Mechanical and electrical equipment fault monitoring method, device and storage medium combined with adjacent sample analysis

Abstract

The present invention relates to a method, device and storage medium for monitoring electromechanical equipment faults in combination with adjacent sample analysis, and relates to the field of control equipment monitoring, including: interacting with electromechanical equipment, receiving electromechanical equipment operation control parameters and electromechanical equipment monitoring feature matrix; performing multi-level adjacent search in non-fault state to obtain adjacent sample tree; recursively aggregating adjacent sample tree to generate electromechanical equipment standard feature matrix; comparing electromechanical equipment monitoring feature matrix and electromechanical equipment standard feature matrix, constructing electromechanical equipment feature deviation vector matrix, inputting electromechanical equipment fault probability prediction model, generating electromechanical equipment fault probability, greater than or equal to fault probability threshold, and generating electromechanical equipment maintenance instructions. By introducing adjacent sample analysis, the sample size of fault monitoring is expanded, while ensuring the accuracy of judgment, realizing the automation and intelligence of electromechanical equipment fault monitoring, and achieving the technical effect of improving fault monitoring efficiency and accuracy.

Description

Technical Field

[0001] The present invention relates to the field of control equipment monitoring, and in particular to a method, equipment and storage medium for monitoring electromechanical equipment faults in combination with adjacent sample analysis. Background Art

[0002] As a key facility in industrial production, the operating status of electromechanical equipment affects production efficiency and product quality. In order to ensure the normal operation of electromechanical equipment, it is crucial to timely detect and warn of equipment failures. Traditional electromechanical equipment fault monitoring mainly relies on manual experience, and determines whether the equipment is faulty by setting threshold values ​​of various state parameters. This method is not only inefficient, but also its accuracy depends largely on the professional level of personnel, and it is difficult to adapt to the increasingly complex needs of industrial production. In recent years, with the development of industrial automation and informatization, some studies have begun to try to use data-driven methods to realize automatic monitoring and diagnosis of electromechanical equipment faults. These methods automatically construct threshold values ​​of various state parameters by collecting and analyzing the massive data generated during the operation of the equipment, but due to the limited sample size of industrial data, the accuracy and reliability are difficult to meet the requirements of practical applications. Summary of the invention

[0003] The present invention aims to solve the technical problem that in the prior art, electromechanical equipment fault monitoring is carried out by manually setting various state threshold values, or the automatic setting is insufficient due to insufficient industrial data samples, resulting in low efficiency and accuracy of automatic fault monitoring. An electromechanical equipment fault monitoring method, device and storage medium combined with adjacent sample analysis are provided 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 monitoring electromechanical equipment faults in combination with adjacent sample analysis, including: interacting with electromechanical equipment to receive electromechanical equipment operation control parameters and an electromechanical equipment monitoring feature matrix; performing a multi-level adjacent search for a non-fault state according to the electromechanical equipment operation control parameters to obtain an adjacent sample tree; recursively aggregating the adjacent sample tree to generate an electromechanical equipment standard feature matrix; comparing the electromechanical equipment monitoring feature matrix and the electromechanical equipment standard feature matrix to construct an electromechanical equipment feature deviation vector matrix; inputting the electromechanical equipment feature deviation vector matrix into an electromechanical equipment failure probability prediction model to generate an electromechanical equipment failure probability; when the electromechanical equipment failure probability is greater than or equal to a failure probability threshold, generating an electromechanical equipment maintenance instruction and sending it to the electromechanical equipment operation and maintenance management end.

[0006] In a second aspect, the present application provides an electronic device, comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is used to execute the electromechanical equipment fault monitoring method combined with adjacent sample analysis provided in the present application.

[0007] In a third aspect, the present application provides a non-transitory computer-readable storage medium storing a computer program for executing the electromechanical equipment fault monitoring method combined with adjacent sample analysis provided by the present application.

