based on the fusion weight to obtain the target loss value corresponding to the initial cluster center.
[0051] Exemplarily, the first loss value of the initial cluster center is calculated, and the second loss value related to the initial cluster center is calculated according to the inter-class quantity constraint, and the first loss value is divided by the second loss value to obtain a ratio result. This ratio result reflects the relative size of the two losses. The number of iterations corresponding to the initial cluster center is obtained, that is, the number of clustering is obtained, and then according to the number of clustering of each initial cluster center, and then according to the ratio between the number of clustering and the preset number, the attenuation parameter corresponding to the ratio is determined using the Euler number. Finally, the adjustment parameter is set based on the attenuation parameter. The purpose of adjusting the parameter is to adjust the loss calculation according to the actual clustering situation.
[0052] Exemplarily, after multiplying the ratio result and the adjustment parameter, the product result is determined as a fusion weight. The fusion weight is used to balance the contribution of the second loss value so as to be used in loss fusion.
[0053] Exemplarily, the calculated fusion weight is used to multiply the fusion weight and the second loss value, and then the multiplication result is added to the first loss value to obtain a target loss value. This target loss value comprehensively considers the first loss value and the second loss value, and reflects the overall performance of the initial cluster center.
[0054] Specifically, the adjustment parameters are determined by the number of clustering times so that the model can adapt to the actual data distribution. This can improve the quality of clustering and make the final cluster center more representative and stable. The fusion weight is determined by the ratio result and the adjustment parameter, so that the model can dynamically adjust the loss weight according to the specific situation. This adaptive adjustment mechanism can improve the robustness and generalization ability of the model for different data sets. Therefore, the fused target loss value can more accurately reflect the actual performance of the initial cluster center, providing support for obtaining accurate clustering results in the future.
[0055] Step S104: Compare the reference data with the second data to obtain second abnormal data corresponding to the second sensor.
[0056] Exemplarily, the reference data is normal data, and then the reference data is standardized or normalized for comparison. The features required for comparison are extracted from the reference data and the second data. Features can be selected by methods such as correlation analysis and principal component analysis (PCA). New features are constructed as needed, which may include statistical features of the data (mean, variance, etc.), time series features, frequency domain features, etc.
[0057] Exemplarily, a suitable model is selected to compare the reference data and the second data. For example, common models include statistical anomaly detection models (such as Z-Score, Grubbs test), machine learning-based anomaly detection models (such as isolation forest, local outlier factor (LOF)), or deep learning models (such as autoencoders). Use the reference data to train the model. If a machine learning or deep learning model is used, the reference data can be divided into a training set and a validation set to facilitate adjustment of model parameters.
[0058] Exemplarily, the trained model is applied to the second data, and the anomaly score or probability of each data point is calculated. According to the output of the model and the actual application requirements, a threshold for anomaly detection is set. The threshold can be determined by statistical methods (such as based on the mean and standard deviation) or by evaluation of the model (such as ROC curve). The data points in the second data are compared with the threshold to identify the second abnormal data.
[0059] In some embodiments, the comparing the reference data and the second data to obtain the second abnormal data corresponding to the second sensor includes: taking the reference data as the first coordinate point corresponding to the horizontal axis and taking the second data as the second coordinate point corresponding to the vertical axis; calculating the difference between the first coordinate point and the second coordinate point, and counting the distribution corresponding to the difference; determining the delay parameters corresponding to the reference data and the second data according to the distribution; aligning the reference data and the second data according to the delay parameters to obtain the difference data between the second data and the reference data; determining the change characteristics and deviation characteristics corresponding to the second data according to the difference data and the delay parameters, the change characteristics being used to characterize the change of the second data, and the deviation characteristics being used to characterize the deviation phenomenon corresponding to the second data; screening from the second data according to the change characteristics and the deviation characteristics to obtain the second abnormal data corresponding to the second sensor.
[0060] Exemplarily, the reference data is used as the coordinate point of the horizontal axis (X axis), and the second data is used as the coordinate point of the vertical axis (Y axis). This can form a two-dimensional coordinate system, in which each pair of reference data and second data corresponds to a coordinate point (x, y). The difference between each pair of coordinate points (x_i, y_i) is calculated, such as y_i - x_i, which can represent the deviation of the second data relative to the reference data.
[0061] Exemplarily, all calculated differences are counted and their distribution is analyzed. For example, statistics such as mean, variance, standard deviation, etc. can be calculated, and a histogram or probability density function diagram can be drawn. Determine the delay parameter based on the distribution of the difference. This can be a decision based on statistical characteristics (such as mean and standard deviation), or estimate the delay parameter through a model (such as an autoregressive model). The delay parameter is used to adjust the time alignment of the reference data and the second data for accurate comparison.
