Fault analysis method and system based on equipment model

Abstract

The invention relates to the technical field of fault analysis, and discloses a fault analysis method and system based on an equipment model, and the method comprises the steps: loading an equipment model of a switch cabinet; acquiring real-time operation data of the switch cabinet, and disassembling the real-time operation data into sub-data in different working forms; for any piece of sub-data, issuing the sub-data to the corresponding form sub-model; according to a time sequence relationship, collecting simulation data obtained after the sub-data is operated by each form sub-model, and according to the collected simulation data, constructing a plurality of groups of simulation time sequence arrays; generating a fault analysis result of each group of simulation time sequence data one by one; and obtaining a fault analysis result of the last simulation time ordinal array, and determining the fault analysis result as a fault analysis result of the switch cabinet. The system has the beneficial effects of more precise monitoring and fault analysis.

Description

Technical Field

[0001] The present application relates to the technical field of fault analysis, and in particular to a fault analysis method and system based on a device model. Background Art

[0002] Current switchgear fault analysis methods have many limitations. First, manual judgment is easily affected by human factors such as operator experience and emotions, resulting in inaccurate or in-time diagnosis of complex or rare faults. Second, traditional simple simulation technology usually only considers typical working modes, ignoring the changes of equipment under complex working conditions, and it is difficult to fully reflect the combined effects of various operating states and fault modes of the equipment, thus affecting the analysis accuracy.

[0003] Therefore, the current problem that needs to be solved is how to improve the accuracy and precision of fault analysis and achieve more sophisticated monitoring and fault analysis. Summary of the invention

[0004] The present application provides a fault analysis method and system based on an equipment model, which solves the technical problem of how to improve the accuracy and precision of fault analysis and achieves the technical effect of more sophisticated monitoring and fault analysis.

[0005] In order to achieve the above objectives, the main technical solutions adopted in this application include: In a first aspect, an embodiment of the present application provides a fault analysis method based on a device model, the method comprising: Loading a device model of a switch cabinet, wherein the device model includes a plurality of state sub-models of the switch cabinet in different working states, and each of the state sub-models is arranged according to a time sequence relationship of the working state; Acquire the real-time operation data of the switch cabinet, and decompose the real-time operation data into sub-data under different working modes; For any sub-data, the sub-data is sent to the corresponding morphological sub-model to run the sub-data through the morphological sub-model; According to the time series relationship, the simulation data obtained after the sub-data of each morphological sub-model is run are collected, and multiple groups of simulation time series arrays are constructed according to the collected simulation data; Generate fault analysis results for each group of simulation timing data one by one, wherein, when generating the fault analysis results of the current simulation timing array, obtain the historical fault analysis results of the previous simulation timing array before the current simulation timing data, and use the historical fault analysis results as guide labels to perform fault prediction on the current simulation timing array; A fault analysis result of the last simulation time series array is obtained, and the fault analysis result is determined as the fault analysis result of the switch cabinet.

[0006] The present embodiment provides a fault analysis method based on a device model, which loads multiple morphological sub-models, and these sub-models are arranged according to different working modes of the device and their timing relationships, providing a basis for subsequent accurate fault analysis. Real-time operation data is obtained and disassembled into sub-data related to different working modes to ensure the pertinence and efficiency of the data. The sub-data is sent to the corresponding morphological sub-model for processing, ensuring that each sub-data can be simulated under appropriate working conditions, and can effectively simulate the performance of the switch cabinet under different working modes. Simulation data is collected according to the timing relationship and multiple groups of simulation timing arrays are constructed, providing sufficient data support for subsequent fault analysis. For each group of simulation timing data, by combining the historical fault analysis results, the fault mode in the previous data can be used to predict the fault of the current simulation data. This historical data-oriented analysis method not only improves the accuracy of the prediction, but also enhances the robustness of the model. Finally, by analyzing the fault analysis results of the last simulation timing array, a comprehensive evaluation and judgment of the switch cabinet fault is provided.

[0007] Optionally, sending the sub-data to the corresponding morphological sub-model includes: Identify the target working form to which the sub-data belongs, and obtain the target timing identifier corresponding to the target working form; The sub-data is transmitted to the morphological sub-model having the target timing identification.

[0008] This embodiment identifies the target working form of the sub-data and obtains the corresponding target timing identifier to ensure that each sub-data is assigned to the appropriate processing path as needed. The sub-data is transferred to the morphological sub-model with the corresponding timing identifier for further processing, which realizes the automation and precise timing management of data processing, not only improving the processing efficiency of the system, but also enhancing flexibility, and being able to optimize the processing process under different working forms and time series requirements.

[0009] Optionally, constructing multiple groups of simulation timing arrays according to the collected simulation data includes: Identify the timing identifiers covered by the multiple morphological sub-models, and for any current timing identifier, obtain the simulation data corresponding to the current timing identifier, and obtain the simulation data corresponding to each of the other timing identifiers before the current timing identifier, and splice the obtained simulation data according to the timing identifiers; The spliced ​​data is filled into a preset first field, and the number of spliced ​​simulation data is filled into a preset second field, and the array formed by the first field and the second field is used as the simulation timing array corresponding to the current timing identifier.

[0010] The timing identification and simulation data acquisition process of this embodiment makes it possible to comprehensively collect and understand the operating status of the switch cabinet at different time nodes, ensuring a complete grasp of the historical evolution process. By splicing the simulation data of different time nodes, a continuous time series is formed, revealing the changing trend of the operating status over time. Filling the spliced ​​data into the preset first field and the number of copies of the simulation data into the second field can ensure the structuring and standardization of the data during storage, making subsequent processing more efficient. The obtained simulation timing array is used as a data carrier to facilitate subsequent tasks such as timing analysis and fault diagnosis.

