Mechanical and electrical equipment fault prediction method, system and intelligent operation and maintenance gateway based on hybrid expert model

Abstract

The present invention relates to a method, system and intelligent operation and maintenance gateway for electromechanical equipment fault prediction that integrates a hybrid expert model, and relates to the field of intelligent predictive maintenance, including: receiving an electromechanical equipment operating state matrix; matching a hybrid expert model according to an equipment model identifier; traversing the expert model layer for priori to obtain a list of fault prediction accuracy rates; performing expert model integration configuration according to the list of fault prediction accuracy rates to obtain an expert model integration scheme; sending the expert model integration scheme to a gated network, processing the electromechanical equipment operating state matrix, and obtaining an electromechanical equipment predicted fault element. By receiving the equipment operating state matrix, matching the corresponding hybrid expert model, traversing the expert model layer to obtain the fault prediction accuracy rate, and then performing expert model integration configuration, the integrated model is used to process the equipment operating state, thereby achieving a technical effect of improving the accuracy of electromechanical equipment fault prediction.

Description

Technical Field

[0001] The present invention relates to the field of intelligent predictive maintenance, and in particular to a method and system for predicting faults of electromechanical equipment integrating a hybrid expert model, and an intelligent operation and maintenance gateway. Background Art

[0002] As a key facility in industrial production, the operation failure of electromechanical equipment will directly lead to a decrease in production efficiency. In the prior art, hybrid expert models are widely used in the field of electromechanical equipment fault prediction as an integrated learning method. The model combines multiple expert sub-models and uses a gating network to weightedly fuse the outputs of each expert, thereby improving the prediction performance. However, in the actual application of traditional hybrid expert models, each expert model is often trained independently for different tasks or data, and there is a lack of effective information interaction and coordination mechanisms between them. This leads to relatively isolated work between expert models, which cannot give full play to the advantages of integrated learning and affects the accuracy of electromechanical equipment fault prediction. Summary of the invention

[0003] The present invention aims to solve the technical problem in the prior art that when traditional hybrid expert models are applied, each expert model works relatively independently and cannot achieve synergy, resulting in low accuracy in electromechanical equipment fault prediction. The present invention provides an electromechanical equipment fault prediction method, system and intelligent operation and maintenance gateway that integrate hybrid expert models to solve the problem.

[0004] The technical solution of the present invention to solve the above technical problems is as follows:

[0005] In a first aspect, the present invention provides a method for predicting electromechanical equipment faults by integrating a hybrid expert model, comprising: receiving an electromechanical equipment operating status matrix, wherein the electromechanical equipment operating status matrix has an equipment model identification, an equipment component service life identification list, and an equipment component maintenance number identification list; matching a hybrid expert model according to the equipment model identification, wherein the hybrid expert model has an expert model layer and a gating network; traversing the expert model layer for priori according to the equipment model identification, the equipment component service life identification list, and the equipment component maintenance number identification list to obtain a fault prediction accuracy list; performing expert model integration configuration according to the fault prediction accuracy list to obtain an expert model integration solution; sending the expert model integration solution to the gating network, processing the electromechanical equipment operating status matrix, and obtaining an electromechanical equipment predicted fault component.

[0006] In a second aspect, the present invention provides an electromechanical equipment fault prediction system integrating a hybrid expert model, comprising: a data receiving module, used to receive an electromechanical equipment operation status matrix, wherein the electromechanical equipment operation status matrix has an equipment model identification, an equipment component service life identification list, and an equipment component maintenance number identification list; a model matching module, used to match the hybrid expert model according to the equipment model identification, wherein the hybrid expert model has an expert model layer and a gating network; an accuracy evaluation module, used to traverse the expert model layer for priori according to the equipment model identification, the equipment component service life identification list, and the equipment component maintenance number identification list, and obtain a fault prediction accuracy list; an integrated configuration module, used to perform expert model integration configuration according to the fault prediction accuracy list to obtain an expert model integration plan; a fault prediction module, used to send the expert model integration plan to the gating network, process the electromechanical equipment operation status matrix, and obtain the electromechanical equipment predicted fault component.

[0007] On the third aspect, the present application provides an intelligent operation and maintenance gateway for electromechanical equipment fault prediction that integrates a hybrid expert model, which is used to implement the provided electromechanical equipment fault prediction method that integrates a hybrid expert model.

[0008] The beneficial effects of the present invention are:

[0009] Receive the electromechanical equipment operation status matrix, where the electromechanical equipment operation status matrix has the equipment model identification, the equipment component service life identification list and the equipment component maintenance number identification list, which provides necessary data for subsequent matching of expert models and fault prediction; match the hybrid expert model according to the equipment model identification, where the hybrid expert model has an expert model layer and a gating network, and select the corresponding hybrid expert model according to different equipment models, which not only takes into account the particularity of the equipment model, but also utilizes the advantages of the expert model layer and the gating network in the hybrid expert model, laying the foundation for improving the fault prediction accuracy; according to the equipment model identification, the equipment component service life identification list and the equipment component maintenance number identification list, The expert model layer is traversed to perform a priori prediction, and a list of fault prediction accuracy is obtained. The equipment operation status information is fully utilized to traverse and perform a priori prediction on each expert model in the expert model layer, and the evaluation results of different experts on the fault prediction accuracy are obtained, which provides a quantitative basis for the subsequent expert model integration. The expert model integration configuration is performed according to the fault prediction accuracy list to obtain the expert model integration solution, which gives full play to the prediction advantages of different experts and improves the fault prediction ability of the integrated model. The expert model integration solution is sent to the gated network, the electromechanical equipment operation status matrix is ​​processed, and the electromechanical equipment predicted fault components are obtained, thereby improving the accuracy and efficiency of electromechanical equipment fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1A schematic diagram of the flow chart of the electromechanical equipment fault prediction method integrating the hybrid expert model provided by the present invention;

[0011] Figure 2 A structural schematic diagram of the electromechanical equipment fault prediction system integrating the hybrid expert model provided by the present invention.