[0062] Exemplarily, the reference data and the second data are time-aligned according to the delay parameter. The data can be aligned using methods such as interpolation and translation. Difference data is calculated in the aligned data. That is, the difference between the aligned reference data and the second data is calculated, and the change of the second data is analyzed based on the difference data. The rate of change, trend line, etc. of the difference data can be calculated to identify the change characteristics of the data. Thereby, the deviation characteristics of the difference data, such as outliers, deviation, etc., are calculated to describe the deviation of the second data relative to the reference data.
[0063] Exemplarily, a change feature score is calculated for each data point. For example, a data point whose change rate exceeds a preset threshold can be marked as abnormal, and a deviation feature score is calculated for each data point. A data point whose deviation exceeds a normal range can be marked as abnormal. According to a set threshold, the change feature score and the deviation feature score are compared, and the data point exceeding the threshold is marked as abnormal, thereby obtaining second abnormal data corresponding to the second sensor.
[0064] Specifically, by accurately calculating the delay parameters, the time misalignment can be eliminated, so that the reference data and the second data are completely aligned in time. This ensures that subsequent analysis is based on accurate time synchronization data, thereby reducing errors caused by delays. The aligned data can more realistically reflect the relationship between the two, ensuring that the calculation of the difference data is accurate, making the results of anomaly detection and feature analysis more reliable, and then by comparing the change features and deviation features, abnormal patterns in the data can be identified, such as sudden changes or deviations from long-term trends. This helps to detect abnormal situations in a timely manner and avoid potential problems.
[0065] Step S105: fusing the first abnormal data and the second abnormal data to obtain a fault monitoring result corresponding to the associated device.
[0066] Exemplarily, key features are extracted from the first abnormal data and the second abnormal data, and these features can effectively describe the fault status of the associated equipment. For example, the frequency and severity of the abnormal values can be extracted. Select features that have a significant impact on the fault monitoring results, avoid redundant information, and simplify the data fusion process. Then perform a weighted average on the abnormal data from different sources to comprehensively consider the impact of each data source. Then use a machine learning model (such as a multivariate classification model) to fuse the abnormal data and combine the information from different data sources for prediction or classification. Based on the fused data, a comprehensive analysis is performed to generate the fault monitoring results of the associated equipment. This may include the type, severity, and time of occurrence of the fault.
[0067] In some embodiments, the data fusion of the first abnormal data and the second abnormal data to obtain the fault monitoring result corresponding to the associated device includes: calculating the degree of association between the first abnormal data and the reference abnormal data, and determining the first abnormal type corresponding to the first abnormal data according to the degree of association; decomposing and reconstructing the second abnormal data to obtain multidimensional features corresponding to the second abnormal data, and classifying the second abnormal data according to the multidimensional features to obtain a second abnormal type; and fusing the first abnormal type and the second abnormal type to obtain the fault monitoring result corresponding to the associated device.
[0068] Exemplarily, the reference abnormal data is known abnormal data corresponding to a known abnormal type, and a common similarity measurement method (such as cosine similarity, Jaccard similarity coefficient, Euclidean distance or other distance measurement) is used to calculate the similarity between the first abnormal data and the reference abnormal data. Alternatively, a clustering algorithm (such as K-means, DBSCAN, etc.) is used to cluster the abnormal data, and the probability that the first abnormal data and the reference abnormal data are in the same cluster is checked, and the degree of association between the first abnormal data and the reference abnormal data is determined based on the probability or similarity.
[0069] Exemplarily, the known abnormality type of the reference abnormality data corresponding to the maximum correlation degree is determined as the first abnormality type corresponding to the first abnormality data.
[0070] Exemplarily, the second abnormal data is decomposed using methods such as Fourier transform and wavelet transform, and important features are extracted from the decomposition results, so that deep learning models (such as autoencoders, convolutional neural networks, etc.) are used to extract and represent data features to obtain multi-dimensional features, and then deep learning models such as convolutional neural networks (CNN), recurrent neural networks (RNN) or long short-term memory networks (LSTM) are used to learn and classify the features of the second abnormal data, thereby determining the second abnormal type corresponding to the second abnormal data.
[0071] Exemplarily, a fusion rule is determined, such as a priority rule. If the first exception type has a higher priority, the first exception type is given priority, and if the second exception type has a higher priority, the second exception type is given priority. For example, a conflict resolution rule is determined. When the first exception type and the second exception type are inconsistent, the conflict is resolved through expert rules or historical data to determine the final exception type.