[0011] Optionally, generating the fault analysis results of each group of simulation timing data one by one includes: For a first simulation timing array and a second simulation timing array that are adjacent to each other, obtaining a first fault analysis result of the first simulation timing array, and setting the first fault analysis result as a guide label of the second simulation timing array; The second simulation timing array with the guide label set is input into the trained fusion network, wherein the fusion network includes a label configuration network and a result prediction network, wherein the label configuration network extracts features of the guide label and determines a weight matrix that matches the features of the guide label, and the weight matrix is ​​loaded into the result prediction network, so that the result prediction network loaded with the weight matrix performs fault prediction on the second simulation timing array.

[0012] This embodiment processes the adjacent first simulation timing array and the second simulation timing array, first obtains the fault analysis result of the first simulation timing array, and uses it as the guide label of the second simulation timing array. It ensures that the analysis of the second simulation timing array can be based on the fault mode in the first simulation timing array, thereby providing effective reference information for subsequent fault prediction. The guide label is input into the trained fusion network. The fusion network includes a label configuration network and a result prediction network. The main function of the label configuration network is to extract features from the guide label and determine the weight matrix matching these features through a learning process. After training, the weight matrix can reflect the degree of influence of different features in the guide label on fault prediction. Subsequently, this weight matrix is ​​loaded into the result prediction network to enhance the performance of the network in the fault prediction task. The embodiment of the present application can effectively associate and share information between different simulation timing arrays, thereby improving the accuracy of the prediction results.

[0013] Optionally, the matching relationship between the guide label and the weight matrix is ​​determined by: Acquire a time series training sample, wherein the time series training sample includes a guide label sample, a time series array sample, and a prediction label; Extracting features from the guide label samples using a label configuration network, and fusing the extracted features with the time series array samples to obtain fused samples; Using a result prediction network to predict the fusion sample, to obtain a fault prediction result; Calculating the error between the fault prediction result and the prediction label, and adjusting the weight matrix of the result prediction network based on the error, so that when the result prediction network after the weight matrix adjustment predicts the fusion sample again, the obtained fault prediction result matches the prediction label; The guide label samples are matched with the adjusted weight matrix.

[0014] This embodiment provides basic data for subsequent model training by obtaining time series training samples including guide label samples, time series array samples and prediction labels. Then the label configuration network extracts features from the guide label samples. The extracted features can effectively reveal the key features of the fault mode. These features are fused with the time series array samples to form fused samples. The fused samples are input into the result prediction network to perform fault prediction and obtain preliminary fault prediction results. By calculating the error between the prediction result and the prediction label, the weight matrix of the result prediction network is adjusted according to the error. This process enables the model to learn the fault mode more accurately in the prediction task, thereby optimizing the prediction result. Finally, the guide label samples are matched with the adjusted weight matrix to ensure that the information sharing and interaction in the entire prediction process reach the optimal state. Not only the accuracy and reliability of the fault prediction model are improved, but also the adaptability of the model is enhanced, and it can cope with various fault modes and time series data, and improve the intelligence level and generalization ability of fault prediction.

[0015] Optionally, fusing the extracted features with the time series array samples to obtain fused samples includes: Identify each row of data in the extracted features, and concatenate each row of data according to the row number to obtain a corresponding concatenation vector; The concatenated vector is used as a suffix of the time series array sample to form a fused sample.

[0016] This embodiment identifies and extracts each row of data in the feature, and then splices these data according to the row number to obtain the corresponding splicing vector. It can effectively combine the row data in the feature in order, retain the time series information and its internal structure in the time series data, so that the order of each row of data is fully reflected, thereby improving the accuracy and effectiveness of subsequent analysis. The splicing vector is then used as a suffix of the time series array sample and combined with the original time series data to form a fused sample. Through this splicing method, the time series array sample and the extracted feature information are organically integrated, so that each sample can not only contain time series information, but also add more valuable information obtained from feature extraction. This fusion enhances the model's ability to understand time series data and improves the accuracy and robustness of fault prediction.

[0017] Optionally, the real-time operation data of the switch cabinet includes at least one of online temperature measurement data of new contacts, online detection data of opening and closing voltage and current, online monitoring data of mechanical properties, and sensing data of partial discharge sensors.

[0018] In a second aspect, an embodiment of the present application provides a fault analysis system based on a device model, the system comprising: A model loading unit, used to load a device model of the switch cabinet, wherein the device model includes a plurality of form sub-models of the switch cabinet in different working modes, and each of the form sub-models is arranged according to a time sequence relationship of the working modes; A data decomposition unit, used for acquiring the real-time operation data of the switch cabinet, and decomposing the real-time operation data into sub-data under different working modes; An operating unit, for sending any sub-data to a corresponding morphological sub-model, so as to operate the sub-data through the morphological sub-model; An array construction unit, used to collect the simulation data obtained after running the sub-data of each morphological sub-model according to the time series relationship, and construct multiple groups of simulation time series arrays according to the collected simulation data; A fault analysis unit is used to generate fault analysis results for each group of simulation timing data one by one, wherein when generating the fault analysis results of the current simulation timing array, a historical fault analysis result of the previous simulation timing array before the current simulation timing data is obtained, and the historical fault analysis result is used as a guide label to perform fault prediction on the current simulation timing array; The result determination unit is used to obtain the fault analysis result of the last simulation time series array, and determine the fault analysis result as the fault analysis result of the switch cabinet.