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

[0009] It interacts with electromechanical equipment, receives electromechanical equipment operation control parameters and electromechanical equipment monitoring feature matrix, and provides a basis for subsequent fault monitoring and analysis; it performs multi-level adjacency search according to the electromechanical equipment operation control parameters to obtain the adjacency sample tree, expands the number of samples, and lays the foundation for the subsequent construction of the standard feature matrix; it recursively aggregates the adjacency sample tree to generate the electromechanical equipment standard feature matrix, and extracts the standard feature performance of the electromechanical equipment under normal operating conditions by summarizing the adjacency sample tree from the bottom up to form a standard feature matrix; it compares the electromechanical equipment monitoring feature matrix and the electromechanical equipment standard feature matrix to construct the electromechanical equipment feature deviation vector matrix, quantitatively characterizes the degree to which the actual operating state of the electromechanical equipment deviates from the normal state, and forms a deviation vector matrix that characterizes the degree of fault; it inputs the electromechanical equipment feature deviation vector matrix into the electromechanical equipment fault probability prediction model to generate the electromechanical equipment fault probability and realize forward-looking warning of faults; when the electromechanical equipment fault probability is greater than or equal to the fault probability threshold, it generates the electromechanical equipment maintenance instruction and sends it to the electromechanical equipment operation and maintenance management end to eliminate the fault in an early state and ensure the continuous normal operation of the equipment. By utilizing adjacent sample analysis technology, the problem of insufficient industrial data samples was overcome. By constructing a deviation vector by comparing it with standard features and combining it with a fault probability prediction model, the automated and intelligent monitoring and early warning of electromechanical equipment failures were achieved, improving the efficiency and accuracy of fault monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A schematic flow chart of a method for monitoring electromechanical equipment faults in combination with adjacent sample analysis provided by the present invention;

[0011] Figure 2 A schematic diagram of the structure of an electronic device provided by the present invention;

[0012] Figure 3 A schematic diagram of the structure of a computer-readable storage medium provided by the present invention.

[0013] In the accompanying drawings, the components represented by the reference numerals are as follows:

[0014] Electronic device 100 , memory 110 , processor 120 , first computer program 111 , computer-readable storage medium 200 , second computer program 211 . DETAILED DESCRIPTION

[0015] 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.

[0016] 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.

[0017] 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.

[0018] Embodiment 1:

[0019] like Figure 1 As shown, an embodiment of the present invention provides a method for monitoring a fault of an electromechanical device in combination with adjacent sample analysis, which is applied to a system for monitoring a fault of an electromechanical device in combination with adjacent sample analysis, and the system is in communication connection with the electromechanical device.

[0020] Specifically, the electromechanical equipment fault monitoring method combined with adjacent sample analysis is applied to the electromechanical equipment fault monitoring system combined with adjacent sample analysis. The system establishes a communication connection with the control unit, sensor, etc. of the electromechanical equipment to be monitored through wired or wireless communication to achieve data interaction and transmission. Through the communication connection between the system and the electromechanical equipment, the real-time operation data of the electromechanical equipment can be continuously obtained, providing a data basis for subsequent adjacent sample analysis, fault monitoring and other steps.

[0021] Mechanical and electrical equipment fault monitoring methods include:

[0022] S100: interacting with electromechanical equipment, receiving electromechanical equipment operation control parameters and electromechanical equipment monitoring characteristic matrix.

[0023] Specifically, during the operation of electromechanical equipment, its control unit will collect and record various operation control parameters in real time, such as voltage, current, speed, etc., to form the operation control parameters of the electromechanical equipment. These parameters reflect the real-time working conditions and status of the electromechanical equipment. At the same time, various sensors equipped on the electromechanical equipment, such as vibration sensors and acoustic sensors, continuously collect monitoring data such as vibration and noise of the equipment, and form the electromechanical equipment monitoring feature matrix after feature extraction.

[0024] The electromechanical equipment fault monitoring system regularly sends data requests to the electromechanical equipment through communication connections with the electromechanical equipment, obtains the electromechanical equipment operation control parameters and the electromechanical equipment monitoring feature matrix, and lays a data foundation for realizing intelligent fault monitoring of electromechanical equipment.

[0025] S200: Perform a non-fault state multi-level adjacency search according to the electromechanical equipment operation control parameters to obtain an adjacency sample tree.

[0026] Specifically, after obtaining the operation control parameters of the electromechanical equipment, it is first assumed that the current electromechanical equipment is in a non-fault state, that is, the equipment is operating normally. Under this premise, a multi-level adjacency search is started. The so-called adjacency search refers to finding data samples similar or adjacent to the current electromechanical equipment operation control parameters in the historical database, corresponding to the electromechanical equipment operation data in the non-fault state. Among them, the multi-level adjacency search is a layer-by-layer in-depth process. First, a first-level adjacency search is performed to find a batch of historical data samples closest to the current electromechanical equipment operation control parameters to form a first-level adjacency sample set; then, based on the first-level adjacency samples, a second-level adjacency search is further performed to find data similar to the first-level adjacency samples to form a second-level adjacency sample set. By analogy, a multi-level adjacency search is performed until the preset search depth or sample quantity requirements are reached. The sample sets at each level obtained by the multi-level adjacency search form a tree structure, namely, an adjacency sample tree. The root node of the adjacency sample tree corresponds to the current electromechanical equipment operation control parameters, and each layer of nodes corresponds to the sample set obtained by the first-level adjacency search. The number of layers of the tree reflects the depth of the adjacency search.