[0012] In the accompanying drawings, the components represented by the reference numerals are as follows:

[0013] Data receiving module 11, model matching module 12, accuracy evaluation module 13, integrated configuration module 14, fault prediction module 15. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0015] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0016] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present invention.

[0017] Embodiment 1:

[0018] like Figure 1 As shown, the embodiment of the present invention provides a method for predicting mechanical and electrical equipment faults by integrating a hybrid expert model, including:

[0019] S100: Receive an electromechanical equipment operation status matrix, wherein the electromechanical equipment operation status matrix has an equipment model identification, an equipment component service life identification list, and an equipment component maintenance number identification list.

[0020] Specifically, first, the electromechanical equipment operation status matrix is ​​received. Among them, the electromechanical equipment can be various types of mechanical and electrical equipment, such as generators, transformers, motors, pumps, compressors, etc. The electromechanical equipment operation status matrix is ​​a data structure used to characterize the current working conditions of the equipment. On the one hand, the electromechanical equipment operation status matrix has an equipment model identification, which is used to distinguish equipment of different types or models; on the other hand, the electromechanical equipment operation status matrix contains multiple lists, namely, the equipment component service life identification list and the equipment component maintenance number identification list, which correspond to the service life and historical maintenance number of each component of the equipment. Among them, the service life is the cumulative time since the commissioning or the equivalent time converted to the current time, which reflects the degree of wear and aging of the component; the number of maintenance times reflects the reliability level and failure risk of the component.

[0021] By obtaining the operating status matrix of electromechanical equipment, the necessary basis is provided for subsequent expert model matching and fault prediction, thereby ensuring the pertinence and accuracy of fault prediction.

[0022] S200: Matching a hybrid expert model according to the device model identifier, wherein the hybrid expert model has an expert model layer and a gating network.

[0023] Specifically, after obtaining the operating state matrix of the electromechanical equipment, the equipment model identifier contained therein is extracted, and the hybrid expert model corresponding to the equipment model is matched from the pre-built hybrid expert model library. The hybrid expert model library is a data structure for storing and managing multiple hybrid expert models. In the process of building the hybrid expert model library, each hybrid expert model is associated with one or more equipment model identifiers. For example, it is represented and stored in the form of a key-value pair, where the key is the equipment model identifier and the value is one or more hybrid expert models corresponding to the equipment model.

[0024] Among them, the hybrid expert model has an expert model layer and a gating network. The expert model layer consists of multiple parallel sub-models, each of which is called an expert model. These expert models can use different machine learning algorithms, such as decision trees, support vector machines, neural networks, etc.; they can also be multiple instances of the same algorithm with differences in training data, feature engineering, hyperparameter configuration, etc. This structural design enables each expert model to have different prediction capabilities in different areas of the state space. When faced with a specific fault prediction task, each expert model can give its own judgment results from different angles. The gating network plays the role of task scheduling in the hybrid expert model. It is responsible for assigning a weight coefficient to each expert model according to the expert model integration scheme, and weighted summing the output results of all expert models according to the weight to generate the final fault prediction output.

[0025] By mapping and matching the device model identification with the hybrid expert model, necessary model support is provided for the follow-up.

[0026] S300: According to the equipment model identifier, the equipment component service life identifier list and the equipment component maintenance times identifier list, traverse the expert model layer for priori to obtain a fault prediction accuracy list.

[0027] Specifically, in order to comprehensively evaluate the prediction performance of each expert model under different equipment states, a priori data set similar to the current electromechanical equipment to be predicted is constructed based on the equipment model identification, the equipment component service life identification list, and the equipment component maintenance number identification list. Among them, by setting a similarity threshold, the electromechanical equipment operation state matrix is ​​matched with the equipment operation state matrix in the historical data set, and cases with similar service status and operation and maintenance status are screened out as the priori data set of the current equipment. After obtaining the priori data set, each expert model in the expert model layer is traversed, and it is used to predict faults on the equipment state matrix in the priori data set, and the prediction results are compared with the actual fault conditions of the corresponding equipment, thereby calculating the fault prediction accuracy of the expert model. The higher the fault prediction accuracy, the better the prediction performance of the expert model on the current priori data set.

[0028] After obtaining the fault prediction accuracy of each expert model, add it to the fault prediction accuracy list according to the expert model number, forming a one-to-one correspondence between the expert model and its prediction performance. The fault prediction accuracy list provides an important basis for the subsequent expert model integration.

[0029] S400: Perform expert model integration configuration according to the fault prediction accuracy list to obtain an expert model integration solution.

[0030] Specifically, the obtained fault prediction accuracy list is used to calculate the weight coefficient of each expert model, thereby forming an expert model integration solution.

[0031] First, according to the preset fault prediction accuracy threshold, the fault prediction accuracy list is screened, and the target expert model with the required accuracy is selected to form a preliminary integrated expert model subset. After obtaining the integrated expert model subset, the weight coefficient of each expert model in the integrated prediction is further determined. For example, according to the fault prediction accuracy of each expert model, it is linearly normalized into a weight coefficient to ensure that the sum of all weights is 1. The generated integrated expert model subset and its weight coefficient are encapsulated to form a complete expert model integration solution.