[0019] Optionally, the fault analysis unit is specifically used to obtain a first fault analysis result of the first simulation timing array and a second adjacent simulation timing array, and set the first fault analysis result as a guide label for the second simulation timing array; input the second simulation timing array with the guide label set into the trained fusion network, the fusion network including a label configuration network and a result prediction network, wherein the label configuration network extracts features of the guide label and determines a weight matrix that matches the features of the guide label, and the weight matrix is ​​loaded into the result prediction network, so that the result prediction network loaded with the weight matrix performs fault prediction on the second simulation timing array.

[0020] Optionally, the fault analysis unit determines the matching relationship between the guide label and the weight matrix in the following manner: obtain a time series training sample, the time series training sample including a guide label sample, a time series array sample and a prediction label; use a label configuration network to extract features from the guide label sample, and fuse the extracted features with the time series array sample to obtain a fused sample; use a result prediction network to predict the fused sample to obtain a fault prediction result; calculate the error between the fault prediction result and the prediction label, and adjust the weight matrix of the result prediction network based on the error, so that when the result prediction network after the weight matrix adjustment predicts the fused sample again, the obtained fault prediction result matches the prediction label; match the guide label sample with the adjusted weight matrix. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the specific implementation methods of the present application or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 A flowchart of a fault analysis method based on a device model provided in an embodiment of the present application; Figure 2 A flowchart of step S5 provided in an embodiment of the present application; Figure 3 A flowchart of step S7 provided in an embodiment of the present application; Figure 4 A flowchart of step S9 provided in an embodiment of the present application; Figure 5 A flowchart of a matching relationship determination method provided in an embodiment of the present application; Figure 6 A flow chart for obtaining a fusion sample provided in an embodiment of the present application; Figure 7 A framework of a fault analysis system based on a device model provided in an embodiment of the present application; Figure 8 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0024] According to an embodiment of the present application, an embodiment of a fault analysis method based on a device model is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0025] Current switchgear fault analysis methods have many limitations. First, manual fault analysis usually relies on the operator's experience and intuition, which is easily affected by human factors such as fatigue, emotions or knowledge blind spots. This results in the analysis results not being accurate or timely when facing complex or rare faults. For example, operators may not fully cover all possible fault modes or ignore some potential abnormal signals, which affects the accuracy and reliability of fault diagnosis.

[0026] In addition, some systems use fault analysis methods based on traditional simulation models. These simulation models are usually simple and often only consider some typical working modes of the switchgear, ignoring the details and changes of the equipment under complex working conditions. This simple simulation method is difficult to fully cover the various working modes of the equipment and cannot accurately reflect the complex behavior of the switchgear in a changing environment, which in turn affects the accuracy of fault analysis. For example, a simple simulation model may not be able to simulate the combined effects of multiple fault modes, nor can it dynamically adapt to the real-time changing state of the equipment, resulting in a large deviation between the predicted results and the actual fault.

[0027] In order to solve the above problems, this embodiment provides a fault analysis method based on a device model. Figure 1A flow chart of a fault analysis method based on a device model provided in an embodiment of the present application, such as Figure 1 As shown, the process includes the following steps: Step S1, loading a device model of a switch cabinet, wherein the device model includes a plurality of form sub-models of the switch cabinet in different working modes, and each form sub-model is arranged according to a time sequence relationship of the working modes.

[0028] Specifically, the working state refers to the set of different working states of the switchgear during actual operation and operation. These states include the operation process of the switchgear equipment, changes during operation, and different working states when the circuit breaker is opened and closed. The morphological sub-model refers to a model established for each specific working state, which is constructed through digital twin technology and is used to simulate and analyze the behavior of the switchgear under a specific working state. Each morphological sub-model is arranged according to the temporal relationship of the working state to ensure that the behavior of the entire switchgear can be simulated continuously and accurately.

[0029] Step S3, acquiring the real-time operation data of the switch cabinet, and breaking down the real-time operation data into sub-data under different working modes.

[0030] Specifically, the operation data of the switch cabinet is collected in real time through sensors and monitoring systems. These data may include at least one of online temperature measurement data of new contacts, online detection data of voltage and current of opening and closing, online monitoring data of mechanical characteristics, and sensing data of partial discharge sensors. The data is disassembled according to the timestamp and state change of the real-time operation data, including dividing the data into time periods according to the time when the event occurs (such as closing, opening, fault, etc.). The data in each time period is classified as a specific working form. By monitoring the state changes of the switch cabinet (such as closing, opening, fault alarm, etc.), these state change points are used as signs of data segmentation. For example, after monitoring the closing operation, relevant data can be extracted from this moment until the next operation or state change point. Each data segment is labeled accordingly, such as "closing operation", "overtemperature state", "fault protection state", etc. In this way, specific working form data can be quickly located in subsequent analysis. The disassembled data is stored in the database, usually organized and indexed by time and working form. By labeling the data, it is convenient to query and further analyze.

[0031] Step S5: for any sub-data, the sub-data is sent to the corresponding morphological sub-model to run the sub-data through the morphological sub-model.

[0032] Specifically, each morphological sub-model is designed for a specific working form and can simulate the behavior of the switch cabinet under that form. The sub-data is sent to the corresponding morphological sub-model. After receiving the sub-data, the morphological sub-model performs physical simulation operation.

