Bridge rectifier fault early warning method based on pattern learning

Abstract

The invention provides a bridge rectifier fault early warning method based on pattern learning, and relates to the technical field of bridge rectifier fault early warning, and the method comprises the steps: carrying out the tracing of a predetermined fault type through employing a plurality of bridge rectifier fault sets, and building a fault index library; a multi-source sensing monitoring array is constructed, monitoring data streams are fed back to a fault early warning unit, and each fault early warning subunit is provided with a BP prediction channel, a random forest decision channel and a support vector prediction channel which are of a parallel structure; generating a fault prediction index topology; and identifying the fault prediction index topology according to a predetermined early warning threshold, and generating a fault early warning signal topology. According to the invention, the technical problems of poor early warning accuracy and timeliness caused by incapability of timely finding potential fault hidden dangers due to insufficient intelligence degree and low analysis fineness of the fault early warning of the bridge rectifier can be solved, the accuracy and timeliness of the fault early warning of the bridge rectifier can be improved, and the fault early warning efficiency of the bridge rectifier can be improved. And the effect of ensuring stable operation of the production system is achieved.

Description

Technical Field

[0001] The present application relates to the technical field of bridge pile fault warning technology, and in particular to a bridge pile fault warning method based on pattern learning. Background Art

[0002] A bridge rectifier is a rectifier device composed of four diodes. Its main function is to convert the input AC signal into a DC signal. It is widely used in many fields, such as mobile communication equipment, industrial automation control equipment, etc. In these applications, the bridge rectifier can convert AC power into DC power, provide a stable power supply, and ensure the normal operation of industrial equipment.

[0003] At present, the traditional rectifier bridge stack fault warning method is usually to manually detect the rectifier bridge stack fault through a multimeter. This method has a low degree of intelligence and detection precision, and requires a lot of detection time. At the same time, the power needs to be cut off during detection, which affects the stability of the production system operation.

[0004] In summary, the existing rectifier bridge stack fault warning method is unable to detect potential fault hazards in a timely manner due to the insufficient intelligence of the fault warning and the low analysis precision, resulting in technical problems such as poor accuracy and timeliness of the rectifier bridge stack fault warning. Summary of the invention

[0005] The purpose of this application is to provide a bridge stack fault warning method based on pattern learning to solve the technical problem that the existing rectifier bridge stack fault warning method cannot detect potential fault hazards in time due to insufficient intelligence of the fault warning and low analysis precision, resulting in poor accuracy and timeliness of the rectifier bridge stack fault warning.

[0006] In view of the above problems, the present application provides a bridge pile fault warning method based on pattern learning, the method comprising: reading the attribute data of the target bridge pile, locating the identification and building the bridge pile structure topology; clustering and dividing the sample bridge pile fault set based on the predetermined fault type to generate multiple bridge pile fault sets; using the multiple bridge pile fault sets to perform a single traceability analysis on the predetermined fault type, and establishing a fault indicator library according to the analysis results, wherein the fault indicator library includes multiple fault type-influence indicator mapping relationships; constructing a multi-source sensor monitoring array based on the bridge pile structure topology and the fault indicator library, and feeding back the multi-source sensor monitoring data stream to a fault warning unit, the fault warning unit including multiple fault warning sub-units, the fault warning sub-units corresponding to the fault type one by one, wherein each fault warning sub-unit has a parallel structure of BP prediction channel, random forest decision channel and support vector prediction channel; predicting and analyzing the multi-source sensor monitoring data stream through the fault warning unit to generate a fault prediction index topology; identifying the fault prediction index topology according to a predetermined warning threshold, generating a fault warning signal topology, and sending the fault warning signal topology to the operation and maintenance inspection unit.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: By reading the attribute data of the target bridge pile, the bridge pile structure topology is constructed by locating the mark; the sample bridge pile fault set is clustered and divided based on the predetermined fault type to generate multiple bridge pile fault sets; the predetermined fault type is single traceable analysis is performed using the multiple bridge pile fault sets, and a fault indicator library is established according to the analysis results, wherein the fault indicator library includes multiple fault type-influence indicator mapping relationships; a multi-source sensor monitoring array is constructed based on the bridge pile structure topology and the fault indicator library, and the multi-source sensor monitoring data stream is fed back to a fault warning unit, wherein the fault warning unit includes multiple fault warning sub-units, and the fault warning sub-units correspond to the fault type one by one, wherein each fault warning sub-unit has a parallel structure of BP prediction channel, random forest decision channel and support vector prediction channel; the multi-source sensor monitoring data stream is predicted and analyzed by the fault warning unit to generate a fault prediction index topology; the fault prediction index topology is identified according to a predetermined warning threshold, a fault warning signal topology is generated, and the fault warning signal topology is sent to an operation and maintenance inspection unit. That is to say, a single traceability analysis is performed on the predetermined fault type based on the sample bridge pile fault set to build a fault index library; then, a multi-source sensor monitoring array is built according to the bridge pile structure topology and the fault index library to perform sensor monitoring on multiple rectifier bridge piles, and a multi-source sensor monitoring data stream is obtained; on the other hand, a prediction channel is built based on BP neural network, random forest decision tree and support vector machine, and the prediction channel is integrated to generate a fault warning unit, which includes multiple fault warning sub-units, each of which has a corresponding relationship with the fault type; then, the fault type is determined based on the multi-source sensor monitoring data stream, and a matching fault warning sub-unit is obtained, and a fault index prediction is performed on the multi-source sensor monitoring data stream based on the matching fault warning sub-unit to obtain a fault prediction index topology; finally, the fault prediction index topology is identified according to the predetermined warning threshold, and a fault warning signal topology is generated and sent to the operation and maintenance unit. The intelligent and refined degree of fault warning analysis of rectifier bridge piles can be improved, thereby improving the accuracy and timeliness of rectifier bridge pile fault warning, achieving the technical effect of timely discovering potential fault hazards of abnormal rectifier bridge piles, thereby reducing or avoiding damage to industrial equipment and ensuring stable operation of the production system.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0010] Figure 1 A schematic diagram of a flow chart of a bridge pile fault early warning method based on pattern learning for this application; Figure 2 This is a schematic diagram of the process of establishing a fault indicator library in a bridge pile fault early warning method based on pattern learning in this application. DETAILED DESCRIPTION