[0032] By reprocessing and optimizing the fault prediction accuracy list, the prediction capability contained in the expert model layer is explored and utilized to the maximum extent, so that the hybrid expert model can dynamically adapt when dealing with complex and changeable electromechanical equipment fault prediction tasks, thereby effectively improving the overall level of fault prediction.

[0033] S500: Send the expert model integration solution to the gate control network, process the electromechanical equipment operation state matrix, and obtain the predicted fault components of the electromechanical equipment.

[0034] Specifically, after obtaining the expert model integration scheme, it is sent to the gating network of the hybrid expert model to guide the gating network to perform weighted fusion on the outputs of each expert model, and give the corresponding predicted fault component results for the input electromechanical equipment operating state matrix to obtain the predicted fault components of the electromechanical equipment.

[0035] After receiving the expert model integration solution, the gating network extracts the integrated expert model subset and the corresponding weight coefficient from it. The former determines which expert model outputs need to be included in the final integrated prediction, while the latter provides the importance of each expert model output. Subsequently, the gating network simultaneously inputs the electromechanical equipment operation state matrix into each expert model in the integrated expert model subset to obtain multiple parallel prediction output results. Although these results all come from the same set of equipment status data, due to the differences in the fault modes and feature dimensions that each expert model is good at, the judgment of possible faulty components may not be the same. Afterwards, the gating network comprehensively processes each prediction output result based on the weight coefficient provided by the expert model integration solution to obtain the predicted faulty components of the electromechanical equipment.

[0036] Through the hybrid expert model, the parallel prediction results of the expert model layer are integrated, taking into account different fault modes and feature dimensions, which improves the comprehensiveness of the prediction. At the same time, the weight coefficient is used to fuse the expert model output to improve the accuracy of the integrated prediction.

[0037] Furthermore, the embodiment of the present application also includes:

[0038] S310: collecting fault samples based on the equipment model identification, the equipment component service life identification list and the equipment component maintenance times identification list as constraints, collecting a priori data set, wherein any piece of priori data includes a priori sample electromechanical equipment operation state matrix and a priori sample electromechanical equipment fault component;

[0039] S320: Obtain a first expert model of the expert model layer, process the prior sample electromechanical equipment operation state matrix, and obtain a first expert model to predict a fault component;

[0040] S330: Counting the ratio of the intersection fault components of the priori sample electromechanical device fault components and the fault components predicted by the first expert model to the union fault components of the priori sample electromechanical device fault components and the fault components predicted by the first expert model, and setting it as the first expert model fault prediction accuracy rate;

[0041] S340: Add the first expert model fault prediction accuracy into the fault prediction accuracy list.

[0042] In a feasible implementation, first, according to the equipment model identification, equipment component service life identification list and equipment component maintenance number identification list of the electromechanical equipment to be predicted, the historical operation data is screened, and the fault sample data similar to the state of the electromechanical equipment to be predicted is collected as a priori data set for the subsequent evaluation of the prediction performance of the expert model. Among them, each data in the priori data set includes two parts: one is the priori sample electromechanical equipment operation state matrix, which records the operation state data of similar equipment at a certain historical moment; the other is the priori sample electromechanical equipment fault component, which records the actual fault component information of the similar equipment at that historical moment.

[0043] Then, traverse each expert model in the expert model layer, select one expert model each time as the first expert model, and use it to process each prior sample electromechanical equipment operating state matrix in the prior data set to obtain the corresponding fault prediction result, that is, the first expert model predicts the fault element. Then, for each data in the prior data set, compare the fault element predicted by the first expert model with the fault element of the prior sample electromechanical equipment, count the number of elements of the intersection of the two, and divide it by the number of elements of the union of the two to obtain the prediction accuracy of the first expert model on the prior data. Take the average of the prediction accuracy of all data in the prior data set as the fault prediction accuracy of the first expert model on the entire prior data set, that is, the fault prediction accuracy of the first expert model, which represents the prediction performance of the first expert model when facing a state similar to that of the electromechanical equipment to be predicted. Afterwards, the fault prediction accuracy of the first expert model is recorded in the fault prediction accuracy list as a reference for screening the integrated expert model in the subsequent steps.

[0044] By collecting prior data sets and traversing the expert model layers, a quantitative evaluation of the prediction performance of each expert model is achieved, providing a basis for subsequent expert model screening and integration.

[0045] Furthermore, the embodiment of the present application also includes:

[0046] S311: Obtain first fault sample data, wherein the first fault sample data has a fault equipment model label, a service life label list of equipment components, and a maintenance number label list of equipment components;

[0047] S312: Calculating the absolute value of service life deviations of components of the same model on the equipment component service life label list and the equipment component service life identification list to obtain an equipment component service life deviation absolute value list;

[0048] S313: Calculating the absolute value of the deviation of the maintenance times of components of the same model on the equipment component maintenance times label list and the equipment component maintenance times identification list to obtain an absolute value list of the deviation of the maintenance times of components of the same model;

[0049] S314: Counting the proportion of the number of components in the equipment component service life deviation absolute value list whose service life deviation absolute value is less than the service life deviation absolute value threshold, and setting it as the service status similarity;

[0050] S315: Counting the proportion of the number of components in the equipment component maintenance frequency deviation absolute value list whose maintenance frequency deviation absolute value is less than the maintenance frequency deviation absolute value threshold, and setting it as the operation and maintenance status similarity;

[0051] S316: When the faulty equipment model label is the same as the equipment model identifier, and the service status similarity is greater than or equal to a first similarity threshold, and the operation and maintenance status similarity is greater than or equal to a second similarity threshold, the first fault sample data is added to the prior data set.