[0033] Step S7, collecting the simulation data obtained after running the sub-data of each morphological sub-model according to the time series relationship, and constructing multiple groups of simulation time series arrays based on the collected simulation data.

[0034] Specifically, according to the time series relationship, the simulation data of each morphological sub-model after the operation sub-data is collected to ensure that the simulation data is ordered in the time series. By collecting these data in chronological order, the evolution process of the switchgear in the entire operation cycle can be reflected. After collecting these time series data, multiple simulation time series arrays will be constructed based on the collected data. The simulation time series array refers to a set of data structures that organize the simulation data in chronological order, and each array corresponds to a set of simulation data under a specific time series.

[0035] Step S9, generating fault analysis results for each group of simulation timing data one by one, wherein, when generating the fault analysis results of the current simulation timing array, obtaining the historical fault analysis results of the previous simulation timing array before the current simulation timing data, and using the historical fault analysis results as guide labels to perform fault prediction on the current simulation timing array.

[0036] Specifically, each set of simulation timing data is analyzed to generate fault analysis results. When generating the fault analysis results of the current simulation timing data, it is necessary to obtain the historical fault analysis results of the previous simulation timing array before the current simulation timing data. These historical results serve as guide tags to provide contextual information for the current fault prediction, and can effectively use historical fault data to optimize the fault prediction of real-time simulation data and improve the accuracy and reliability of the prediction. This method combines historical fault data and real-time simulation data in order to obtain more accurate prediction results.

[0037] Step S11, obtaining the fault analysis result of the last simulation time series array, and determining the fault analysis result as the fault analysis result of the switch cabinet.

[0038] Specifically, after completing the fault analysis of all simulation time series arrays, the fault analysis result obtained by the last simulation time series array is collected. The fault analysis result is the fault analysis result of the switch cabinet, which takes into account the influence of the latest operation data and historical fault analysis results.

[0039] The present embodiment provides a fault analysis method based on a device model, which loads multiple morphological sub-models, and these sub-models are arranged according to different working modes of the device and their timing relationships, providing a basis for subsequent accurate fault analysis. Real-time operation data is obtained and disassembled into sub-data related to different working modes to ensure the pertinence and efficiency of the data. The sub-data is sent to the corresponding morphological sub-model for processing, ensuring that each sub-data can be simulated under appropriate working conditions, and can effectively simulate the performance of the switch cabinet under different working modes. Simulation data is collected according to the timing relationship and multiple groups of simulation timing arrays are constructed, providing sufficient data support for subsequent fault analysis. For each group of simulation timing data, by combining the historical fault analysis results, the fault mode in the previous data can be used to predict the fault of the current simulation data. This historical data-oriented analysis method not only improves the accuracy of the prediction, but also enhances the robustness of the model. Finally, by analyzing the fault analysis results of the last simulation timing array, a comprehensive evaluation and judgment of the switch cabinet fault is provided.

[0040] Figure 2 The flowchart of step S5 provided in the embodiment of the present application may include the following steps: Step S51, identifying the target working form to which the sub-data belongs, and obtaining the target timing identifier corresponding to the target working form.

[0041] Specifically, in order to identify the working state of the switch cabinet corresponding to each sub-data in the real-time operation data, machine learning algorithms such as decision trees, random forests, support vector machines, etc. can be used for classification. First, key features that can represent different working states are extracted from the sub-data. For example, characteristic parameters of the opening and closing current waveform, vibration frequency and amplitude of the circuit breaker, etc. can be extracted. These features can reflect the different working states of the switch cabinet. Then the extracted features are formed into feature vectors as the input of the model. Using a supervised learning-based method, the model is trained through historical data to obtain a pre-trained classification model. The model can judge the working state corresponding to the sub-data based on the input feature vector. Finally, through the trained classification model, each sub-data is classified into different working states according to the feature vector, thereby identifying the target working state to which the sub-data belongs.

[0042] After identifying the target working form, the corresponding target timing identifier needs to be obtained. Here, a unique timing identifier corresponding to each working form of the switch cabinet needs to be pre-assigned, that is, to identify each working state. Then, the target working form and the corresponding target timing identifier are found according to the predefined mapping relationship working form-timing identifier.

[0043] Step S53, transmitting the sub-data to the morphological sub-model with the target timing identification.

[0044] Specifically, once the target working form and corresponding target timing identifier of the sub-data are determined, the sub-data is transmitted to the form sub-model with the corresponding timing identifier to ensure that the data can be correctly sent to the form sub-model for further analysis and processing.

[0045] and Figure 1 Compared with the embodiment shown in the figure, this embodiment ensures that each sub-data is assigned to a suitable processing path as required by identifying the target working form of the sub-data and obtaining the corresponding target timing identifier. The sub-data is transmitted to the morphological sub-model with the corresponding timing identifier for further processing, which realizes the automation and precise timing management of data processing, not only improves the processing efficiency of the system, but also enhances the flexibility, and can optimize the processing process under different working forms and time series requirements.

[0046] Figure 3 The flowchart of step S7 provided in the embodiment of the present application may include the following steps: Step S71, identify the timing identifiers covered by multiple morphological sub-models, and for any current timing identifier, obtain the simulation data corresponding to the current timing identifier, as well as obtain the simulation data corresponding to each of the other timing identifiers located before the current timing identifier, and splice the obtained simulation data according to the timing identifiers.