[0011] This application provides a bridge stack fault warning method based on pattern learning, which solves the existing rectifier bridge stack fault warning method. Due to the insufficient intelligence of fault warning and low analysis precision, potential fault hazards cannot be discovered in time, resulting in poor accuracy and timeliness of rectifier bridge stack fault warning. The intelligence and refinement of rectifier bridge stack fault warning analysis can be improved, thereby improving the accuracy and timeliness of rectifier bridge stack fault warning, achieving the technical effect of timely discovering potential fault hazards of abnormal rectifier bridge stacks, thereby reducing or avoiding damage to industrial equipment and ensuring stable operation of production systems.

[0012] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0013] Example Please refer to the attached Figure 1 The present application provides a bridge pile fault early warning method based on pattern learning, and the method specifically comprises the following steps: Step 1: Read the attribute data of the target bridge pile, locate the mark and build the bridge pile structure topology; Specifically, first, read the attribute data of the target bridge stack, which includes multiple rectifier bridge stacks in industrial equipment, where the industrial equipment can be selected according to actual conditions, such as large motors, large generators, large power transmission systems, etc.; wherein the attribute data includes the position coordinates of the rectifier bridge stack in the industrial equipment, and then based on the position coordinates, simulate and model multiple rectifier bridge stacks with attribute identifiers and distribute them to generate the bridge stack structure topology. By constructing the bridge stack structure topology, the precision of rectifier bridge stack fault identification can be improved, and at the same time, it provides support for the subsequent construction of a multi-source sensor monitoring array for the bridge stack structure topology.

[0014] Step 2: clustering the sample bridge pile fault set based on the predetermined fault type to generate multiple bridge pile fault sets; Specifically, a predetermined fault type is set, and the predetermined fault type includes at least a breakdown fault, an open circuit fault, and a heating fault of the rectifier bridge stack. Technical personnel in this field can also add fault types according to actual conditions. A sample bridge stack fault set is retrieved based on historical bridge stack fault records, and the sample bridge stack fault set is clustered and divided according to the predetermined fault type to obtain multiple bridge stack fault sets. By obtaining multiple bridge stack fault sets, data support is provided for the next step of tracing the fault type and determining the fault type indicator.

[0015] Step 3: Perform a single source tracing analysis on the predetermined fault type using the multiple bridge stack fault sets, and establish a fault index library according to the analysis results, wherein the fault index library includes multiple fault type-impact index mapping relationships; Specifically, based on the single analysis principle and the fault tree principle, the predetermined fault type is traced and analyzed using the multiple bridge stack fault sets to determine the multiple key influencing indicators corresponding to each fault type, where the key influencing indicators refer to indicators that have a greater impact on the rectifier bridge stack fault, such as rectifier temperature, current, voltage, abnormal vibration, etc.; then, based on the mapping association between the fault type and the multiple key influencing indicators, multiple fault type-influencing indicator mapping relationships are formed, and a fault indicator library is constructed based on the multiple fault type-influencing indicator mapping relationships. By tracing the predetermined fault type based on the single analysis principle and the fault tree principle, the accuracy of the key influencing indicators can be improved; at the same time, a fault indicator library is constructed based on the mapping association between the fault type and the multiple key influencing indicators, which provides a basis for the subsequent sensor monitoring and fault type identification of the rectifier bridge stack, and can improve the accuracy and efficiency of fault type identification.

[0016] Step 4: construct a multi-source sensor monitoring array based on the bridge pile structure topology and the fault indicator library, and feed back the multi-source sensor monitoring data stream to the fault warning unit, the fault warning unit includes a plurality of fault warning sub-units, the fault warning sub-units correspond to the fault types one by one, wherein each fault warning sub-unit has a BP prediction channel, a random forest decision channel and a support vector prediction channel of a parallel structure; Specifically, based on the multiple key influencing indicators in the fault indicator library, an adaptive monitoring sensor is set to obtain a multi-source sensor set. For example, if the key influencing indicator is the rectifier temperature, the corresponding monitoring sensor is a temperature sensor. Then, based on the bridge stack structure topology, the multi-source sensor set is configured to generate a multi-source sensor monitoring array, wherein each rectifier bridge stack corresponds to a multi-source sensor set, and each multi-source sensor set has a positioning identifier.