[0027] By obtaining the adjacent sample tree, a wealth of reference data is provided for subsequent fault monitoring and analysis. Through the aggregation analysis of adjacent samples, the standard characteristic performance of electromechanical equipment in a non-fault state can be obtained, thus providing a judgment basis for fault diagnosis. At the same time, multi-level adjacent search expands the number of samples and improves the accuracy of the analysis results.

[0028] S300: recursively aggregate the adjacent sample tree to generate a standard feature matrix of electromechanical equipment.

[0029] Specifically, the adjacent sample tree contains a large number of historical operation data samples, which are distributed on each node of the tree. In order to extract the standard characteristic performance of electromechanical equipment in a non-fault state, an aggregation analysis is performed on the adjacent sample tree.

[0030] The recursive aggregation method is adopted to process the adjacent sample tree layer by layer from bottom to top. First, starting from the bottom leaf node of the adjacent sample tree, the feature matrix of the samples contained in each leaf node is extracted. These feature matrices reflect the actual operating characteristics of electromechanical equipment at various historical moments. Then, the feature matrices of all leaf nodes under the same parent node are aggregated. Among them, the aggregation method can be to take the average value, weighted average, maximum value, minimum value, etc., to extract the common characteristics of the sample set represented by the parent node. The feature matrix obtained by aggregation is called the fusion feature matrix of the parent node. After completing the aggregation of the bottom leaf nodes, the adjacent sample tree is recursively processed upward. For each non-leaf node, all the child nodes under it are first pruned, and then the fusion feature matrices of these child nodes are aggregated again to obtain the fusion feature matrix of the non-leaf node. By analogy, through continuous pruning and feature aggregation, the adjacent sample tree is finally reduced to the root node. The fusion feature matrix contained in the root node is the standard feature matrix of the electromechanical equipment, which reflects the standard feature performance of the electromechanical equipment under normal operating conditions.

[0031] Through recursive layer-by-layer pruning and feature fusion, the scattered sample data is condensed into a standard feature matrix of electromechanical equipment using the hierarchical structure of the adjacent sample tree, which not only reduces the computational complexity but also improves the accuracy of feature extraction. The obtained standard feature matrix of electromechanical equipment provides a reference for subsequent fault diagnosis, thereby realizing intelligent fault monitoring.

[0032] S400: Compare the electromechanical equipment monitoring characteristic matrix with the electromechanical equipment standard characteristic matrix to construct an electromechanical equipment characteristic deviation vector matrix.

[0033] Specifically, first, the electromechanical equipment monitoring feature matrix and the electromechanical equipment standard feature matrix are arranged in the same feature order so that the same position of the two matrices corresponds to the same feature parameter. Then, the difference between the corresponding elements in the two matrices is calculated one by one to obtain a difference matrix. Each element of the difference matrix represents the degree of deviation between the actual monitoring feature and the standard feature in the corresponding parameter. In order to better characterize the direction and magnitude of the deviation, the difference matrix is ​​further converted into an electromechanical equipment feature deviation vector matrix. Specifically, each row of the difference matrix is ​​regarded as a feature deviation vector, whose direction indicates the positive or negative of the deviation, and the difference indicates the magnitude of the deviation, thereby forming an electromechanical equipment feature deviation vector matrix. Each vector in the electromechanical equipment feature deviation vector matrix represents the difference between the actual performance and the standard performance of the electromechanical equipment on a certain feature parameter.

[0034] Through matrix comparison and vector processing, quantitative comparison of real-time monitoring features and standard features is achieved, and a characteristic deviation vector matrix of electromechanical equipment is constructed, which not only retains the direction information of the deviation, but also takes into account the size information of the deviation, providing support for further intelligent fault prediction and diagnostic analysis.

[0035] S500: Inputting the electromechanical equipment characteristic deviation vector matrix into an electromechanical equipment failure probability prediction model to generate an electromechanical equipment failure probability.