[0047] Specifically, firstly, the time series identifiers covered by multiple morphological sub-models are identified, and these time series identifiers represent different working forms. Each time series identifier corresponds to specific simulation data, which records the operation status of the switch cabinet at the time series node. For any current time series identifier, the simulation data related to it is obtained. These simulation data usually contain key status information at the time series point, such as the operating parameters, fault information or other monitoring indicators of the equipment, which are used to describe the current working status. At the same time, it is also necessary to obtain the simulation data corresponding to other time series identifiers before the current time series identifier. By obtaining the simulation data corresponding to these historical time series identifiers, it is possible to fully understand how the current state has evolved and the change trend in the historical data. Finally, all the obtained simulation data are spliced ​​in the order of the time series identifiers. This process connects the simulation data of different time nodes in series to form a continuous time series, providing a comprehensive view of the operation process of the switch cabinet. Through the spliced ​​simulation data, the entire working form can be analyzed and predicted more deeply, potential problems can be identified, operation strategies can be optimized, and more accurate fault warnings and maintenance solutions can be provided.

[0048] Step S73, fill the spliced ​​data into the preset first field, fill the number of spliced ​​simulation data into the preset second field, and use the array formed by the first field and the second field as the simulation timing array corresponding to the current timing identifier.

[0049] Specifically, the spliced ​​simulation data is filled into the preset first field, which represents all the simulation data under the current timing identifier. The number of copies of the spliced ​​simulation data is filled into the preset second field, which records the number of simulation data under the current timing identifier, which helps to understand the integrity and coverage of the data. The first field and the second field are combined to form a simulation timing array corresponding to the current timing identifier. This array not only contains the simulation data, but also contains the statistical information of the data, providing comprehensive data support for fault analysis.

[0050] and Figure 1 Compared with the embodiment shown, the timing identification and simulation data acquisition process of this embodiment makes it possible to comprehensively collect and understand the operating status of the switch cabinet at different time nodes, ensuring a complete grasp of the historical evolution process. By splicing the simulation data of different time nodes, a continuous time series is formed, revealing the changing trend of the operating status over time. Filling the spliced ​​data into the preset first field and filling the number of copies of the simulation data into the second field can ensure the structuring and standardization of the data during storage, making subsequent processing more efficient. The obtained simulation timing array is used as a data carrier to facilitate subsequent tasks such as timing analysis and fault diagnosis.

[0051] Figure 4 The flowchart of step S9 provided in the embodiment of the present application may include the following steps: Step S91 , for adjacent first simulation timing array and second simulation timing array, obtain a first fault analysis result of the first simulation timing array, and set the first fault analysis result as a guide label of the second simulation timing array.

[0052] Specifically, the fault analysis results obtained from the first simulation timing array are actually based on historical data and previous simulation results, reflecting the existing fault modes and laws. These fault analysis results are used as guide labels and applied to the second simulation timing array. This guide label acts as a representative of the historical fault data and provides guidance for subsequent simulation data. In the subsequent fault prediction process, historical fault data can not only be combined with real-time data to help identify possible faults, but also improve the accuracy of predictions through known fault modes. Therefore, this method can effectively improve the prediction accuracy and ensure that fault prediction not only relies on current data, but also integrates historical experience.

[0053] Step S93, input the second simulation timing array with the guide label set into the trained fusion network, the fusion network includes a label configuration network and a result prediction network, wherein the label configuration network extracts features of the guide label and determines a weight matrix that matches the features of the guide label, and the weight matrix is ​​loaded into the result prediction network, so that the result prediction network loaded with the weight matrix performs fault prediction on the second simulation timing array.

[0054] Specifically, first, the label configuration network is responsible for feature extraction of the guide label. This process can be implemented through a deep learning model, such as using a convolutional neural network (CNN) or a long short-term memory network (LSTM) to identify key features in the guide label. These features can represent important information of the guide label and provide support for subsequent weight matrix determination and fault prediction. Once the features of the guide label are extracted, the label configuration network determines a weight matrix that matches these features. This weight matrix is ​​obtained through learning during the training process and can reflect the degree of influence of the guide label features on fault prediction. The determination of the weight matrix can be achieved through optimization algorithms, such as stochastic gradient descent (SGD) or Adam optimizer, which can adjust the weights during the training process to minimize the prediction error. Next, the determined weight matrix is ​​loaded into the result prediction network, which can be an LSTM network or other models suitable for time series prediction. The result prediction network loaded with the weight matrix can use the information of the guide label to improve the accuracy of fault prediction. The loading of the weight matrix is ​​equivalent to providing the network with prior knowledge, so that the network can pay more attention to those features that match the historical fault mode when processing the second simulation time series array. The result prediction network loaded with the adaptation weight matrix performs fault prediction on the second simulation time series array. This prediction process combines historical fault data and real-time simulation data to obtain more accurate prediction results.

[0055] It should be noted that by constructing the association between the guide label and the weight matrix, the adapted weight matrix can be determined according to the current guide label. This process is similar to training a neural network, in which the features of the guide label are used to adjust the weights of the result prediction network to adapt to the current prediction task. The role of the weight matrix in the result prediction network is to adjust the response of the network to pay more attention to those features that match the historical failure mode. In this way, the result prediction network can better capture the key factors that may cause failures. In this way, historical failure data can be used to optimize the fault prediction of real-time simulation data and improve the accuracy and reliability of the prediction.