[0017] The multi-source sensor monitoring array is used to perform sensor monitoring on multiple rectifier bridge piles in the bridge pile structure topology to obtain a multi-source sensor monitoring data stream; a communication connection is established between the multi-source sensor monitoring array and the fault warning unit, and the multi-source sensor monitoring data stream is fed back to the fault warning unit at a preset time node. The preset time node can be set according to the actual situation, for example: feedback every 2 minutes. The fault warning unit includes multiple fault warning sub-units, and each fault warning sub-unit corresponds to a fault type one by one. In each fault warning sub-unit, an integrated BP prediction channel, a random forest decision channel, and a support vector prediction channel with a parallel structure are included, that is, the output result of each fault warning sub-unit is obtained by fusion analysis of the output results of the BP prediction channel, the random forest decision channel, and the support vector prediction channel.

[0018] By integrating the BP prediction channel, random forest decision channel and support vector prediction channel based on the principle of ensemble learning and building a fault warning sub-unit, the advantages of multiple models can be integrated to improve the accuracy, reliability and efficiency of the fault prediction index, thereby improving the accuracy, rationality and efficiency of rectifier bridge stack fault prediction.

[0019] Step 5: performing prediction analysis on the multi-source sensor monitoring data stream through the fault warning unit to generate a fault prediction index topology; Specifically, the multi-source sensor monitoring data stream is input into the fault warning unit. First, the fault type is determined based on the analysis of the multi-source sensor monitoring data stream, and a fault warning sub-unit is obtained based on the matching of the fault type; then, the multi-source sensor monitoring data stream is predicted and analyzed by the matched fault warning sub-unit, and multiple fault prediction indexes corresponding to multiple rectifier bridge stacks are obtained, and a fault prediction index topology is constructed based on the multiple fault prediction indexes. By generating a fault prediction index topology, the fault risk of each rectifier bridge stack can be clearly and intuitively obtained, providing a basis for accurate fault warning in the next step.

[0020] Step 6: Identify the fault prediction index topology according to a predetermined warning threshold, generate a fault warning signal topology, and send the fault warning signal topology to an operation and maintenance unit.

[0021] Specifically, a predetermined warning threshold is obtained, and the predetermined warning threshold includes multiple warning index thresholds corresponding to multiple fault types, wherein the fault type and the warning index threshold correspond one to one, and the warning index threshold is a standard warning index, which can be set by technical personnel in this field based on actual conditions; then, the fault prediction index topology is judged according to the predetermined warning threshold, and the rectifier bridge stack corresponding to the fault prediction index greater than the predetermined warning threshold is marked for warning, and a fault warning signal topology is generated based on the warning marking result and the fault warning type; finally, the fault warning signal topology is sent to the operation and maintenance unit, and targeted operation and maintenance are performed based on the fault warning signal topology. By generating a fault prediction index topology, the accuracy of the rectifier bridge stack fault warning can be improved, thereby improving the efficiency and accuracy of operation and maintenance.

[0022] The above method can solve the technical problem that the existing rectifier bridge stack fault warning method is not able to detect potential fault hazards in time due to the insufficient intelligence level of the fault warning and the low analysis precision, resulting in poor accuracy and timeliness of the rectifier bridge stack fault warning. First, the attribute data of the target bridge pile is read, and the bridge pile structure topology is constructed by locating the identification; then, the sample bridge pile fault set is clustered and divided based on the predetermined fault type to generate multiple bridge pile fault sets; then, the predetermined fault type is single traceability analysis is performed using the multiple bridge pile fault sets, and a fault indicator library is established according to the analysis results, wherein the fault indicator library includes multiple fault type-influencing indicator mapping relationships; next, a multi-source sensor monitoring array is constructed based on the bridge pile structure topology and the fault indicator library, and the multi-source sensor monitoring data stream is fed back to a fault warning unit, wherein the fault warning unit includes multiple fault warning sub-units, and the fault warning sub-units correspond to the fault type one by one, wherein each fault warning sub-unit has a BP prediction channel, a random forest decision channel, and a support vector prediction channel of a parallel structure; in addition, the multi-source sensor monitoring data stream is predicted and analyzed by the fault warning unit to generate a fault prediction index topology; finally, the fault prediction index topology is identified according to a predetermined warning threshold, a fault warning signal topology is generated, and the fault warning signal topology is sent to the operation and maintenance inspection unit. By conducting a single source tracing analysis on the predetermined fault type based on the sample bridge pile fault set, a fault index library is constructed; then, a multi-source sensor monitoring array is constructed according to the bridge pile structure topology and the fault index library to perform sensor monitoring on multiple rectifier bridge piles, and a multi-source sensor monitoring data stream is obtained; on the other hand, a prediction channel is constructed based on BP neural network, random forest decision tree and support vector machine, and the prediction channel is integrated to generate a fault warning unit, which includes multiple fault warning sub-units, each of which has a corresponding relationship with the fault type; then, the fault type is determined based on the multi-source sensor monitoring data stream, and a matching fault warning sub-unit is obtained, and a fault index prediction is performed on the multi-source sensor monitoring data stream based on the matching fault warning sub-unit to obtain a fault prediction index topology; finally, the fault prediction index topology is identified according to the predetermined warning threshold, and a fault warning signal topology is generated and sent to the operation and maintenance unit. The intelligent and refined degree of fault warning analysis of rectifier bridge piles can be improved, thereby improving the accuracy and timeliness of rectifier bridge pile fault warning, achieving the technical effect of timely discovering potential fault hazards of abnormal rectifier bridge piles, thereby reducing or avoiding damage to industrial equipment and ensuring stable operation of the production system.