[0036] Specifically, the electromechanical equipment failure probability prediction model is an intelligent algorithm model based on big data and machine learning. The model learns the intrinsic relationship between feature deviation and failure probability by training massive historical fault data and the corresponding feature deviation vector matrix. The input of the model is the feature deviation vector matrix of electromechanical equipment, and the output is the failure probability of electromechanical equipment. After obtaining the feature deviation vector matrix of electromechanical equipment, it is input into the electromechanical equipment failure probability prediction model. The model can calculate the probability of failure of the current electromechanical equipment and obtain the failure probability of the electromechanical equipment, reflecting the current health status and failure risk level of the electromechanical equipment. Specifically, first, the electromechanical equipment failure probability prediction model extracts the deviation features from the input electromechanical equipment feature deviation vector matrix and converts it into a mathematical representation that the model can handle; then, the model matches the extracted deviation features with the failure mode stored internally to find the most likely cause of the current deviation; then, the model estimates the probability value of the failure based on the matching results and historical data.

[0037] Through big data-driven machine learning methods, the electromechanical equipment fault probability prediction model can automatically extract the laws and characteristics of fault occurrence from massive historical operation data and apply them to real-time fault prediction. Compared with traditional fault judgment methods based on experience and thresholds, it has a higher level of intelligence and prediction accuracy.

[0038] S600: When the failure probability of the electromechanical equipment is greater than or equal to the failure probability threshold, a maintenance instruction for the electromechanical equipment is generated and sent to the electromechanical equipment operation and maintenance management terminal.

[0039] Specifically, after obtaining the failure probability of the electromechanical equipment, the electromechanical equipment failure monitoring system determines whether the failure probability of the electromechanical equipment is greater than or equal to the preset failure probability threshold. If so, the electromechanical equipment maintenance instruction is automatically generated and sent to the electromechanical equipment operation and maintenance management end, prompting relevant personnel to carry out equipment maintenance and fault handling in a timely manner.

[0040] The failure probability threshold is a critical value pre-set based on factors such as the type, importance, and historical failure data of the electromechanical equipment. When the real-time failure probability of the electromechanical equipment exceeds the threshold, it indicates that the equipment is already in a risky state and necessary maintenance and repair measures must be taken to avoid the occurrence and expansion of the failure. The maintenance instructions include information such as the equipment number, fault type, and fault location to guide maintenance personnel to quickly and accurately carry out fault diagnosis and repair work. In order to ensure the timely communication and execution of the instructions, the generated electromechanical equipment maintenance instructions are automatically sent to the electromechanical equipment operation and maintenance management end. The operation and maintenance management end is usually a centralized information platform that is connected to the electromechanical equipment fault monitoring system and is responsible for coordinating and supervising daily equipment operation and maintenance work.

[0041] By automatically generating and sending maintenance instructions, closed-loop management of electromechanical equipment failures is achieved. Based on the real-time failure probability prediction results, maintenance reminders are proactively sent to the operation and maintenance end, minimizing the response time from failure warning to maintenance processing. Predictive maintenance strategies can avoid unplanned equipment downtime, reduce losses caused by failures, and improve the continuity and reliability of equipment operation.

[0042] Furthermore, the embodiment of the present application also includes:

[0043] S210: Obtain a static tag of an electromechanical device, wherein the static tag of the electromechanical device includes a model of the electromechanical device and a service time of the electromechanical device.

[0044] S220: Perform a non-fault state first-level adjacency search according to the electromechanical equipment model, the service time of the electromechanical equipment and the operation control parameters of the electromechanical equipment to obtain a first-level adjacency sample, wherein the first-level adjacency sample has the electromechanical equipment first-level sample operation control parameters.

[0045] S230: Based on the electromechanical equipment primary sample operation control parameter, traverse the primary adjacent samples to perform a non-fault state secondary adjacent search to obtain a secondary adjacent sample, wherein the secondary adjacent sample has the electromechanical equipment secondary sample operation control parameter.

[0046] S240: Based on the operation control parameters of the electromechanical equipment secondary sample, traverse the secondary adjacent samples to perform a non-fault state tertiary adjacent search to obtain a tertiary adjacent sample.

[0047] S250: Configure the root node to be empty, set the first-level adjacent sample to be a first-level leaf node, set the second-level adjacent sample to be a second-level leaf node until the third-level adjacent sample is set to be a third-level leaf node, and generate the adjacent sample tree.