[0072] Exemplarily, the abnormal time corresponding to the abnormality of the associated device is determined based on the first abnormal data and the second abnormal data, and then the abnormal description statement corresponding to the associated device is generated according to the abnormal time and the final abnormal type, so that the abnormal description statement is determined as the fault monitoring result corresponding to the associated device.
[0073] Specifically, by calculating the degree of association, the type of the first abnormal data can be more accurately identified, which helps to more clearly understand and classify abnormal situations. Decomposing and reconstructing the second abnormal data and extracting multidimensional features can reveal more complex abnormal patterns, thereby enhancing the ability to classify the second abnormal data and improving the accuracy of classification. Fusion of abnormal information from different sources and different types makes fault monitoring more comprehensive and can cover more potential fault situations. By combining the two types of abnormal information, equipment failures can be more comprehensively identified and monitored, reducing the possibility of missed detection.
[0074] In some embodiments, the calculating the degree of association between the first abnormal data and the reference abnormal data includes: obtaining the reference abnormal data from an abnormal reference set, and then performing matrix representation on the reference abnormal data to obtain a corresponding first matrix; performing matrix representation on the first abnormal data to obtain a corresponding second matrix; calculating an association parameter and a difference parameter between the first matrix and the second matrix, and calculating second distance information between the first matrix and the second matrix; determining the degree of association between the first abnormal data and the reference abnormal data according to the association parameter, the difference parameter and the second distance information; wherein the degree of association is obtained according to the following formula:
[0075]
[0076] r represents the degree of association, relation represents the association parameter, diff represents the difference parameter, represents the first matrix, represents the second matrix, dis represents the second distance information, and m represents the number corresponding to the first abnormal data.
[0077] Exemplarily, reference anomaly data related to the target analysis is selected from the anomaly reference set. These data can come from historical records, known failure cases, or expert-defined anomaly patterns. The reference anomaly data is preprocessed, including noise removal, missing value filling, and standardization to ensure data quality and consistency.
[0078] Exemplarily, key features are extracted from the reference abnormal data, and the extracted features are organized into a matrix form to obtain a first matrix. Similarly, feature extraction is performed on the first abnormal data to obtain features that are the same or similar to the reference abnormal data. The features of the first abnormal data are organized into a matrix to obtain a second matrix. The construction method should be consistent with the reference abnormal data matrix for comparison and analysis.
[0079] Exemplarily, the Pearson correlation coefficient is used to measure the association parameters between corresponding elements in the first matrix and the second matrix, and the degree of difference between the first matrix and the second matrix is calculated. For example, the mean square error is used to measure the difference between the first matrix and the second matrix, and the mean of the sum of squares of the differences between corresponding elements of the first matrix and the second matrix is calculated to determine the difference parameter.
[0080] Exemplarily, the second distance information between the first matrix and the second matrix is calculated using the Manhattan distance, and then the correlation parameter, the difference parameter and the second distance information are comprehensively considered to determine the correlation degree between the first abnormal data and the reference abnormal data. Wherein, the correlation degree is obtained according to the following formula:
[0081]
[0082] r represents the degree of association, relation represents the association parameter, and diff represents the difference parameter. represents the first matrix, represents the second matrix, dis represents the second distance information, and m represents the number corresponding to the first abnormal data.
[0083] The above formula combines the correlation parameter, the difference parameter and the second distance information to quantify the degree of correlation between the first matrix and the second matrix. This comprehensive evaluation helps to understand the relationship between the data more comprehensively. By combining different indicators to calculate the degree of correlation, the similarities and differences of the data can be more accurately reflected, avoiding the deviation that may be caused by a single indicator. Therefore, by calculating the degree of correlation, the relationship between the first abnormal data and the reference abnormal data can be better identified, thereby locating the cause of the fault more accurately.
[0084] Specifically, the calculation of the association parameter and the difference parameter can accurately evaluate the similarity and difference between the first matrix and the second matrix. This evaluation helps to understand the characteristics and patterns of the abnormal data. By analyzing the association parameter and the difference parameter, it is possible to better identify whether the first abnormal data has a similar abnormal pattern to the reference abnormal data, thereby improving the accuracy of abnormality detection, and then combining the association parameter, the difference parameter and the second distance information to provide a comprehensive correlation degree indicator. This comprehensive analysis method can more comprehensively determine the degree of correlation between the first abnormal data and the reference abnormal data.
[0085] In some embodiments, the fusing of the first abnormality type and the second abnormality type to obtain the fault monitoring result corresponding to the associated device includes: determining a first recognition probability corresponding to the first abnormality type according to the degree of association, and determining a second recognition probability corresponding to the second abnormality type; fusing the first recognition probability and the second recognition probability to determine a target abnormality type corresponding to the associated device; and determining the fault monitoring result corresponding to the associated device according to the target abnormality type.