[0023] Further, the target bridge pile attribute data is read, the bridge pile structure topology is constructed by locating the mark, and step one of the present application includes: Collecting attribute data of the target bridge stack, the target bridge stack includes multiple rectifier bridge stacks in industrial equipment, the attribute data includes model specifications, electrical parameters, aging index and location coordinates; In the simulation platform, attribute identification is performed on the plurality of rectifier bridge stacks according to the attribute data to generate a plurality of identified rectifier bridge stacks; The plurality of identified rectifier bridge stacks are distributed and arranged based on the position coordinates to generate the bridge stack structure topology.

[0024] Specifically, first, the attribute data of the target bridge stack is collected, wherein the target bridge stack includes multiple rectifier bridge stacks in industrial equipment; wherein the attribute data includes model specifications, electrical parameters, aging index and position coordinates, the electrical parameters include current threshold, voltage threshold, etc., the aging index is set based on the remaining service life of the rectifier bridge stack, wherein the remaining service life is inversely proportional to the aging index, and the position coordinates refer to the specific position of the rectifier bridge stack in the industrial equipment.

[0025] The simulation platform is a software tool specifically used to build, simulate and analyze various systems, processes and phenomena. It combines mathematical modeling, simulation algorithms, graphical interfaces and other aspects, allowing users to conduct various complex simulations and experiments more conveniently and intuitively. In the simulation platform, the multiple rectifier bridge stacks are attributed according to the attribute data to obtain multiple identified rectifier bridge stacks. Then, the multiple identified rectifier bridge stacks are distributed and arranged based on the position coordinates to obtain the bridge stack structure topology.

[0026] Furthermore, the sample bridge stack fault set is clustered and divided based on the predetermined fault type to generate multiple bridge stack fault sets. Step 2 of the present application includes: Acquiring operating characteristics of the target bridge stack, wherein the operating characteristics include operating environment characteristics and workload characteristics; Taking the model specification, the electrical parameters and the working characteristics as constraints, searching historical bridge stack fault records to obtain a sample bridge stack fault set; Obtaining a predetermined fault type, wherein the predetermined fault type at least includes a breakdown fault, an open circuit fault, and a heating fault; The sample bridge stack fault sets are clustered and divided based on the predetermined fault types to obtain the multiple bridge stack fault sets, and the bridge stack fault sets correspond to the fault types one by one.

[0027] Specifically, first, the working characteristics of the target bridge stack are obtained, wherein the working characteristics include working environment characteristics and workload characteristics, wherein the working environment characteristics include working environment temperature, working environment humidity, etc., and the workload characteristics include working status, daily load, etc. Then, the model specifications, electrical parameters, and working characteristics are used as retrieval constraints to retrieve and screen historical bridge stack fault records to obtain a sample bridge stack fault set. By setting the model specifications, electrical parameters, and working characteristics as retrieval constraints, the compatibility of the sample bridge stack fault set with the target bridge stack can be improved, thereby improving the accuracy and reliability of the sample bridge stack fault set.

[0028] Obtain predetermined fault types, wherein the predetermined fault types include at least breakdown faults, open circuit faults and heating faults, and new fault types may also be added according to actual conditions. Then, the sample bridge stack fault sets are clustered and divided based on the predetermined fault types to obtain the multiple bridge stack fault sets, wherein the bridge stack fault sets correspond to the fault types one by one.

[0029] Further, the plurality of bridge stack fault sets are used to perform a single source tracing analysis on the predetermined fault type, and a fault indicator library is established according to the analysis results, as shown in the attached Figure 2 As shown, step three of this application includes: Performing standardization processing on the multiple bridge pile fault sets based on a predetermined data processing strategy to obtain multiple standard bridge pile fault sets, wherein the predetermined data processing strategy includes a predetermined data cleaning strategy and a predetermined dimensionless processing strategy; Selecting a first fault type from the predetermined fault types, and obtaining a first standard bridge stack fault set corresponding to the first fault type; Based on the fault tree principle, taking the first fault type as the top event, tracing and analyzing the first standard bridge pile fault set in turn to generate multiple fault influencing factor sequences; configuring factor weights in the factor sequence, fusing the multiple fault impact factor sequences based on the factor weights to generate a first fault impact factor sequence, wherein the factor weights gradually decrease as the factor sequence decreases from the earliest to the latest; Extracting the influencing factors of preset rankings in the first fault influencing factor sequence to determine a first key influencing indicator; A first mapping relationship between the first fault type and the first key influencing indicator is established, and the fault indicator library is constructed according to the first mapping relationship.