[0056] and Figure 1Compared with the embodiment shown, this embodiment processes the adjacent first simulation timing array and the second simulation timing array, first obtains the fault analysis result of the first simulation timing array, and uses it as the guide label of the second simulation timing array. It ensures that the analysis of the second simulation timing array can be based on the fault mode in the first simulation timing array, thereby providing effective reference information for subsequent fault prediction. The guide label is input into the trained fusion network. The fusion network includes a label configuration network and a result prediction network. The main function of the label configuration network is to extract features from the guide label and determine the weight matrix matching these features through a learning process. After training, the weight matrix can reflect the degree of influence of different features in the guide label on fault prediction. Subsequently, this weight matrix is ​​loaded into the result prediction network to enhance the performance of the network in the fault prediction task. The embodiment of the present application can effectively associate and share information between different simulation timing arrays, thereby improving the accuracy of the prediction results.

[0057] Figure 5 A flowchart of a matching relationship determination method provided in an embodiment of the present application, the process may include the following steps: Step S931, obtaining time series training samples, which include guide label samples, time series array samples and prediction labels.

[0058] Specifically, the guide label samples are extracted from historical data and are past fault records of switch cabinets, which are used to indicate past fault events. The time series array samples are real-time operating data corresponding to the guide labels, such as time series data such as voltage, current, and temperature of the switch cabinet. The predicted labels are the expected outputs corresponding to the time series array samples, which are usually known fault labels and are used to train the model to identify faults.

[0059] Step S933, using the label configuration network to extract features from the guide label samples, and fusing the extracted features with the time series array samples to obtain fused samples.

[0060] Use a label configuration network, such as a convolutional neural network (CNN) or a long short-term memory network (LSTM), to identify and extract features from the guide label samples, and combine the extracted features with the time series array samples to form a new dataset, namely the fused sample, which contains the original time series information and historical fault features.

[0061] Step S935: Use the result prediction network to predict the fusion sample to obtain a fault prediction result.

[0062] Specifically, the fused samples are input into a result prediction network, such as a convolutional neural network (CNN) or a long short-term memory network (LSTM), and the network outputs a fault prediction result based on the fused samples.

[0063] Step S937, calculate the error between the fault prediction result and the prediction label, and adjust the weight matrix of the result prediction network based on the error, so that when the result prediction network after the weight matrix adjustment predicts the fusion sample again, the obtained fault prediction result matches the prediction label.

[0064] Specifically, the error between the fault prediction result and the predicted label is calculated using the mean square error (MSE) or cross entropy error. Then, based on the calculated error, an optimization algorithm (such as SGD or Adam) is used to adjust the weight matrix of the result prediction network until the error does not exceed a preset threshold.

[0065] Step S939, matching the guide label sample with the adjusted weight matrix.

[0066] Specifically, the adjusted weight matrix is ​​matched with the guide label sample to obtain the corresponding relationship between the weight matrix and the guide label sample for subsequent optimization of the network's prediction ability for new samples.

[0067] and Figure 4 Compared with the embodiment shown in the figure, this embodiment provides basic data for subsequent model training by obtaining time series training samples including guide label samples, time series array samples and prediction labels. Then the label configuration network extracts features from the guide label samples. The extracted features can effectively reveal the key features of the fault mode. These features are fused with the time series array samples to form fused samples. The fused samples are input into the result prediction network to perform fault prediction and obtain preliminary fault prediction results. By calculating the error between the prediction result and the prediction label, the weight matrix of the result prediction network is adjusted according to the error. This process enables the model to learn the fault mode more accurately in the prediction task, thereby optimizing the prediction result. Finally, the guide label samples are matched with the adjusted weight matrix to ensure that the information sharing and interaction in the entire prediction process reach the optimal state. Not only the accuracy and reliability of the fault prediction model are improved, but also the adaptability of the model is enhanced, and it can cope with various fault modes and time series data, and improve the intelligence level and generalization ability of fault prediction.

[0068] Figure 6 The flowchart of obtaining the fusion sample provided in the embodiment of the present application may include the following steps: Step S9331, identifying each row of data in the extracted features, and splicing each row of data according to the row number to obtain a corresponding splicing vector.

[0069] Specifically, each row of data is identified from the features extracted by the label configuration network. The identified feature row data is spliced ​​in the order of their original row numbers. This splicing process integrates the multi-dimensional feature information into a one-dimensional vector so that it can be fused with the time series array samples. After the splicing is completed, a splicing vector is obtained, which contains the continuous information of all row data, providing a basis for subsequent fusion.

[0070] Step S9333, using the concatenated vector as a suffix of the time series array sample to form a fused sample.

[0071] Specifically, the concatenated vector is used as a suffix of the time series array sample to form a fused sample. This means that the concatenated vector is appended to the end of the time series array sample to form a longer vector that contains both the time series data and the feature information extracted from the guide label. The fused sample integrates the feature information of the original time series data and the guide label, providing more comprehensive input data for the result prediction network.

[0072] and Figure 5 Compared with the embodiment shown, this embodiment identifies and extracts each row of data in the feature, and then splices these data according to the row number to obtain the corresponding splicing vector. It can effectively combine the row data in the feature in order, retain the time series information and its internal structure in the time series data, so that the order of each row of data is fully reflected, thereby improving the accuracy and effectiveness of subsequent analysis. The splicing vector is then used as a suffix of the time series array sample and combined with the original time series data to form a fused sample. Through this splicing method, the time series array sample and the extracted feature information are organically integrated, so that each sample can add more valuable information obtained from feature extraction on the basis of containing the time series information. This fusion enhances the model's ability to understand time series data and improves the accuracy and robustness of fault prediction.