[0086] Exemplarily, the correlation degree between the first abnormal data and the reference abnormal data is determined according to the first matrix and the second matrix, so as to determine the first recognition probability corresponding to the first abnormal type according to the correlation degree. The second recognition probability corresponding to the second abnormal type is obtained according to the abnormal classification model.
[0087] Exemplarily, weights are set to fuse the first recognition probability and the second recognition probability. The weights can be based on factors such as model performance and data reliability. Bayesian theorem is applied to synthesize the first recognition probability and the second recognition probability. Bayesian fusion takes into account prior probability and conditional probability to update the posterior probability. Then rules are set, for example, selecting the anomaly type with the highest recognition probability as the target anomaly type, or using other decision rules (such as the maximum entropy principle) to determine the fused result.
[0088] Exemplarily, according to the selected fusion method, the first recognition probability and the second recognition probability are calculated comprehensively to obtain the final target abnormality type. The target abnormality type is the abnormality type with the highest probability, or the type with the highest fusion probability.
[0089] Exemplarily, a mapping relationship between the target abnormality type and the fault monitoring result is established. This is usually based on experience or historical data. For example, a certain abnormality type corresponds to a specific fault mode and processing method, so that the corresponding fault monitoring result is queried from the mapping relationship according to the target abnormality type. The fault monitoring result includes the target abnormality type and the abnormality processing method corresponding to the target abnormality type.
[0090] Specifically, fusing the first recognition probability and the second recognition probability helps to combine information from different sources to more comprehensively evaluate the abnormal situation. This can reduce the reliance on a single source of information and enhance the robustness of the system. By fusing multiple recognition probabilities, the target abnormality type can be identified more accurately, reducing the probability of misjudgment and missed reports.
[0091] In some embodiments, after obtaining the fault monitoring result corresponding to the associated device, it also includes: obtaining the target user corresponding to the associated device and the target position corresponding to the associated device from a database; obtaining the corresponding exception description statement using a text generation algorithm according to the target position and the fault monitoring result; and sending the exception description statement to the target device corresponding to the target user, so that the target user receives the exception description statement of the associated device according to the target device, and then performs exception processing according to the exception description statement.
[0092] Exemplarily, the target user corresponding to the associated equipment is queried from the database, and the target user is the staff responsible for the inspection and maintenance of the associated equipment. The target position corresponding to the associated equipment is also queried from the database, and the target position is the location information of the associated equipment in the target production line, so that the target user can quickly reach the corresponding position to inspect and repair the equipment.
[0093] For example, a description template including the fault monitoring result and the target location is designed. For example: "Device [name of the associated device] experienced [fault monitoring result] at [target location]. " Thus, the target location and the fault monitoring result are filled into the description template to obtain the abnormal description statement of the associated device.
[0094] For example, a suitable transmission method is selected according to the communication method supported by the target user (such as email, SMS, push notification, instant message, etc.), and the generated exception description statement is sent to the target user through the communication method corresponding to the target user, ensuring that the message is correctly delivered to the terminal device of the target user and can be viewed by the target user in a timely manner. The target user can then take corresponding processing measures according to the received exception description statement, such as contacting technical support, performing troubleshooting operations, etc.
[0095] Specifically, the efficiency of information transmission is improved by automatically generating and sending exception description statements. Text generation algorithms and database automation processing are used to reduce errors in manual operations. Exception information is immediately communicated to relevant personnel to help quickly respond to and handle faults. Clear exception descriptions are provided to help users quickly understand the fault situation, so that fault handling can be carried out more effectively. This ensures that the entire process from the generation to the transmission of exception information is completed efficiently and accurately.
[0096] In some embodiments, it also includes: collecting video monitoring results corresponding to the video monitoring equipment of the key workstation; presenting the video monitoring results in real time in a terminal device that is communicatively connected to the video monitoring equipment, so that the terminal device can display the operating results of the key workstation in real time according to the video monitoring results.
[0097] For example, ensure that the surveillance cameras at all key workstations are properly installed and functioning properly. Set the output format and transmission protocol of the video surveillance device to enable effective data communication with the terminal device. Select an appropriate transmission protocol (such as RTSP, RTMP, HTTP, etc.) to obtain video stream data. Ensure that the surveillance device supports the selected protocol. Ensure that the network connection between the video surveillance device and the terminal device is stable and can support real-time data transmission.
[0098] For example, a terminal device suitable for video display is selected, such as a computer, a smart phone, a tablet or a dedicated monitoring display. Depending on the type of terminal device, corresponding software or applications are configured to receive and display the video stream. Settings for receiving the video stream are configured on the terminal device, including inputting the URL or stream address of the video stream. Ensure that the terminal device can receive video data from the video monitoring device. The real-time video stream is displayed on the screen of the terminal device through the configured software or application.