[0030] Specifically, first, a predetermined data processing strategy is obtained, wherein the predetermined data processing strategy includes a predetermined data cleaning strategy and a predetermined dimensionless processing strategy, wherein the predetermined data cleaning strategy includes at least redundant elimination, anomaly correction and missing supplementation methods, and the predetermined dimensionless processing strategy refers to converting data into dimensionless data through certain mathematical transformations, thereby facilitating data comparison and analysis, and commonly used dimensionless processing methods include normalization method, standard method, mean method, etc. Then, based on the predetermined data processing strategy, the multiple bridge stack fault sets are standardized to obtain multiple standard bridge stack fault sets. By configuring a predetermined data processing strategy to perform standardized processing on multiple bridge stack fault sets, noise data interference can be removed, and the accuracy and reliability of the bridge stack fault set can be improved, thereby improving the accuracy of subsequent analysis of bridge stack fault impact indicators.

[0031] A first fault type is selected from the predetermined fault types, the first fault type being any one of the predetermined fault types, and a first standard bridge stack fault set corresponding to the first fault type is obtained. The fault tree principle is based on a top-down deductive failure analysis method, which takes the fault state of the rectifier bridge stack as the target of identification and estimation, that is, the top event, and then finds out the necessary and sufficient direct cause of the top event as the second-level event, and then finds out the direct factor that causes the second-level fault event as the third-level event, and expands step by step until the final fault influencing factor is determined. Based on the fault tree principle, the first fault type is taken as the top event, and the first standard bridge stack fault set is traced and analyzed in turn to determine multiple fault influencing factor sets, and the fault influencing factors are arranged from large to small based on the degree of influence of the fault influencing factors to obtain multiple fault influencing factor sequences.

[0032] Configure the factor weights in the factor sequence, wherein the factor weights gradually decrease as the factor order decreases from first to last, that is, the higher the factor ranking, the greater the corresponding weight; configure the weights of the fault impact factors in the multiple fault impact factor sequences according to the factor weights, and perform weighted summation based on the weight configuration results and the frequency of occurrence of the factors, and arrange the fault impact factors from large to small according to the weighted summation results to generate a first fault impact factor sequence. Then extract the impact factors of a preset ranking in the first fault impact factor sequence, and the preset ranking can be set based on actual conditions, such as setting the preset ranking to 3, that is, extracting the top 3 fault impact factors in the first fault impact factor sequence, and setting the fault impact factor as the first key impact indicator. Then establish a first mapping relationship between the first fault type and the first key impact indicator, and obtain multiple mapping relationships in sequence to construct the fault indicator library.

[0033] Further, step 4 of this application includes: Obtain a prediction operator, and build an initial BP prediction channel, an initial random forest decision channel, and an initial support vector prediction channel based on the prediction operator; Generate multiple training sample sets based on the multiple standard bridge stack fault sets and the fault indicator library; Specifically, first, a prediction operator is obtained, and the prediction operator includes a BP neural network, a random forest decision tree, and a support vector machine. The BP neural network is a multi-layer feedforward neural network trained according to the error back propagation algorithm. Its working principle is to optimize the weights and biases in the network through the back propagation algorithm to minimize the prediction error. It has the ability to approximate any function, can learn nonlinear relationships, and has a certain generalization ability; random forest is an integrated learning algorithm, which is integrated by multiple decision trees. It improves the prediction accuracy and stability by building multiple decision trees and integrating their prediction results. Each decision tree is built based on a randomly sampled subset of training data and randomly selected features, which helps to increase the diversity and generalization ability of the model; support vector machine is a supervised learning algorithm for classifying and predicting data. Then, based on the prediction operator, the initial BP prediction channel, the initial random forest decision channel, and the initial support vector prediction channel are respectively built. Further, multiple training sample sets are generated through the multiple standard bridge pile fault sets and the fault indicator library.

[0034] Furthermore, the present application also includes the following steps: Selecting a first standard bridge stack fault set and a first mapping relationship, and obtaining a first fault type and a first key impact indicator of the first mapping relationship; Extracting information from the first standard bridge pile fault set based on the first key influencing indicator to establish a first key data set; Sequentially evaluating the criticality and trend slope of the bridge pile faults in the first key bridge pile fault data set, and obtaining a first fault index set by weighted calculation based on the evaluation results; The first key data set and the first fault index set are mapped and fused to obtain a first training sample set, and the first training sample set is added to the multiple training sample sets.

[0035] Specifically, the method for generating multiple training sample sets based on the multiple standard bridge stack fault sets and the fault index library is as follows: first, select the first standard bridge stack fault set and the first mapping relationship, and obtain the first fault type and the first key influencing indicator of the first mapping relationship, the first standard bridge stack fault set and the first fault type have a mapping association, and the first standard bridge stack fault set is any one of the multiple standard bridge stack fault sets. Then, based on the first key influencing indicator, information is extracted from the first standard bridge stack fault set to obtain the first key data set. Then, the bridge pile faults in the first key bridge pile fault data set are evaluated for hazard and trend slope in turn to obtain hazard evaluation results and trend slope evaluation results, wherein the hazard is set based on the destructiveness caused by the bridge pile fault, wherein the greater the destructiveness, the greater the hazard; the trend slope is set based on the occurrence rate of the rectifier bridge pile fault, wherein the faster the occurrence rate, the greater the trend slope; weight coefficients are configured for the hazard and trend slope respectively, wherein the weight coefficient is set based on the impact of the indicator on the overall fault, wherein the greater the impact, the greater the corresponding weight coefficient; based on the weight coefficient, the hazard evaluation results and trend slope evaluation results are weightedly calculated, and the weighted calculation results are used as the fault index to obtain the first fault index set. Finally, the first key data set and the first fault index set are mapped and fused, that is, the first key data and the first fault index are mapped to each other, to obtain the first training sample set and add it to multiple training sample sets to obtain the multiple sample sets.