[0073] Accordingly, please refer to Figure 7 A framework of a fault analysis system based on a device model provided in an embodiment of the present application, the system comprising: The model loading unit 101 is used to load the equipment model of the switch cabinet, wherein the equipment model includes a plurality of form sub-models of the switch cabinet in different working modes, and each form sub-model is arranged according to the time sequence relationship of the working mode; The data decomposition unit 103 is used to obtain the real-time operation data of the switch cabinet and decompose the real-time operation data into sub-data under different working modes; The running unit 105 is used for sending any sub-data to the corresponding morphological sub-model to run the sub-data through the morphological sub-model; An array construction unit 107 is used to collect the simulation data obtained after running the sub-data of each morphological sub-model according to the time series relationship, and construct multiple groups of simulation time series arrays according to the collected simulation data; The fault analysis unit 109 is used to generate fault analysis results of each group of simulation timing data one by one, wherein when generating the fault analysis results of the current simulation timing array, the historical fault analysis results of the previous simulation timing array before the current simulation timing data are obtained, and the historical fault analysis results are used as guide labels to perform fault prediction on the current simulation timing array; The result determination unit 111 is used to obtain the fault analysis result of the last simulation time series array, and determine the fault analysis result as the fault analysis result of the switch cabinet.

[0074] In some optional embodiments, the fault analysis unit 109 is specifically used to obtain a first fault analysis result of the first simulation timing array and a second adjacent simulation timing array, and set the first fault analysis result as a guide label of the second simulation timing array; the second simulation timing array with the guide label set is input into the trained fusion network, and the fusion network includes a label configuration network and a result prediction network, wherein the label configuration network extracts features of the guide label and determines a weight matrix that matches the features of the guide label, and the weight matrix is ​​loaded into the result prediction network, so that the result prediction network loaded with the weight matrix performs fault prediction on the second simulation timing array.

[0075] In some optional embodiments, the fault analysis unit 109 determines the matching relationship between the guide label and the weight matrix in the following manner: obtain a time series training sample, which includes a guide label sample, a time series array sample, and a prediction label; use a label configuration network to extract features from the guide label sample, and fuse the extracted features with the time series array sample to obtain a fused sample; use a result prediction network to predict the fused sample to obtain a fault prediction result; calculate the error between the fault prediction result and the prediction label, and adjust the weight matrix of the result prediction network based on the error, so that when the result prediction network after the weight matrix adjustment is completed predicts the fused sample again, the obtained fault prediction result matches the prediction label; match the guide label sample with the adjusted weight matrix.

[0076] In some optional implementations, sending the sub-data to the corresponding morphological sub-model includes: Identify the target work form to which the sub-data belongs, and obtain the target timing identifier corresponding to the target work form; The sub-data is transferred to the morphological sub-model with the target timing identification.

[0077] In some optional implementations, constructing multiple sets of simulation timing arrays according to the collected simulation data includes: Identify the timing identifiers covered by multiple morphological sub-models, obtain the simulation data corresponding to any current timing identifier, and obtain the simulation data corresponding to other timing identifiers before the current timing identifier, and splice the obtained simulation data according to the timing identifiers; The spliced ​​data is filled into a preset first field, and the number of spliced ​​simulation data is filled into a preset second field, and the array formed by the first field and the second field is used as the simulation timing array corresponding to the current timing identifier.

[0078] In some optional implementations, generating the fault analysis results of each group of simulation timing data one by one includes: For the adjacent first simulation timing array and the second simulation timing array, obtaining a first fault analysis result of the first simulation timing array, and setting the first fault analysis result as a guide label of the second simulation timing array; The second simulation timing array with the guide label set is input into the trained fusion network, which includes a label configuration network and a result prediction network, wherein the label configuration network extracts features of the guide label and determines a weight matrix that matches the features of the guide label, and the weight matrix is ​​loaded into the result prediction network, so that the result prediction network loaded with the weight matrix performs fault prediction on the second simulation timing array.

[0079] In some optional embodiments, the real-time operation data of the switch cabinet includes at least one of online temperature measurement data of new contacts, online detection data of opening and closing voltage and current, online monitoring data of mechanical properties, and sensing data of partial discharge sensors.

[0080] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0081] In this embodiment, a fault analysis system based on a device model is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0082] See also Figure 8 , Figure 8 A schematic diagram of the structure of a computer device provided in an embodiment of the present application is shown in FIG. Figure 8As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 8 A processor 10 is taken as an example.

[0083] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0084] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0085] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0086] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0087] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0088] The embodiment of the present application also provides a computer-readable storage medium. The above method according to the embodiment of the present application can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0089] The systems or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0090] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0091] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods and systems. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0092] The present application is described with reference to the flowcharts and / or block diagrams of the methods and systems according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0093] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0094] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0095] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0096] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0097] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

[0098] Although the embodiments of the present application have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present application, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A fault analysis method based on a device model, characterized in that: The method comprises: Loading a device model of a switch cabinet, wherein the device model includes a plurality of state sub-models of the switch cabinet in different working states, and each of the state sub-models is arranged according to a time sequence relationship of the working state; Acquire the real-time operation data of the switch cabinet, and decompose the real-time operation data into sub-data under different working modes; For any sub-data, the sub-data is sent to the corresponding morphological sub-model to run the sub-data through the morphological sub-model; According to the time series relationship, the simulation data obtained after the sub-data of each morphological sub-model is run are collected, and multiple groups of simulation time series arrays are constructed according to the collected simulation data; Generate fault analysis results for each group of simulation timing data one by one, wherein, when generating the fault analysis results of the current simulation timing array, obtain the historical fault analysis results of the previous simulation timing array before the current simulation timing data, and use the historical fault analysis results as guide labels to perform fault prediction on the current simulation timing array; A fault analysis result of the last simulation time series array is obtained, and the fault analysis result is determined as the fault analysis result of the switch cabinet.