[0099] For example, design a user interface based on actual needs to ensure that the video monitoring results are clearly visible on the terminal device. This includes setting the video window size, clarity, resolution, etc. Provide users with a control interface so that they can perform basic operations on the video stream, such as play, pause, zoom, etc. Regularly check the quality and stability of the video stream to ensure that the terminal device can continue to receive high-quality real-time video. If there is a video stream interruption or quality problem, troubleshoot and repair it in a timely manner to ensure the continuity and accuracy of video monitoring.
[0100] Specifically, configure video surveillance equipment and terminal devices to establish a stable video stream transmission channel. Configure appropriate software or applications on the terminal devices to display video surveillance results in real time. Optimize the display interface on the terminal devices to ensure that the monitoring results are clearly visible and provide necessary operation functions. In this way, the terminal devices can display the operating status of key workstations in real time, helping users monitor and manage the operation of key workstations.
[0101] The embodiment of the present invention provides a fault monitoring method and system for an intelligent plate processing workshop, the method comprising obtaining the operating status corresponding to the three-color alarm light of the key station, and being able to quickly screen out the target station with abnormalities. This real-time monitoring method can greatly improve the efficiency of fault location, so that the problem is discovered in time, and the time of production stagnation is reduced. The discrete data and continuous data of the target station under different sensors are extracted from the database, so that the operating status of the equipment can be fully understood. This detailed data support not only helps technicians to conduct in-depth analysis, but also provides accurate equipment operating status information, which is helpful for formulating effective maintenance plans and optimization measures. When clustering the data of the first sensor, the number constraint between classes is introduced, and the first abnormal data of the first sensor can be effectively identified. This method improves the sensitivity and accuracy of detection, reduces the false alarm rate, and ensures the reliability of abnormal data. By comparing the reference data with the data of the second sensor, the second abnormal data of the second sensor can be accurately identified. This comparative analysis further improves the accuracy of fault diagnosis and ensures comprehensive monitoring of the equipment status. The first abnormal data and the second abnormal data are fused to comprehensively analyze the overall status of the equipment. This data fusion method provides a more comprehensive fault monitoring result, makes problem identification more reliable, and helps to carry out maintenance and optimization in time to prevent the spread of equipment failure. This method effectively solves the problem in related technologies that it is difficult to quickly identify and promptly handle faults or anomalies in intelligent production equipment. Through real-time data monitoring, anomaly detection and data fusion, the speed of fault response and processing effect are significantly improved, thereby improving the stability and efficiency of the production line.
[0102] See also Figure 2 , Figure 2The present application provides an intelligent plate processing workshop fault monitoring system 200, which includes a station determination module 201, a data acquisition module 202, a first identification module 203, a second identification module 204, and a fault determination module 205, wherein the station determination module 201 is used to obtain the operating status corresponding to the three-color alarm light of the key station, and perform abnormal screening on the key station according to the operating status to determine the target station; the data acquisition module 202 is used to obtain the first data corresponding to the first sensor and the second data corresponding to the second sensor of the target station from the database. Data, wherein the first data is discrete data of the associated equipment corresponding to the target workstation, and the second data is continuous data of the associated equipment corresponding to the target workstation; a first identification module 203 is used to introduce inter-class quantity constraints in the process of data clustering of the first data, and obtain first abnormal data corresponding to the first sensor; a second identification module 204 is used to compare the reference data with the second data to obtain second abnormal data corresponding to the second sensor; a fault determination module 205 is used to fuse the first abnormal data with the second abnormal data to obtain a fault monitoring result corresponding to the associated equipment.
[0103] In some implementations, the first identification module 203 introduces the inter-class quantity constraint in the process of clustering the first data and obtaining the first abnormal data corresponding to the first sensor, and executes:
[0104] Determining an initial cluster center, and determining a degree of membership between the first data and the initial cluster center;
[0105] Calculating first distance information between the first data and the initial cluster center;
[0106] Determine a first loss value corresponding to the initial cluster center according to the first distance information and the degree of membership;
[0107] Setting the inter-class quantity constraint according to the membership degree to determine a second loss value corresponding to the initial cluster center;
[0108] Performing loss fusion on the first loss value and the second loss value to obtain a target loss value corresponding to the initial cluster center;
[0109] The initial cluster center is updated according to the target loss value to obtain a target cluster center corresponding to the first data and cluster data corresponding to the target cluster center;
[0110] Performing abnormal classification on the cluster data to obtain the first abnormal data corresponding to the first sensor;
[0111] The first loss value is obtained according to the following formula:
[0112]
[0113] represents the first loss value, N represents the number of the initial cluster centers, sum represents the number of the first data, represents the degree of membership between the nth first data and the mth initial clustering center, represents the degree of membership between the i-th first data and the m-th initial cluster center, Represents the first distance information between the nth first data and the mth initial clustering center.