[0036] Selecting a first training sample set corresponding to a first fault type, and performing equal value division on the first training sample set to obtain a plurality of divided training sample sets; Based on the multiple divided training sample sets, supervised training and cross-validation are performed on the initial BP prediction channel, the initial random forest decision channel and the initial support vector prediction channel to obtain converged BP prediction channel, converged random forest decision channel and converged support vector prediction channel that meet the expected constraints and the corresponding convergence degree; An integrated fusion strategy is set according to the convergence degree, and based on the integrated fusion strategy, the converged BP prediction channel, the converged random forest decision channel and the converged support vector prediction channel are integrated to obtain a first fault warning subunit; Specifically, a first training sample set corresponding to a first fault type is selected, and the first fault type is any one of the predetermined fault types; then the first training sample set is equally divided to obtain a plurality of divided training sample sets. The plurality of divided training sample sets are used as training data, and supervised training is performed on the initial BP prediction channel, the initial random forest decision channel, and the initial support vector prediction channel respectively, and after training to a preset number of thresholds, the training data is exchanged for cross-validation, and the output accuracy of the initial BP prediction channel, the initial random forest decision channel, and the initial support vector prediction channel is obtained based on the cross-validation results, and the output accuracy is judged based on the expected constraints. If the output accuracy does not meet the expected constraints, the training data is exchanged to continue iterative training; if it meets the expected constraints, the converged BP prediction channel, the converged random forest decision channel, the converged support vector prediction channel, and the final output accuracy that meet the expected constraints are obtained, and the final output accuracy is set as the convergence degree.

[0037] An integrated fusion strategy is set according to the convergence degree, and then the convergence BP prediction channel, the convergence random forest decision channel and the convergence support vector prediction channel are integrated and fused based on the integrated fusion strategy to obtain a first fault warning subunit, wherein the integrated fusion strategy is a determination strategy for the output result of the first fault warning subunit.

[0038] Furthermore, the present application also includes the following steps: Determine a first convergence degree, a second convergence degree, and a third convergence degree of the convergent BP prediction channel, the convergent random forest decision channel, and the convergent support vector prediction channel; configuring a first weight coefficient, a second weight coefficient, and a third weight coefficient based on the first convergence degree, the second convergence degree, and the third convergence degree; The first weight coefficient, the second weight coefficient and the third weight coefficient are set as the confidence of the channel output result to obtain the integrated fusion strategy.

[0039] Specifically, the method for setting the integrated fusion strategy according to the convergence degree is as follows: first, determine the first convergence degree, the second convergence degree and the third convergence degree of the convergent BP prediction channel, the convergent random forest decision channel and the convergent support vector prediction channel. Then, configure the first weight coefficient, the second weight coefficient and the third weight coefficient according to the first convergence degree, the second convergence degree and the third convergence degree, wherein the convergence degree is positively correlated with the weight coefficient, that is, the greater the convergence degree, the greater the corresponding weight coefficient. Finally, the first weight coefficient, the second weight coefficient and the third weight coefficient are set as the confidence of the channel output result, and the integrated fusion strategy is obtained, that is, the output result of the fault warning subunit is the weighted sum of the output results of the convergent BP prediction channel, the convergent random forest decision channel and the convergent support vector prediction channel.

[0040] The fault early warning unit is generated based on the first fault type and the first fault early warning subunit.

[0041] Specifically, a mapping association is established between the first fault type and the first fault warning subunit, a plurality of fault warning subunits of a plurality of fault types are obtained in sequence, and the fault warning unit is constructed based on the mapping association fusion.

[0042] Further, the multi-source sensor monitoring data stream is predicted and analyzed by the fault warning unit to generate a fault prediction index topology. Step five of the present application includes: Selecting first multi-source sensor monitoring data of a first rectifier bridge stack, identifying the first multi-source sensor monitoring data based on a preset standard value, and determining first abnormal monitoring data; Determining a first matching fault type and a first matching fault early warning subunit based on the first abnormal monitoring data, and performing a prediction analysis on the first abnormal monitoring data by the first matching fault early warning subunit to obtain a first fault prediction index; The first rectifier bridge stack, the first matching fault type and the first fault prediction index are integrated to generate the fault prediction index topology.

[0043] Specifically, first, select the first multi-source sensor monitoring data of the first rectifier bridge stack, the first rectifier bridge stack is any rectifier bridge stack in the bridge stack structure topology; obtain a predetermined standard value, the predetermined standard value is the maximum value of the key influencing indicator under normal conditions, and can be set according to actual conditions; then identify the first multi-source sensor monitoring data based on the preset standard value, and mark the monitoring data greater than the preset standard value as abnormal, and obtain the first abnormal monitoring data. Then, reverse match the first abnormal monitoring data based on the fault indicator library to determine the first matching fault type, and determine the first matching fault warning subunit based on the first matching fault type. Then input the first abnormal monitoring data into the first matching fault warning subunit, and output the first fault prediction index. Finally, integrate the first rectifier bridge stack, the first matching fault type and the first fault prediction index to obtain the fault prediction index topology.