[0048] In a feasible implementation, first, obtain the static label of the electromechanical equipment, including the model and service time of the electromechanical equipment. Among them, the model of the electromechanical equipment is used to identify the basic information of the equipment, such as the type, specification, and manufacturer; the service time of the electromechanical equipment reflects the length of time the equipment has been running. Then, the model, service time and operation control parameters of the electromechanical equipment are used as retrieval conditions to search for similar sample data in the historical database to obtain the first-level adjacent samples. These first-level adjacent samples come from other electromechanical equipment of the same model, similar service time, and similar operating conditions. Their operation control parameters have a high correlation with the electromechanical equipment to be monitored, which are called the first-level sample operation control parameters of the electromechanical equipment. Then, further carry out the second-level adjacent search and the third-level adjacent search. The second-level adjacent search is based on the model, service time and first-level sample operation control parameters of the electromechanical equipment, and similar samples are found in the historical database to obtain the second-level adjacent samples. The second-level adjacent sample operation control parameters are the second-level sample operation control parameters of the electromechanical equipment. The third-level search is based on the second-level adjacent samples, and takes the model, service time and operation control parameters of the second-level samples of electromechanical equipment as the benchmark, and performs similarity matching again to obtain the third-level adjacent samples. Through multi-level search, adjacent sample data that is highly correlated with the operating status of electromechanical equipment and distributed at different similarity levels is obtained. Subsequently, an empty root node is created, and the first-, second-, and third-level adjacent samples are used as the first-, second-, and third-level leaf nodes of the tree, respectively, and a hierarchical parent-child relationship is established with the root node. In this way, all adjacent samples are organized into a tree with a hierarchical structure, and each layer of nodes represents a subset of samples at a similarity level.

[0049] By obtaining the adjacent sample tree through the three-level adjacent search, sufficient reference data can be obtained while ensuring the sample relevance. The three-level search achieves a good balance in breadth and depth, obtaining historical sample data that is highly relevant to the status of the equipment to be monitored, avoiding the problem of too little data or too low relevance.

[0050] Furthermore, the embodiment of the present application also includes:

[0051] S221: Constructing a first adjacency constraint condition according to the electromechanical device model, and when the sample electromechanical device model is the same as the electromechanical device model, it is considered that the first adjacency constraint condition is satisfied.

[0052] S222: According to the service life of the electromechanical equipment and based on a service life deviation threshold, a second adjacency constraint condition is constructed; when the service life deviation between the sample electromechanical equipment and the service life of the electromechanical equipment is less than or equal to the service life deviation threshold, the second adjacency constraint condition is deemed to be satisfied.

[0053] S223: According to the electromechanical equipment operation control parameters and based on the control parameter attribute deviation threshold set, a third adjacency constraint condition is constructed; when the control parameter deviation set of the sample electromechanical equipment operation control parameters and the electromechanical equipment operation control parameters is less than or equal to the one-to-one corresponding control parameter attribute deviation threshold set, the third adjacency constraint condition is deemed to be satisfied.

[0054] S224: Taking the electromechanical equipment monitoring characteristic attributes as a target, collecting at least 50 of the first-level adjacency samples that simultaneously satisfy the first adjacency constraint condition, the second adjacency constraint condition, and the third adjacency constraint condition.

[0055] In a preferred embodiment, when obtaining the first-level adjacent samples, first, a first adjacent constraint is constructed according to the model of the electromechanical equipment to ensure the consistency of the model between the adjacent sample and the electromechanical equipment, thereby ensuring the availability of the adjacent sample. When the model of the sample electromechanical equipment is the same as the model of the electromechanical equipment, the first adjacent constraint is deemed to be satisfied. Secondly, a service time deviation threshold is preset, and when the difference between the service time of the sample electromechanical equipment and the service time of the electromechanical equipment is less than or equal to the threshold, the second adjacent constraint is deemed to be satisfied. By limiting the deviation range of the service time, it is ensured that the adjacent sample and the equipment to be monitored are in a similar wear stage and health level, thereby improving the comparability and correlation of the sample data. Then, each electromechanical equipment operation control parameter is used as an electromechanical equipment monitoring characteristic attribute, and a deviation threshold is preset for each electromechanical equipment monitoring characteristic attribute to form a control parameter attribute deviation threshold set. When the deviation of the control parameters of the sample electromechanical equipment and the control parameters of the electromechanical equipment to be monitored in each attribute is less than or equal to the corresponding deviation threshold, it is considered to meet the third adjacency constraint condition, thereby ensuring that the operating conditions of the adjacent sample and the equipment to be monitored are similar, and improving the supporting role of the sample data for fault prediction and diagnosis. Subsequently, with the monitoring characteristic attributes of electromechanical equipment as the target, sample data that simultaneously meets the three adjacency constraints of the first adjacency constraint condition, the second adjacency constraint condition and the third adjacency constraint condition are collected, and the number of samples is not less than 50, and the first-level adjacency samples are obtained. High-quality samples are screened by the intersection of multi-dimensional constraints, and the statistical significance of the samples is guaranteed by the quantity requirements, laying a solid foundation for subsequent data analysis.