[0114] In some implementations, the first recognition module 203 performs, during the process of performing loss fusion on the first loss value and the second loss value to obtain the target loss value corresponding to the initial cluster center:
[0115] Performing a ratio operation on the first loss value and the second loss value to obtain a ratio result;
[0116] Obtaining the number of clusterings corresponding to the initial clustering center, and determining the adjustment parameter corresponding to the initial clustering center according to the number of clusterings;
[0117] Determine a fusion weight corresponding to the second loss value according to the ratio result and the adjustment parameter;
[0118] The first loss value and the second loss value are fused according to the fusion weight to obtain the target loss value corresponding to the initial cluster center.
[0119] In some implementations, the second identification module 204 performs, during the process of comparing the reference data with the second data to obtain the second abnormal data corresponding to the second sensor:
[0120] Taking the reference data as a first coordinate point corresponding to a horizontal axis and taking the second data as a second coordinate point corresponding to a vertical axis;
[0121] Calculating the difference between the first coordinate point and the second coordinate point, and collecting statistics on the distribution of the difference;
[0122] Determine, according to the distribution, delay parameters corresponding to the reference data and the second data;
[0123] Performing data alignment on the reference data and the second data according to the delay parameter to obtain difference data between the second data and the reference data;
[0124] Determine, according to the difference data and the delay parameter, a change feature and a deviation feature corresponding to the second data, wherein the change feature is used to characterize a change of the second data, and the deviation feature is used to characterize a deviation phenomenon corresponding to the second data;
[0125] The second data is screened according to the change feature and the deviation feature to obtain the second abnormal data corresponding to the second sensor.
[0126] In some implementations, the fault determination module 205 performs, during the process of fusing the first abnormal data and the second abnormal data to obtain the fault monitoring result corresponding to the associated device:
[0127] Calculating a correlation degree between the first abnormal data and reference abnormal data, and determining a first abnormal type corresponding to the first abnormal data according to the correlation degree;
[0128] Decomposing and reconstructing the second abnormal data to obtain multidimensional features corresponding to the second abnormal data, and classifying the second abnormal data according to the multidimensional features to obtain a second abnormal type;
[0129] The first abnormality type and the second abnormality type are merged to obtain the fault monitoring result corresponding to the associated device.
[0130] In some implementations, the fault determination module 205, during the process of calculating the degree of association between the first abnormal data and the reference abnormal data, performs:
[0131] Obtain the reference abnormal data from the abnormal reference set, and then perform matrix representation on the reference abnormal data to obtain a corresponding first matrix;
[0132] Performing matrix representation on the first abnormal data to obtain a corresponding second matrix;
[0133] Calculating an association parameter and a difference parameter between the first matrix and the second matrix, and calculating second distance information between the first matrix and the second matrix;
[0134] determining the degree of association between the first abnormal data and the reference abnormal data according to the association parameter, the difference parameter and the second distance information;
[0135] The degree of association is obtained according to the following formula:
[0136]
[0137] r represents the degree of association, relation represents the association parameter, diff represents the difference parameter, represents the first matrix, represents the second matrix, dis represents the second distance information, and m represents the number corresponding to the first abnormal data.
[0138] In some implementations, the fault determination module 205, in the process of fusing the first abnormality type and the second abnormality type to obtain the fault monitoring result corresponding to the associated device, performs:
[0139] Determine a first recognition probability corresponding to the first abnormality type according to the association degree, and determine a second recognition probability corresponding to the second abnormality type;
[0140] Determine the target abnormality type corresponding to the associated device by fusing the first recognition probability and the second recognition probability;
[0141] The fault monitoring result corresponding to the associated device is determined according to the target abnormality type.
[0142] In some implementations, after obtaining the fault monitoring result corresponding to the associated device, the fault determination module 205 further performs:
[0143] Obtaining from a database a target user corresponding to the associated device and a target location corresponding to the associated device;
[0144] Obtaining a corresponding abnormal description statement using a text generation algorithm according to the target location and the fault monitoring result;
[0145] The exception description statement is sent to the target device corresponding to the target user, so that the target user receives the exception description statement of the associated device according to the target device, and then performs exception processing according to the exception description statement.