[0044] Furthermore, before identifying the fault prediction index topology according to the predetermined warning threshold, step six of the present application also includes: determining a plurality of initial warning thresholds of the predetermined fault type; Identify the rectifier bridge stack aging index corresponding to the fault prediction index in the fault prediction index topology; The multiple initial warning thresholds are adaptively adjusted based on the aging index of the rectifier bridge stack to obtain the predetermined warning threshold, wherein the warning threshold and the rectifier bridge stack have a one-to-one correspondence.

[0045] Specifically, first, multiple initial warning thresholds of the predetermined fault type are determined, and the initial warning threshold is a standard warning index, which can be set by technicians in this field according to actual conditions. Then, the corresponding rectifier bridge stack aging index of the fault prediction index in the fault prediction index topology is identified, and then the optimization coefficient is determined according to the rectifier bridge stack aging index, and the multiple initial warning thresholds are optimized and adjusted according to the optimization coefficient to obtain the predetermined warning threshold, where the optimization coefficient is a decimal between 0 and 1, and is inversely proportional to the aging index, that is, the larger the aging index, the smaller the optimization coefficient, and the warning threshold corresponds to the rectifier bridge stack one by one.

[0046] By optimizing and correcting the initial warning threshold based on the aging degree of the rectifier bridge stack, the adaptability of the preset warning threshold to the actual state of the rectifier bridge stack can be improved, thereby improving the accuracy and rationality of the rectifier bridge stack fault prediction.

[0047] In summary, the bridge pile fault early warning method based on pattern learning provided by the present application has the following technical effects: 1. Through single source tracing analysis of the predetermined fault type based on the sample bridge pile fault set, a fault index library is constructed; then, according to the bridge pile structure topology and the fault index library, a multi-source sensor monitoring array is built to perform sensor monitoring on multiple rectifier bridge piles to obtain a multi-source sensor monitoring data stream; on the other hand, a prediction channel is built based on BP neural network, random forest decision tree and support vector machine, and the prediction channel is integrated to generate a fault warning unit, which includes multiple fault warning sub-units, each of which has a corresponding relationship with the fault type; then, the fault type is determined based on the multi-source sensor monitoring data stream, and a matching fault warning sub-unit is obtained, and the fault index prediction of the multi-source sensor monitoring data stream is performed based on the matching fault warning sub-unit to obtain a fault prediction index topology; finally, the fault prediction index topology is identified according to the predetermined warning threshold, and a fault warning signal topology is generated and sent to the operation and maintenance unit. The intelligent and refined degree of fault warning analysis of rectifier bridge piles can be improved, thereby improving the accuracy and timeliness of fault warning of rectifier bridge piles, achieving the technical effect of timely discovering potential fault hazards of abnormal rectifier bridge piles, thereby reducing or avoiding damage to industrial equipment and ensuring stable operation of the production system.

[0048] 2. By integrating the BP prediction channel, random forest decision channel and support vector prediction channel based on the principle of integrated learning and building a fault warning sub-unit, the advantages of multiple models can be integrated to improve the accuracy, reliability and efficiency of the fault prediction index, thereby improving the accuracy, rationality and efficiency of rectifier bridge stack fault prediction.

[0049] 3. By optimizing and correcting the initial warning threshold based on the aging degree of the rectifier bridge stack, the adaptability of the predetermined warning threshold to the actual state of the rectifier bridge stack can be improved, thereby improving the accuracy and rationality of the rectifier bridge stack fault prediction.

[0050] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0051] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A bridge pile fault early warning method based on pattern learning, characterized in that: The method comprises: Read the attribute data of the target bridge pile, locate and identify the bridge pile structure topology; Clustering the sample bridge pile fault set based on the predetermined fault type to generate multiple bridge pile fault sets; Performing a single source tracing analysis on the predetermined fault type using the multiple bridge stack fault sets, and building a fault index library according to the analysis results, wherein the fault index library includes multiple fault type-impact index mapping relationships; A multi-source sensor monitoring array is constructed based on the bridge pile structure topology and the fault indicator library, and the multi-source sensor monitoring data stream is fed back to a fault warning unit, wherein the fault warning unit includes a plurality of fault warning sub-units, and the fault warning sub-units correspond to the fault types one by one, wherein each fault warning sub-unit has a BP prediction channel, a random forest decision channel, and a support vector prediction channel of a parallel structure; Performing prediction analysis on the multi-source sensor monitoring data stream through the fault warning unit to generate a fault prediction index topology; The fault prediction index topology is identified according to a predetermined warning threshold, a fault warning signal topology is generated, and the fault warning signal topology is sent to an operation and maintenance unit.

2. The method according to claim 1, characterized in that The method of reading the attribute data of the target bridge pile and locating and identifying the bridge pile structure topology includes: Collecting attribute data of the target bridge stack, the target bridge stack includes multiple rectifier bridge stacks in industrial equipment, the attribute data includes model specifications, electrical parameters, aging index and location coordinates; In the simulation platform, attribute identification is performed on the plurality of rectifier bridge stacks according to the attribute data to generate a plurality of identified rectifier bridge stacks; The plurality of identified rectifier bridge stacks are distributed and arranged based on the position coordinates to generate the bridge stack structure topology.