[0056] By constructing three adjacent constraints from the three perspectives of equipment model, service time, and operating conditions, the similarity between adjacent samples and the equipment to be monitored is ensured in multiple dimensions. On this basis, three constraints are applied, and the target and quantity of sample collection are specified to obtain high-quality and highly reliable first-level adjacent sample data, which provides important support for improving the accuracy and practicality of electromechanical equipment fault monitoring.

[0057] Furthermore, the embodiment of the present application also includes:

[0058] S310: Step 1: According to the adjacent sample tree, extract the first record feature matrix of the first adjacent sample node, the second record feature matrix of the second adjacent sample node, and so on until the Mth record feature matrix of the Mth adjacent sample node, wherein the first adjacent sample node until the Mth adjacent sample node are bottom-level leaf nodes and have the same parent node.

[0059] S320: Step 2: Perform frequent fitting on the first record feature matrix, the second record feature matrix, and up to the Mth record feature matrix to generate a first fusion feature matrix.

[0060] S330: Step 3: Prune the first adjacent sample node to the Mth adjacent sample node, and add the first fusion feature matrix into the parent node.

[0061] S340: Repeat steps 1 to 3, and when only the root node remains in the adjacent sample tree, output the electromechanical equipment standard feature matrix of the root node.

[0062] In a feasible implementation, when generating a standard feature matrix of electromechanical equipment, first, according to the adjacent sample tree, the first record feature matrix of the first adjacent sample node, the second record feature matrix of the second adjacent sample node, and the Mth record feature matrix of the Mth adjacent sample node are extracted, so as to obtain the original sample data at the bottom of the adjacent sample tree, and provide input for subsequent feature aggregation. Among them, the first adjacent sample node to the Mth adjacent sample node are all bottom leaf nodes of the adjacent sample tree, and have the same parent node. Then, the first record feature matrix, the second record feature matrix to the Mth record feature matrix are used as input, and the common features contained in these matrices are extracted by the method of frequent fitting, and a fused feature matrix is ​​generated, which is called the first fused feature matrix. Among them, frequent fitting can use statistical methods, such as mode, mean, median, etc., to obtain the most representative and stable feature representation. Subsequently, in the adjacent sample tree, the first adjacent sample node to the Mth adjacent sample node are pruned, and the generated first fused feature matrix is ​​added to the corresponding parent node. Through the pruning operation, the original sample nodes at the bottom of the adjacent sample tree are removed and replaced with the fused feature matrix. While simplifying the sample tree structure, the aggregation and abstraction of features are achieved, so that the parent node can represent the feature information of its child nodes at a higher level.

[0063] Repeat the node matrix aggregation until only the root node remains in the adjacent sample tree. In the iterative process, the adjacent sample tree is pruned and feature aggregated layer by layer, and the feature representation of the sample data is continuously refined and abstracted. When the pruning process reaches the root node, the entire adjacent sample tree is reduced to a root node, and the fused feature matrix it carries is the standard feature matrix of the electromechanical equipment. This matrix integrates the feature information of the entire adjacent sample tree and reflects the standard feature performance of the electromechanical equipment in a non-fault state.

[0064] By making full use of the hierarchical structure of the adjacent sample tree and using the frequent fitting method, the common features of sample data at different levels are effectively extracted and integrated, and a highly condensed and abstract standard feature representation, namely the standard feature matrix of electromechanical equipment, is obtained. Based on this standard feature matrix, the health status of electromechanical equipment can be accurately evaluated, and potential fault signs can be discovered early, thereby achieving accurate and efficient equipment monitoring and maintenance. At the same time, recursive aggregation also greatly compresses the data size, reduces storage and computing overhead, and improves operational efficiency and real-time performance.

[0065] Furthermore, the embodiment of the present application also includes:

[0066] S321: Extracting a first feature attribute record value set to a Qth feature attribute record value set according to the first record feature matrix, the second record feature matrix to the Mth record feature matrix.

[0067] S322: performing outlier removal processing on the first characteristic attribute record value set to obtain a first characteristic attribute set of record values, and counting distribution intervals of the first characteristic attribute set of record values ​​to set the distribution intervals as a first characteristic attribute fusion value.

[0068] S323: until the Qth feature attribute record value set is processed to remove outliers, a Qth feature attribute set record value is obtained, and a distribution interval of the Qth feature attribute set record value is counted and set as the Qth feature attribute fusion value.

[0069] S324: Constructing the first fused feature matrix according to the first feature attribute fusion value up to the Qth feature attribute fusion value.