[0146] In some implementations, the intelligent plate processing workshop fault monitoring system 200 further performs:
[0147] Collecting video surveillance results corresponding to the video surveillance equipment at the key workstations;
[0148] The video monitoring result is presented in real time in a terminal device that is communicatively connected to the video monitoring device, so that the terminal device can display the operation result of the key workstation in real time according to the video monitoring result.
[0149] It should be noted that technical personnel in the relevant field can clearly understand that for the convenience and conciseness of description, the specific working process of the intelligent plate processing workshop fault monitoring system 200 described above can refer to the corresponding process in the aforementioned intelligent plate processing workshop fault monitoring method embodiment, and will not be repeated here.
[0150] An embodiment of the present invention also provides a storage medium for computer-readable storage, wherein the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any intelligent plate processing workshop fault monitoring method provided in the description of the embodiment of the present invention.
[0151] The storage medium may be an internal storage unit of the terminal device in the aforementioned embodiment, such as a hard disk or memory of the terminal device. The storage medium may also be an external storage device of the terminal device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the terminal device.
[0152] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware embodiment, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.
[0153] It should be understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations. It should be noted that, in this article, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "including a..." does not exclude the presence of other identical elements in the process, method, article or system including the element.
[0154] The serial numbers of the embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.
Claims
1. A fault monitoring method for an intelligent plate processing workshop, It is characterized in that The method comprises: Obtaining the operating status corresponding to the three-color alarm light of the key workstation, and performing abnormal screening on the key workstation according to the operating status to determine the target workstation; Obtaining, from a database, first data corresponding to the target workstation under a first sensor and second data corresponding to the target workstation under a second sensor, wherein the first data is discrete data of an associated device corresponding to the target workstation, and the second data is continuous data of the associated device corresponding to the target workstation; In the process of clustering the first sensor data, a quantity constraint between classes is introduced to obtain first abnormal data corresponding to the first sensor; Comparing the reference data with the second data to obtain second abnormal data corresponding to the second sensor; fusing the first abnormal data and the second abnormal data to obtain a fault monitoring result corresponding to the associated device; The step of introducing a quantity constraint between classes in a process of clustering the first sensor data to obtain first abnormal data corresponding to the first sensor includes: Determining an initial cluster center, and determining a degree of membership between the first sensor data and the initial cluster center; Calculating first distance information between the first sensor data and the initial cluster center; Determine a first loss value corresponding to the initial cluster center according to the first distance information and the membership degree; Setting the inter-class quantity constraint according to the membership degree to determine a second loss value corresponding to the initial cluster center; Performing loss fusion on the first loss value and the second loss value to obtain a target loss value corresponding to the initial cluster center; The initial cluster center is updated according to the target loss value to obtain the target cluster center corresponding to the first sensor data and the cluster data corresponding to the target cluster center; Performing abnormal classification on the cluster data to obtain the first abnormal data corresponding to the first sensor; The first loss value is obtained according to the following formula: ; represents the first loss value, N represents the number of the initial cluster centers, sum represents the number of the first sensor data, represents the degree of membership between the nth first sensor data and the mth initial clustering center, represents the degree of membership between the i-th first sensor data and the m-th initial clustering center, represents the first distance information between the nth first sensor data and the mth initial clustering center, and t represents a constant.
2. The method according to claim 1, It is characterized in that The performing loss fusion on the first loss value and the second loss value to obtain a target loss value corresponding to the initial cluster center includes: Performing a ratio operation on the first loss value and the second loss value to obtain a ratio result; Obtaining the number of clusterings corresponding to the initial clustering center, and determining the adjustment parameter corresponding to the initial clustering center according to the number of clusterings; Determine a fusion weight corresponding to the second loss value according to the ratio result and the adjustment parameter; The first loss value and the second loss value are fused according to the fusion weight to obtain the target loss value corresponding to the initial cluster center.
3. The method according to claim 1, It is characterized in that The step of comparing the reference data with the second data to obtain second abnormal data corresponding to the second sensor includes: Taking the reference data as a first coordinate point corresponding to a horizontal axis and taking the second data as a second coordinate point corresponding to a vertical axis; Calculating the difference between the first coordinate point and the second coordinate point, and collecting statistics on the distribution of the difference; Determine, according to the distribution, delay parameters corresponding to the reference data and the second data; Performing data alignment on the reference data and the second data according to the delay parameter to obtain difference data between the second data and the reference data; Determine, according to the difference data and the delay parameter, a change feature and a deviation feature corresponding to the second data, wherein the change feature is used to characterize a change of the second data, and the deviation feature is used to characterize a deviation phenomenon corresponding to the second data; The second data is screened according to the change feature and the deviation feature to obtain the second abnormal data corresponding to the second sensor.