[0052] In a preferred embodiment, when obtaining a priori data sets, first, each fault sample data in the historical operation data is traversed as the first fault sample data. The first fault sample data contains the attribute information of the electromechanical equipment that failed at a certain historical moment, including the faulty equipment model label, the equipment component service life label list, and the equipment component maintenance times label list. Among them, the faulty equipment model label indicates the model of the equipment; the equipment component service life label list records the service life of each component of the equipment when the failure occurs; the equipment component maintenance times label list records the historical maintenance times of each component of the equipment before the failure occurs.

[0053] Then, the service life label list of the equipment components of the first fault sample data and the service life identification list of the equipment components of the electromechanical equipment to be predicted are used to calculate the absolute value of the service life deviation of the two on the same type of components, and obtain the absolute value list of the service life deviation of the equipment components, which quantifies the difference in the service level of the first fault sample equipment and the electromechanical equipment to be predicted in the same type of components. At the same time, the equipment component maintenance times label list of the first fault sample data and the equipment component maintenance times identification list of the electromechanical equipment to be predicted are used to calculate the absolute value of the maintenance times deviation of the two on the same type of components, and obtain the absolute value list of the maintenance times deviation of the equipment components, which quantifies the difference in the reliability level of the first fault sample equipment and the electromechanical equipment to be predicted in the same type of components. Subsequently, the number of absolute values ​​of deviations less than the preset threshold in the equipment component service life deviation absolute value list is counted and divided by the total length of the list to obtain the service status similarity, which reflects the overall similarity between the first fault sample equipment and the electromechanical equipment to be predicted in the service level, and the value is between 0 and 1. At the same time, the number of absolute values ​​of deviations in the absolute value list of equipment component maintenance frequency deviations that are less than the preset threshold is counted and divided by the total length of the list to obtain the similarity of operation and maintenance status, which reflects the overall similarity between the reliability level of the first fault sample equipment and the electromechanical equipment to be predicted, and the value is between 0 and 1.

[0054] Subsequently, determine whether the faulty device model label of the first fault sample data is consistent with the device model identifier of the electromechanical device to be predicted, and whether the service status similarity is greater than or equal to the first similarity threshold, and the operation and maintenance status similarity is greater than or equal to the second similarity threshold. Among them, the first similarity threshold refers to the lower limit of the service status similarity; the second similarity threshold refers to the lower limit of the operation and maintenance status similarity. If all three conditions are met at the same time, it is considered that the state of the first fault sample device is similar to that of the electromechanical device to be predicted, and the first fault sample data is added to the priori data set; otherwise, the first fault sample data is discarded and not included in the priori data set. Repeat the execution until all fault sample data are traversed. After the traversal is completed, the priori data set contains all fault samples that are similar to the state of the electromechanical equipment to be predicted.

[0055] By comparing the similarity between the faulty sample equipment and the electromechanical equipment to be predicted in terms of equipment model, service status, and operation and maintenance status, a highly targeted prior data set collection is achieved. This collection method makes full use of the structured attribute information of electromechanical equipment, can effectively filter out irrelevant samples with large status differences, improve the representativeness of the prior data set, and provide high-quality data support for subsequent expert model evaluation.

[0056] Furthermore, the embodiment of the present application also includes:

[0057] S410: Based on the fault prediction accuracy list, extracting a target expert model set and a fault prediction accuracy set whose fault prediction accuracy is greater than or equal to a fault prediction accuracy threshold from the expert model layer;

[0058] S420: Performing weighted full connection on the target expert model set according to the fault prediction accuracy set to obtain the expert model integration solution.

[0059] In a preferred embodiment, when obtaining an expert model integration scheme, first, based on the obtained fault prediction accuracy list, expert models with excellent prediction performance are screened out from the expert model layer as candidates for integration. Specifically, a fault prediction accuracy threshold is set, each item in the fault prediction accuracy list is traversed, and expert models with fault prediction accuracy greater than or equal to the threshold are extracted to form a target expert model set, and the corresponding fault prediction accuracy values ​​are extracted to form a fault prediction accuracy set. The size of the fault prediction accuracy threshold determines the number and quality of the integrated expert models, which is determined comprehensively based on factors such as specific application requirements and computing resource limitations. The larger the threshold, the fewer the number of integrated expert models, the higher the quality, the better the prediction accuracy may be, but the diversity and robustness of the integration may be reduced; conversely, the larger the integration scale, the better the diversity, but the computational overhead will also increase.

[0060] Subsequently, the target expert model set is weighted and combined using the fault prediction accuracy set to generate an expert model integration scheme. This scheme not only clarifies the weight of each expert model in the integrated prediction, but also determines the connection method between the expert models. Specifically, a weighted fully connected network structure is adopted to fully connect each expert model in the target expert model set with the input layer of the gating network, and the connection weight is the corresponding fault prediction accuracy. In this way, the expert model with higher prediction accuracy has a greater weight in the gating network, and the greater the impact on the final integrated prediction result. At the same time, due to the use of a fully connected structure, the prediction results of different expert models can fully interact, which is conducive to exploring the complementarity and synergy between expert models and improving the performance of integrated prediction.