[0070] In a preferred embodiment, when generating the first fusion feature matrix, first, from the first record feature matrix, the second record feature matrix to the M record feature matrix, respectively extract the record value set corresponding to each feature attribute to form the first feature attribute record value set to the Qth feature attribute record value set. Each set contains the values ​​of M samples on the same feature attribute, so that the record values ​​of the same feature attribute of different samples are concentrated together to prepare for the subsequent frequent fitting. Then, the first feature attribute record value set is processed to remove outliers, and the abnormal values ​​and noise interference in the first feature attribute record value set are removed to obtain a more concentrated and stable record value subset, which is called the first feature attribute set record value. Next, the distribution interval of the record value in the first feature attribute set is counted and set as the first feature attribute fusion value. For example, by calculating the mean, the central tendency of the record value in the first feature attribute set is characterized. Repeat the process until the Qth feature attribute record value set is processed to remove outliers, obtain the corresponding feature attribute set record values, and count their distribution intervals, which are set as the corresponding feature attribute fusion values, thereby obtaining the first feature attribute fusion value until the Qth feature attribute fusion value. Each fusion value represents the frequency performance of a feature attribute in all samples.

[0071] Subsequently, the first fusion feature matrix is ​​constructed based on the fusion value of the first feature attribute to the fusion value of the Qth feature attribute. Specifically, the fusion value of each feature attribute is combined in the original feature attribute order to form a new matrix, namely the first fusion feature matrix. Each column of the matrix corresponds to a feature attribute, and the elements in the column are the frequency expression of the feature attribute, reflecting the central tendency of all samples on the feature attribute.

[0072] Through the frequent fitting process, the feature attribute record value set is extracted, outliers are removed, and distribution interval statistics are performed, which effectively suppresses the influence of outliers and noise, extracts the frequent performance of each feature attribute in all samples, and constructs a fusion feature matrix based on this, improving the quality and credibility of the fusion feature matrix. Based on the high-quality fusion feature matrix, the standard feature performance of electromechanical equipment can be more accurately characterized, laying a solid data foundation for fault diagnosis and prediction. At the same time, frequent fitting also plays a role in feature compression and noise reduction, reducing the data scale, reducing the computational complexity, and improving the operation efficiency.

[0073] Furthermore, the embodiment of the present application also includes:

[0074] S3221: Obtain the first characteristic attribute record value of the first characteristic attribute record value set.

[0075] S3222: Delete the first characteristic attribute record value from the first characteristic attribute record value set to obtain a first characteristic attribute update record value set.

[0076] S3223: Calculate the average of the first characteristic attribute record values ​​of the first characteristic attribute update record value set.

[0077] S3224: traverse the first characteristic attribute update record value set and the first characteristic attribute record value mean to calculate the deviation modulus, obtain the first characteristic attribute update record value deviation modulus set, calculate the mean of the first characteristic attribute deviation modulus set, and generate the first characteristic attribute average deviation modulus.

[0078] S3225: Calculate the deviation modulus between the first characteristic attribute record value and the mean value of the first characteristic attribute record value to obtain the deviation modulus of the first characteristic attribute record value.

[0079] S3226: When the deviation modulus of the first characteristic attribute record value is greater than 4 times the average deviation modulus of the first characteristic attribute, the first characteristic attribute record value is deleted.

[0080] S3227: After traversing the first characteristic attribute record value set, output the record values ​​in the first characteristic attribute set.

[0081] In a preferred embodiment, when obtaining the record values ​​in the first characteristic attribute set, first, extract the record values ​​of each sample on the first characteristic attribute one by one from the first characteristic attribute record value set as the first characteristic attribute record value currently being processed. Then, delete the first characteristic attribute record value from the first characteristic attribute record value set to form a new record value subset, and obtain the first characteristic attribute update record value set as the first characteristic attribute update record value set. Subsequently, calculate the arithmetic average of all the record values ​​in the first characteristic attribute update record value set to obtain the first characteristic attribute record value mean, which reflects the central tendency of other record values ​​except the current first characteristic attribute record value. Then, traverse the first characteristic attribute update record value set, calculate the absolute value of the difference between each record value and the first characteristic attribute record value mean, and obtain the first characteristic attribute update record value deviation modulus value set. Calculate the arithmetic average of all the deviation modulus values ​​in the deviation modulus value set to generate the first characteristic attribute average deviation modulus value, which reflects the average deviation degree of the record values ​​in the first characteristic attribute update record value set relative to the mean.