2. The method according to claim 1, characterized in that Sending the sub-data to the corresponding morphological sub-model includes: Identify the target working form to which the sub-data belongs, and obtain the target timing identifier corresponding to the target working form; The sub-data is transmitted to the morphological sub-model having the target timing identification.

3. The method according to claim 1 or 2, characterized in that: According to the collected simulation data, constructing multiple sets of simulation timing arrays includes: Identify the timing identifiers covered by the multiple morphological sub-models, and for any current timing identifier, obtain the simulation data corresponding to the current timing identifier, and obtain the simulation data corresponding to each of the other timing identifiers before the current timing identifier, and splice the obtained simulation data according to the timing identifiers; The spliced ​​data is filled into a preset first field, and the number of spliced ​​simulation data is filled into a preset second field, and the array formed by the first field and the second field is used as the simulation timing array corresponding to the current timing identifier.

4. The method according to claim 1, characterized in that The fault analysis results generated for each set of simulation timing data include: For a first simulation timing array and a second simulation timing array that are adjacent to each other, obtaining a first fault analysis result of the first simulation timing array, and setting the first fault analysis result as a guide label of the second simulation timing array; The second simulation timing array with the guide label set is input into the trained fusion network, wherein the fusion network includes a label configuration network and a result prediction network, wherein the label configuration network extracts features of the guide label and determines a weight matrix that matches the features of the guide label, and the weight matrix is ​​loaded into the result prediction network, so that the result prediction network loaded with the weight matrix performs fault prediction on the second simulation timing array.

5. The method according to claim 4, characterized in that The matching relationship between the guide label and the weight matrix is ​​determined in the following way: Acquire a time series training sample, wherein the time series training sample includes a guide label sample, a time series array sample, and a prediction label; Extracting features from the guide label samples using a label configuration network, and fusing the extracted features with the time series array samples to obtain fused samples; Using a result prediction network to predict the fusion sample, to obtain a fault prediction result; Calculating the error between the fault prediction result and the prediction label, and adjusting the weight matrix of the result prediction network based on the error, so that when the result prediction network after the weight matrix adjustment predicts the fusion sample again, the obtained fault prediction result matches the prediction label; The guide label samples are matched with the adjusted weight matrix.

6. The method according to claim 5, characterized in that The extracted features are fused with the time series array samples to obtain fused samples including: Identify each row of data in the extracted features, and concatenate each row of data according to the row number to obtain a corresponding concatenation vector; The concatenated vector is used as a suffix of the time series array sample to form a fused sample.

7. The method according to claim 1, characterized in that The real-time operation data of the switch cabinet includes at least one of online temperature measurement data of new contacts, online detection data of opening and closing voltage and current, online monitoring data of mechanical characteristics, and sensing data of partial discharge sensors.

8. A fault analysis system based on a device model, characterized in that: The system comprises: A model loading unit, used to load a device model of the switch cabinet, wherein the device model includes a plurality of form sub-models of the switch cabinet in different working modes, and each of the form sub-models is arranged according to a time sequence relationship of the working modes; A data decomposition unit, used for acquiring the real-time operation data of the switch cabinet, and decomposing the real-time operation data into sub-data under different working modes; An operating unit, for sending any sub-data to a corresponding morphological sub-model, so as to operate the sub-data through the morphological sub-model; An array construction unit, used to collect the simulation data obtained after running the sub-data of each morphological sub-model according to the time series relationship, and construct multiple groups of simulation time series arrays according to the collected simulation data; A fault analysis unit is used to generate fault analysis results for each group of simulation timing data one by one, wherein when generating the fault analysis results of the current simulation timing array, a historical fault analysis result of the previous simulation timing array before the current simulation timing data is obtained, and the historical fault analysis result is used as a guide label to perform fault prediction on the current simulation timing array; The result determination unit is used to obtain the fault analysis result of the last simulation time series array, and determine the fault analysis result as the fault analysis result of the switch cabinet.

9. The system according to claim 8, characterized in that The fault analysis unit is specifically used to obtain a first fault analysis result of the first simulation timing array and a second adjacent simulation timing array, and set the first fault analysis result as a guide label of the second simulation timing array; input the second simulation timing array with the guide label set into the trained fusion network, the fusion network includes a label configuration network and a result prediction network, wherein the label configuration network extracts features of the guide label and determines a weight matrix that matches the features of the guide label, and the weight matrix is ​​loaded into the result prediction network, so that the result prediction network loaded with the weight matrix performs fault prediction on the second simulation timing array.

10. The system according to claim 9, characterized in that The fault analysis unit determines the matching relationship between the guide label and the weight matrix in the following manner: obtaining a time series training sample, wherein the time series training sample includes a guide label sample, a time series array sample, and a prediction label; using a label configuration network to extract features from the guide label sample, and fusing the extracted features with the time series array sample to obtain a fused sample; using a result prediction network to predict the fused sample to obtain a fault prediction result; Calculate the error between the fault prediction result and the prediction label, and adjust the weight matrix of the result prediction network based on the error, so that when the result prediction network after the weight matrix adjustment predicts the fusion sample again, the obtained fault prediction result matches the prediction label; match the guide label sample with the adjusted weight matrix.

Images