[0061] Through the selection and organization of integrated expert models, a weighted integration scheme oriented towards fault prediction accuracy is formed, which makes full use of the performance information of the evaluated individual expert models. Through two-stage screening and integration, the prediction ability of the integrated expert model is guaranteed, while taking into account the limitations of the integration scale on computing resources, thus realizing adaptive expert model integration.

[0062] Furthermore, the present application is an embodiment and also includes:

[0063] S421: summing up the fault prediction accuracy set to generate a fault prediction accuracy sum result;

[0064] S422: traverse the fault prediction accuracy set, compare them with the sum of the fault prediction accuracy respectively, and set them as the expert model integration weight;

[0065] S423: Perform weighted full connection on the target expert model set according to the expert model integration weight to obtain the expert model integration solution.

[0066] In a preferred embodiment, first, all elements in the fault prediction accuracy set are summed to obtain the fault prediction accuracy sum result. This result reflects the sum of the prediction accuracy of all expert models in the target expert model set, and serves as the reference value for the subsequent calculation of the integration weight of each expert model. Then, each element in the fault prediction accuracy set is traversed, and it is divided by the obtained fault prediction accuracy sum result to obtain the corresponding expert model integration weight, so as to realize the conversion of the fault prediction accuracy of each expert model into a weight coefficient, and the sum of all weights is 1. Then, the target expert model set is weighted fully connected using the obtained expert model integration weight to form an expert model integration scheme. Specifically, each expert model in the target expert model set is fully connected to the input layer of the gated network, and the connection weight is the corresponding expert model integration weight. The larger the weight of the connection, the greater the impact of the output of the corresponding expert model on the gated network, and the more critical its role in the final integrated prediction. At the same time, due to the adoption of the weighted full connection mechanism, the prediction results of different expert models can be adaptively integrated, thereby maximizing the strengths of each expert model, making up for the shortcomings of a single model, and improving the prediction performance of the entire integrated model.

[0067] By summing up the fault prediction accuracy, calculating the weights, and weighted full connection, an efficient expert model integration scheme generation mechanism is realized, which makes the integrated expert model more optimized in terms of organizational form and performance, and provides strong support for achieving high-accuracy and strong robustness fault prediction of electromechanical equipment.

[0068] Furthermore, the embodiment of the present application also includes:

[0069] S411: Obtaining an integrated model prediction log of the device model identification, wherein the integrated model prediction log includes a model integration quantity record value and an output accuracy record value;

[0070] S412: Counting the mode value of the model integration quantity record value whose output accuracy record value is greater than or equal to the output accuracy threshold, and setting it as the recommended model integration quantity;

[0071] S413: extracting the target expert model set and the fault prediction accuracy set whose fault prediction accuracy is greater than or equal to a fault prediction accuracy threshold from the expert model layer according to the recommended model integration quantity.

[0072] In a preferred embodiment, when obtaining the target expert model set and the fault prediction accuracy set, in order to minimize the scale of the integrated expert model and reduce unnecessary computing resource consumption while meeting the fault prediction accuracy requirements, first, obtain the integrated model prediction log corresponding to the equipment model identification of the electromechanical equipment to be predicted. This log is a statistical record of the prediction effect when different integrated model scales are used in historical prediction tasks, and contains model integration quantity record values ​​and output accuracy record values. Among them, the model integration quantity record value represents the number of integrated expert models, and the output accuracy record value represents the fault prediction accuracy on the verification data set when the corresponding number of integrated expert models is used. By analyzing the historical prediction logs, the relationship between different integration scales and prediction performance is revealed, providing a reference for optimizing the scale of the integrated expert model for the current prediction task.

[0073] Subsequently, all records that meet the prediction performance requirements are screened out from the integrated model prediction log, that is, records whose output accuracy record values ​​are greater than or equal to the output accuracy threshold, and the corresponding model integration quantity record values ​​are extracted. The mode of these record values ​​is counted as the recommended model integration quantity. The mode reflects the number of integrated expert models most commonly used in historical prediction tasks under the premise of meeting the prediction performance requirements, and represents the optimal balance point, that is, obtaining the best prediction effect with the smallest integration scale. Using it as the number of integrated expert models for the current prediction task can avoid the waste of computing resources caused by blindly increasing the integration scale, and at the same time ensure that the prediction performance will not be significantly reduced.

[0074] Subsequently, taking the number of recommended model integrations obtained as a benchmark, an equal number of candidate expert models are extracted from the expert model layer as the target expert model set, and the corresponding fault prediction accuracy is extracted as the fault prediction accuracy set. Specifically, the first N expert models are selected in descending order of fault prediction accuracy, where N is equal to the number of recommended model integrations. In this way, it is ensured that the integrated expert models have the best prediction performance, and the integrated scale is controlled not to expand excessively, taking into account both the accuracy and efficiency of the integrated prediction.

[0075] By introducing the analysis mechanism of historical integrated model prediction logs, adaptive integrated expert model scale optimization is achieved, which fully utilizes the existing prediction task knowledge and significantly reduces the redundancy of the integrated expert model. While ensuring the prediction performance, the consumption of computing resources is minimized, thus improving the practicality and economy of fault prediction.

[0076] Furthermore, the embodiment of the present application also includes:

[0077] S4131: Obtaining a set of first-level target expert models in the expert model layer whose fault prediction accuracy is greater than or equal to a fault prediction accuracy threshold;

[0078] S4132: When the number of the first-level target expert model sets is less than the number of the recommended model integration, the expert model with the maximum fault prediction accuracy of the first-level target expert model set is twin-copied to a number equal to the number of the recommended model integration to obtain the target expert model set and the fault prediction accuracy set.