4. The method according to claim 1, It is characterized in that The fusing the first abnormal data and the second abnormal data to obtain the fault monitoring result corresponding to the associated device includes: Calculating a correlation degree between the first abnormal data and reference abnormal data, and determining a first abnormal type corresponding to the first abnormal data according to the correlation degree; Decomposing and reconstructing the second abnormal data to obtain multidimensional features corresponding to the second abnormal data, and classifying the second abnormal data according to the multidimensional features to obtain a second abnormal type; The first abnormality type and the second abnormality type are merged to obtain the fault monitoring result corresponding to the associated device.
5. The method according to claim 4, It is characterized in that The calculating the correlation degree between the first abnormal data and the reference abnormal data includes: Obtain the reference abnormal data from the abnormal reference set, and then perform matrix representation on the reference abnormal data to obtain a corresponding first matrix; Performing matrix representation on the first abnormal data to obtain a corresponding second matrix; Calculating an association parameter and a difference parameter between the first matrix and the second matrix, and calculating second distance information between the first matrix and the second matrix; determining the degree of association between the first abnormal data and the reference abnormal data according to the association parameter, the difference parameter and the second distance information; The degree of association is obtained according to the following formula: ; r represents the degree of association, relation represents the association parameter, diff represents the difference parameter, represents the first matrix, represents the second matrix, dis represents the second distance information, and m represents the number corresponding to the first abnormal data.
6. The method according to claim 4, It is characterized in that The fusing the first abnormality type and the second abnormality type to obtain the fault monitoring result corresponding to the associated device includes: Determine a first recognition probability corresponding to the first abnormality type according to the association degree, and determine a second recognition probability corresponding to the second abnormality type; Determine the target abnormality type corresponding to the associated device by fusing the first recognition probability and the second recognition probability; The fault monitoring result corresponding to the associated device is determined according to the target abnormality type.
7. The method according to any one of claims 1 to 6, It is characterized in that After obtaining the fault monitoring result corresponding to the associated device, the method further includes: Obtaining from a database a target user corresponding to the associated device and a target location corresponding to the associated device; Obtaining a corresponding abnormal description statement using a text generation algorithm according to the target location and the fault monitoring result; The exception description statement is sent to the associated device corresponding to the target user, so that the target user receives the exception description statement of the associated device through the associated device, and then performs exception processing according to the exception description statement.
8. The method according to claim 7, It is characterized in that The method further comprises: Collecting video surveillance results corresponding to the video surveillance equipment at the key workstations; The video monitoring result is presented in real time in a terminal device that is communicatively connected to the video monitoring device, so that the terminal device can display the operation result of the key workstation in real time according to the video monitoring result.
9. An intelligent plate processing workshop fault monitoring system, It is characterized in that include: A workstation determination module is used to obtain the operating status corresponding to the three-color alarm light of the key workstation, and to perform abnormal screening on the key workstation according to the operating status to determine the target workstation; A data acquisition module, used to obtain from a database first data corresponding to the target station under the first sensor and second data corresponding to the second sensor, wherein the first data is discrete data of an associated device corresponding to the target station, and the second data is continuous data of the associated device corresponding to the target station; The first identification module is used to introduce the inter-class quantity constraint in the process of clustering the data of the first sensor to obtain the first abnormal data corresponding to the first sensor, wherein the introduction of the inter-class quantity constraint in the process of clustering the data of the first sensor to obtain the first abnormal data corresponding to the first sensor includes: determining an initial cluster center and determining the membership between the first sensor data and the initial cluster center; calculating the first distance information between the first sensor data and the initial cluster center; determining the first loss value corresponding to the initial cluster center according to the first distance information and the membership; setting the inter-class quantity constraint according to the membership to determine the second loss value corresponding to the initial cluster center; performing loss fusion on the first loss value and the second loss value to obtain a target loss value corresponding to the initial cluster center; updating the initial cluster center according to the target loss value to obtain a target cluster center corresponding to the first sensor data and cluster data corresponding to the target cluster center; performing abnormal classification on the cluster data to obtain the first abnormal data corresponding to the first sensor; wherein the first loss value is obtained according to the following formula: ; represents the first loss value, N represents the number of the initial cluster centers, sum represents the number of the first sensor data, represents the degree of membership between the nth first sensor data and the mth initial clustering center, represents the degree of membership between the i-th first sensor data and the m-th initial clustering center, represents the first distance information between the nth first sensor data and the mth initial cluster center, and t represents a constant; A second identification module, used for comparing the reference data with the second data to obtain second abnormal data corresponding to the second sensor; The fault determination module is used to fuse the first abnormal data and the second abnormal data to obtain a fault monitoring result corresponding to the associated device.