3. The method according to claim 2, characterized in that The clustering of the sample bridge stack fault set based on the predetermined fault type to generate multiple bridge stack fault sets includes: Acquiring operating characteristics of the target bridge stack, wherein the operating characteristics include operating environment characteristics and workload characteristics; Taking the model specification, the electrical parameters and the working characteristics as constraints, searching historical bridge stack fault records to obtain a sample bridge stack fault set; Obtaining a predetermined fault type, wherein the predetermined fault type at least includes a breakdown fault, an open circuit fault, and a heating fault; The sample bridge stack fault sets are clustered and divided based on the predetermined fault types to obtain the multiple bridge stack fault sets, and the bridge stack fault sets correspond to the fault types one by one.

4. The method according to claim 3, characterized in that: The method of using the multiple bridge stack fault sets to perform a single source tracing analysis on the predetermined fault type and establishing a fault indicator library according to the analysis results includes: Performing standardization processing on the multiple bridge pile fault sets based on a predetermined data processing strategy to obtain multiple standard bridge pile fault sets, wherein the predetermined data processing strategy includes a predetermined data cleaning strategy and a predetermined dimensionless processing strategy; Selecting a first fault type from the predetermined fault types, and obtaining a first standard bridge stack fault set corresponding to the first fault type; Based on the fault tree principle, taking the first fault type as the top event, tracing and analyzing the first standard bridge pile fault set in turn to generate multiple fault influencing factor sequences; configuring factor weights in the factor sequence, fusing the multiple fault impact factor sequences based on the factor weights to generate a first fault impact factor sequence, wherein the factor weights gradually decrease as the factor sequence decreases from the earliest to the latest; Extracting the influencing factors of preset rankings in the first fault influencing factor sequence to determine a first key influencing indicator; A first mapping relationship between the first fault type and the first key influencing indicator is established, and the fault indicator library is constructed according to the first mapping relationship.

5. The method according to claim 4, characterized in that The method further comprises: Obtain a prediction operator, and build an initial BP prediction channel, an initial random forest decision channel, and an initial support vector prediction channel based on the prediction operator; Generate multiple training sample sets based on the multiple standard bridge stack fault sets and the fault indicator library; Selecting a first training sample set corresponding to a first fault type, and performing equal value division on the first training sample set to obtain a plurality of divided training sample sets; Based on the multiple divided training sample sets, supervised training and cross-validation are performed on the initial BP prediction channel, the initial random forest decision channel, and the initial support vector prediction channel to obtain converged BP prediction channels, converged random forest decision channels, and converged support vector prediction channels that meet expected constraints and corresponding convergence degrees; An integrated fusion strategy is set according to the convergence degree, and based on the integrated fusion strategy, the converged BP prediction channel, the converged random forest decision channel and the converged support vector prediction channel are integrated to obtain a first fault warning subunit; The fault early warning unit is generated based on the first fault type and the first fault early warning subunit.

6. The method according to claim 5, characterized in that The generating a plurality of training sample sets based on the plurality of standard bridge stack fault sets and the fault indicator library comprises: Selecting a first standard bridge stack fault set and a first mapping relationship, and obtaining a first fault type and a first key impact indicator of the first mapping relationship; Extracting information from the first standard bridge pile fault set based on the first key influencing indicator to establish a first key data set; Sequentially evaluating the criticality and trend slope of the bridge pile faults in the first key bridge pile fault data set, and obtaining a first fault index set by weighted calculation based on the evaluation results; The first key data set and the first fault index set are mapped and fused to obtain a first training sample set, and the first training sample set is added to the multiple training sample sets.

7. The method according to claim 5, characterized in that The step of setting an integrated fusion strategy according to the convergence degree includes: Determine a first convergence degree, a second convergence degree, and a third convergence degree of the convergent BP prediction channel, the convergent random forest decision channel, and the convergent support vector prediction channel; configuring a first weight coefficient, a second weight coefficient, and a third weight coefficient based on the first convergence degree, the second convergence degree, and the third convergence degree; The first weight coefficient, the second weight coefficient and the third weight coefficient are set as the confidence of the channel output result to obtain the integrated fusion strategy.

8. The method according to claim 2, characterized in that: The performing predictive analysis on the multi-source sensor monitoring data stream by the fault early warning unit to generate a fault prediction index topology includes: Selecting first multi-source sensor monitoring data of a first rectifier bridge stack, identifying the first multi-source sensor monitoring data based on a preset standard value, and determining first abnormal monitoring data; Determining a first matching fault type and a first matching fault early warning subunit based on the first abnormal monitoring data, and performing a prediction analysis on the first abnormal monitoring data by the first matching fault early warning subunit to obtain a first fault prediction index; The first rectifier bridge stack, the first matching fault type and the first fault prediction index are integrated to generate the fault prediction index topology.

9. The method according to claim 2, characterized in that: The step of identifying the fault prediction index topology according to a predetermined warning threshold value also includes: determining a plurality of initial warning thresholds of the predetermined fault type; Identify the rectifier bridge stack aging index corresponding to the fault prediction index in the fault prediction index topology; The multiple initial warning thresholds are adaptively adjusted based on the aging index of the rectifier bridge stack to obtain the predetermined warning threshold, wherein the warning threshold and the rectifier bridge stack have a one-to-one correspondence.

Images