[0079] In a preferred embodiment, in actual applications, the number of candidate expert models that meet the fault prediction accuracy threshold requirements may be less than the number of recommended model integrations. In order to ensure that the scale of the integrated expert model is consistent with the number of recommended model integrations. First, all expert models whose fault prediction accuracy is greater than or equal to the fault prediction accuracy threshold are selected from the expert model layer to form a first-level target expert model set. This set contains the currently available optimal expert models and is the preferred object for building an integrated model. However, since the number of first-level target expert model sets may be smaller than the number of recommended model integrations, it may be difficult to give full play to the advantages of integrated learning by directly using this set for integration, so it is necessary to expand on this basis.

[0080] Subsequently, determine whether the number of the first-level target expert model set is less than the recommended model integration number. If so, select the expert model with the highest fault prediction accuracy from the first-level target expert model set and perform twin replication on it until the number of expanded expert model sets is equal to the recommended model integration number. The so-called twin replication refers to a complete replication of a new model instance with the same structure and parameters as the original expert model, which has the same prediction ability and accuracy as the original model. Through twin replication, while maintaining the existing optimal model in the first-level target expert model set, the gap in the integration scale is filled, thereby meeting the recommended integration quantity requirements and not introducing expert models with poor prediction capabilities, ensuring the lower limit of the overall performance of the integrated model. After the twin replication is completed, the expanded expert model set is used as the final target expert model set, and its corresponding fault prediction accuracy is extracted as the fault prediction accuracy set to prepare for the subsequent generation of expert model integration solutions.

[0081] By providing an adaptive expansion mechanism to deal with the insufficient number of candidate expert models, it can flexibly cope with the situation of insufficient number of expert models in practical applications, and has stronger robustness and practicality.

[0082] The electromechanical equipment fault prediction method integrating the hybrid expert model provided by the embodiment of the present invention has at least the following technical effects:

[0083] The operating state matrix of electromechanical equipment is received, wherein the operating state matrix of electromechanical equipment has equipment model identification, equipment component service life identification list and equipment component maintenance number identification list, providing the necessary data basis for subsequent expert model matching, etc. According to the equipment model identification, the hybrid expert model is matched, wherein the hybrid expert model has an expert model layer and a gating network. For different types of equipment, a hybrid expert model suitable for it is selected, and the equipment characteristics are fully considered to improve the pertinence and prediction effect of the model. At the same time, the expert model layer and the gating network lay the architectural foundation for subsequent integrated prediction. According to the equipment model identification, the equipment component service life identification list and the equipment component maintenance number identification list, the expert model layer is traversed for priori to obtain a list of fault prediction accuracy, and the existing fault cases and expert experience are fully utilized to evaluate the prediction ability of each expert model, and the quantitative results of the fault prediction accuracy of different expert models are obtained, which provides a decision-making basis for the subsequent expert model integration. According to the fault prediction accuracy list, the expert model integration configuration is performed to obtain the expert model integration solution, which maximizes the advantages of different expert models, makes up for the shortcomings of a single model, and significantly improves the overall effect of fault prediction. The expert model integration scheme is sent to the gated network to process the operating state matrix of the electromechanical equipment, obtain the predicted fault components of the electromechanical equipment, and improve the accuracy of electromechanical equipment fault prediction.

[0084] Embodiment 2:

[0085] like Figure 2 As shown, based on the same inventive concept of the electromechanical equipment fault prediction method integrating the hybrid expert model provided in the first embodiment, the embodiment of the present invention also provides an electromechanical equipment fault prediction system integrating the hybrid expert model, including:

[0086] The data receiving module 11 is used to receive the electromechanical equipment operation status matrix, wherein the electromechanical equipment operation status matrix has an equipment model identification, an equipment component service life identification list, and an equipment component maintenance number identification list;

[0087] A model matching module 12, used for matching a hybrid expert model according to the device model identifier, wherein the hybrid expert model has an expert model layer and a gating network;

[0088] The accuracy evaluation module 13 is used to traverse the expert model layer for priori according to the equipment model identification, the service life identification list of the equipment components and the maintenance number identification list of the equipment components to obtain a fault prediction accuracy list;

[0089] An integration configuration module 14 is used to perform expert model integration configuration according to the fault prediction accuracy list to obtain an expert model integration solution;

[0090] The fault prediction module 15 is used to send the expert model integration solution to the gate control network, process the electromechanical equipment operation state matrix, and obtain the predicted fault components of the electromechanical equipment.

[0091] Furthermore, the accuracy evaluation module 13 includes the following execution steps:

[0092] Fault samples are collected based on the equipment model identification, the equipment component service life identification list and the equipment component maintenance times identification list as constraints, and a priori data set is collected, wherein any piece of priori data includes a priori sample electromechanical equipment operation state matrix and a priori sample electromechanical equipment fault component;

[0093] Obtaining a first expert model of the expert model layer, processing the prior sample electromechanical equipment operation state matrix, and obtaining a first expert model to predict a faulty component;

[0094] Counting the ratio of the intersection fault components of the priori sample electromechanical device fault components and the fault components predicted by the first expert model to the union fault components of the priori sample electromechanical device fault components and the fault components predicted by the first expert model, and setting it as the first expert model fault prediction accuracy rate;

[0095] The first expert model fault prediction accuracy is added to the fault prediction accuracy list.

[0096] Furthermore, the accuracy evaluation module 13 also includes the following execution steps:

[0097] Obtaining first fault sample data, wherein the first fault sample data has a fault equipment model label, a service life label list of equipment components, and a maintenance number label list of equipment components;

[0098] Calculate the absolute value of service life deviation of the same type of components on the equipment component service life label list and the equipment component service life identification list to obtain an equipment component service life deviation absolute value list;

[0099] Calculate the absolute value of the deviation of the maintenance times of components of the same model on the equipment component maintenance times label list and the equipment component maintenance times identification list to obtain an absolute value list of the deviation of the maintenance times of equipment components;

[0100] Counting the proportion of the number of components in the equipment component service life deviation absolute value list whose service life deviation absolute value is less than the service life deviation absolute value threshold, and setting it as the service status similarity;

[0101] Counting the proportion of components in the equipment component maintenance frequency deviation absolute value list whose maintenance frequency deviation absolute value is less than the maintenance frequency deviation absolute value threshold, and setting it as the operation and maintenance status similarity;

[0102] When the faulty equipment model label is the same as the equipment model identifier, and the service status similarity is greater than or equal to a first similarity threshold, and the operation and maintenance status similarity is greater than or equal to a second similarity threshold, the first fault sample data is added to the prior data set.

[0103] Furthermore, the integrated configuration module 14 includes the following execution steps:

[0104] Based on the fault prediction accuracy list, extracting a target expert model set and a fault prediction accuracy set whose fault prediction accuracy is greater than or equal to a fault prediction accuracy threshold from the expert model layer;

[0105] The target expert model set is weighted fully connected according to the fault prediction accuracy set to obtain the expert model integration solution.

[0106] Furthermore, the integrated configuration module 14 also includes the following execution steps:

[0107] Adding the fault prediction accuracy set to generate a fault prediction accuracy sum result;

[0108] Traversing the fault prediction accuracy set, respectively comparing the result with the sum of the fault prediction accuracy, and setting it as the expert model integration weight;

[0109] The target expert model set is weighted fully connected according to the expert model integration weight to obtain the expert model integration solution.

[0110] Furthermore, the integrated configuration module 14 also includes the following execution steps:

[0111] Obtaining an integrated model prediction log of the device model identification, wherein the integrated model prediction log includes a model integration quantity record value and an output accuracy record value;

[0112] Counting the mode value of the model integration quantity record value whose output accuracy record value is greater than or equal to the output accuracy threshold, and setting it as the recommended model integration quantity;

[0113] The target expert model set and the fault prediction accuracy set whose fault prediction accuracy is greater than or equal to a fault prediction accuracy threshold are extracted from the expert model layer according to the recommended model integration quantity.

[0114] Furthermore, the integrated configuration module 14 also includes the following execution steps:

[0115] Obtaining a set of first-level target expert models in the expert model layer whose fault prediction accuracy is greater than or equal to a fault prediction accuracy threshold;

[0116] When the number of the first-level target expert model sets is less than the number of the recommended model integration, the expert model with the maximum fault prediction accuracy of the first-level target expert model set is twin-copied to a number equal to the number of the recommended model integration to obtain the target expert model set and the fault prediction accuracy set.

[0117] Embodiment three:

[0118] The embodiment of the present invention provides a fault prediction intelligent operation and maintenance gateway for electromechanical equipment embodiment integrating a hybrid expert model, which is used to implement a fault prediction method for electromechanical equipment integrating a hybrid expert model, including:

[0119] Receiving an electromechanical equipment operation status matrix, wherein the electromechanical equipment operation status matrix has an equipment model identification, an equipment component service life identification list, and an equipment component maintenance number identification list;

[0120] Matching a hybrid expert model according to the device model identifier, wherein the hybrid expert model has an expert model layer and a gating network;

[0121] According to the equipment model identification, the service life identification list of the equipment components and the maintenance number identification list of the equipment components, the expert model layer is traversed for priori to obtain a list of fault prediction accuracy rates;

[0122] Perform expert model integration configuration according to the fault prediction accuracy list to obtain an expert model integration solution;

[0123] The expert model integration solution is sent to the gate control network, the electromechanical equipment operation state matrix is ​​processed, and the predicted fault components of the electromechanical equipment are obtained.

[0124] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0125] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention 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 code.

[0126] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of 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 computer, or other programmable data processing device to produce 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 system that specifies the functions of a box or multiple boxes.

[0127] 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.

[0128] 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.

[0129] Although preferred embodiments of the present invention have been described, additional changes and modifications may occur to these embodiments once those skilled in the art understand the basic inventive concepts.

[0130] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention belong to the scope of the present invention and its equivalent technology, the present invention is also intended to include these changes and variations.

Claims

1. A method for predicting mechanical and electrical equipment faults by integrating a hybrid expert model, characterized in that: include: Receiving an electromechanical equipment operation status matrix, wherein the electromechanical equipment operation status matrix has an equipment model identification, an equipment component service life identification list, and an equipment component maintenance number identification list; Matching a hybrid expert model according to the device model identifier, wherein the hybrid expert model has an expert model layer and a gating network; According to the equipment model identification, the service life identification list of the equipment components and the maintenance number identification list of the equipment components, the expert model layer is traversed for priori to obtain a list of fault prediction accuracy rates; Perform expert model integration configuration according to the fault prediction accuracy list to obtain an expert model integration solution; Sending the expert model integration solution to the gated network, processing the electromechanical equipment operation state matrix, and obtaining the predicted fault components of the electromechanical equipment; According to the equipment model identification, the equipment component service life identification list and the equipment component maintenance number identification list, the expert model layer is traversed for priori to obtain a list of fault prediction accuracy, including: Fault samples are collected based on the equipment model identification, the equipment component service life identification list and the equipment component maintenance times identification list as constraints, and a priori data set is collected, wherein any piece of priori data includes a priori sample electromechanical equipment operation state matrix and a priori sample electromechanical equipment fault component; Obtaining a first expert model of the expert model layer, processing the prior sample electromechanical equipment operation state matrix, and obtaining a first expert model to predict a faulty component; Counting the ratio of the intersection fault components of the priori sample electromechanical device fault components and the fault components predicted by the first expert model to the union fault components of the priori sample electromechanical device fault components and the fault components predicted by the first expert model, and setting it as the first expert model fault prediction accuracy rate; Adding the first expert model fault prediction accuracy rate to the fault prediction accuracy rate list; Among them, the expert model integration configuration is performed according to the fault prediction accuracy list to obtain the expert model integration solution, including: Based on the fault prediction accuracy list, extracting a target expert model set and a fault prediction accuracy set whose fault prediction accuracy is greater than or equal to a fault prediction accuracy threshold from the expert model layer; The target expert model set is weighted fully connected according to the fault prediction accuracy set to obtain the expert model integration solution.

2. The method according to claim 1, characterized in that Fault samples are collected based on the equipment model identification, the equipment component service life identification list, and the equipment component maintenance times identification list as constraints, and a priori data sets are collected, including: Obtaining first fault sample data, wherein the first fault sample data has a fault equipment model label, a service life label list of equipment components, and a maintenance number label list of equipment components; Calculate the absolute value of service life deviation of the same type of components on the equipment component service life label list and the equipment component service life identification list to obtain an equipment component service life deviation absolute value list; Calculate the absolute value of the deviation of the maintenance times of components of the same model on the equipment component maintenance times label list and the equipment component maintenance times identification list to obtain an absolute value list of the deviation of the maintenance times of equipment components; Counting the proportion of the number of components in the equipment component service life deviation absolute value list whose service life deviation absolute value is less than the service life deviation absolute value threshold, and setting it as the service status similarity; Counting the proportion of components in the equipment component maintenance frequency deviation absolute value list whose maintenance frequency deviation absolute value is less than the maintenance frequency deviation absolute value threshold, and setting it as the operation and maintenance status similarity; When the faulty equipment model label is the same as the equipment model identifier, and the service status similarity is greater than or equal to a first similarity threshold, and the operation and maintenance status similarity is greater than or equal to a second similarity threshold, the first fault sample data is added to the prior data set.

3. The method according to claim 1, characterized in that The target expert model set is weighted fully connected according to the fault prediction accuracy set to obtain the expert model integration solution, including: Adding the fault prediction accuracy set to generate a fault prediction accuracy sum result; Traversing the fault prediction accuracy set, respectively comparing the result with the sum of the fault prediction accuracy, and setting it as the expert model integration weight; The target expert model set is weighted fully connected according to the expert model integration weight to obtain the expert model integration solution.

4. The method according to claim 3, characterized in that Extracting a target expert model set and a fault prediction accuracy set whose fault prediction accuracy is greater than or equal to a fault prediction accuracy threshold from the expert model layer, further comprising: Obtaining an integrated model prediction log of the device model identification, wherein the integrated model prediction log includes a model integration quantity record value and an output accuracy record value; Counting the mode value of the model integration quantity record value whose output accuracy record value is greater than or equal to the output accuracy threshold, and setting it as the recommended model integration quantity; The target expert model set and the fault prediction accuracy set whose fault prediction accuracy is greater than or equal to a fault prediction accuracy threshold are extracted from the expert model layer according to the recommended model integration quantity.

5. The method according to claim 4, characterized in that The target expert model set and the fault prediction accuracy set whose fault prediction accuracy is greater than or equal to the fault prediction accuracy threshold are extracted from the expert model layer according to the recommended model integration quantity, and further comprising: Obtaining a set of first-level target expert models in the expert model layer whose fault prediction accuracy is greater than or equal to a fault prediction accuracy threshold; When the number of the first-level target expert model sets is less than the number of the recommended model integration, the expert model with the maximum fault prediction accuracy of the first-level target expert model set is twin-copied to a number equal to the number of the recommended model integration to obtain the target expert model set and the fault prediction accuracy set.

6. The electromechanical equipment fault prediction system integrating hybrid expert model is characterized by: The method for predicting electromechanical equipment faults using the fused hybrid expert model according to any one of claims 1 to 5 comprises: A data receiving module, the data receiving module is used to receive an electromechanical equipment operation status matrix, wherein the electromechanical equipment operation status matrix has an equipment model identification, an equipment component service life identification list, and an equipment component maintenance number identification list; A model matching module, the model matching module is used to match a hybrid expert model according to the device model identifier, wherein the hybrid expert model has an expert model layer and a gating network; An accuracy evaluation module, the accuracy evaluation module is used to traverse the expert model layer for priori according to the equipment model identification, the service life identification list of the equipment components and the maintenance number identification list of the equipment components to obtain a fault prediction accuracy list; An integrated configuration module, the integrated configuration module is used to perform expert model integration configuration according to the fault prediction accuracy list to obtain an expert model integration solution; A fault prediction module is used to send the expert model integration solution to the gate control network, process the electromechanical equipment operation state matrix, and obtain the predicted fault element of the electromechanical equipment.

7. The intelligent operation and maintenance gateway for mechanical and electrical equipment fault prediction based on hybrid expert model is characterized by: A method for predicting electromechanical equipment faults using a fused hybrid expert model as described in any one of claims 1 to 5.

Images