[0082] Afterwards, the absolute value of the difference between the first feature attribute record value and the mean of the first feature attribute record value is calculated to obtain the first feature attribute record value deviation modulus, which reflects the degree of deviation of the current record value relative to the mean of other record values. Subsequently, it is determined whether the first feature attribute record value deviation modulus is greater than 4 times the first feature attribute average deviation modulus. If so, the first feature attribute record value is marked as an outlier and deleted from the first feature attribute record value set. Repeat the execution until all record values ​​in the first feature attribute record value set are traversed. During the iteration process, the record value set and the mean are continuously updated, and each record value is determined one by one to see if it is an outlier. When the traversal is completed, the remaining record values ​​are the record values ​​in the first feature attribute set after the outliers are removed.

[0083] By identifying and eliminating outliers in the set of recorded values ​​of the first characteristic attribute, a more concentrated and stable set of recorded values ​​of the first characteristic attribute is obtained, thereby more accurately characterizing the frequent performance of the first characteristic attribute, reducing the interference of outliers on frequent fitting, and improving the quality and reliability of the fusion feature matrix.

[0084] Furthermore, the embodiment of the present application also includes:

[0085] S510: Configuring a plurality of electromechanical device characteristic deviation vector record matrices of a preset electromechanical device model.

[0086] S520: traverse the plurality of electromechanical equipment characteristic deviation vector record matrices to collect a plurality of groups of equipment detection results of the preset electromechanical equipment models.

[0087] S530: Traverse the multiple groups of equipment detection results, count the fault detection ratios, and obtain the true values ​​of the fault probabilities of multiple electromechanical equipment.

[0088] S540: Using the true values ​​of the multiple electromechanical equipment failure probabilities as supervision data and the multiple electromechanical equipment characteristic deviation vector record matrices as input data, the electromechanical equipment failure probability prediction model is trained.

[0089] In a preferred embodiment, in order to construct a prediction model for the probability of failure of electromechanical equipment, first, multiple electromechanical equipment of the same model as the electromechanical equipment to be monitored are selected from the historical database, the characteristic deviation vectors of these equipment at different time points are extracted, and multiple electromechanical equipment characteristic deviation vector record matrices are constructed, each matrix containing the characteristic deviation data of the same model of electromechanical equipment in different states. At the same time, multiple electromechanical equipment characteristic deviation vector record matrices are traversed, and the results of previous fault detection and health assessment are collected for the electromechanical equipment corresponding to each electromechanical equipment characteristic deviation vector record matrix, forming multiple groups of equipment detection results, indicating whether the corresponding electromechanical equipment fails at a specific time point. Subsequently, multiple groups of equipment detection results are traversed, and the detection results of each electromechanical equipment are statistically analyzed, and the proportion of the frequency of failure to the total number of detections is calculated as the true value of the failure probability of the equipment in the corresponding state, and multiple true values ​​of the failure probability of electromechanical equipment are obtained. The true value of the failure probability of electromechanical equipment reflects the true probability of failure of electromechanical equipment under the condition of given characteristic deviation.

[0090] Afterwards, the true values ​​of the failure probabilities of multiple electromechanical equipment are used as supervisory data, and the characteristic deviation vector record matrices of multiple electromechanical equipment are used as input data to train the electromechanical equipment failure probability prediction model. Specifically, machine learning algorithms such as logistic regression, decision tree, support vector machine, etc. are selected to build the electromechanical equipment failure probability prediction model. During the training process, the characteristic deviation vector record matrix is ​​used as the input feature of the model, and the corresponding true value of the failure probability is used as the target output of the model, and the model parameters are optimized using supervised learning. Through repeated iterations and tuning, the model gradually learns the intrinsic relationship between the characteristic deviation and the failure probability, and has the ability to predict the failure probability based on the new characteristic deviation vector.

[0091] By constructing a failure probability prediction model for electromechanical equipment, we fully utilize historical operation data and fault detection data, and establish a mapping relationship between feature deviation and failure probability. This enables the model to accurately predict the failure risk of electromechanical equipment based on the feature deviation data collected in real time, and to provide early warning of potential failure risks, thereby providing strong support for condition-based maintenance and fault prevention, and minimizing failure downtime and economic losses.

[0092] Furthermore, the electromechanical equipment monitoring characteristic attributes include at least time domain vibration feature descriptors, time domain audio feature descriptors, frequency domain vibration feature descriptors, frequency domain audio feature descriptors, time-frequency domain vibration feature descriptors, perceptual audio feature descriptors and statistical audio feature descriptors.

[0093] In a feasible implementation, the monitoring characteristic attributes of electromechanical equipment include at least time domain vibration feature descriptors, time domain audio feature descriptors, frequency domain vibration feature descriptors, frequency domain audio feature descriptors, time-frequency domain vibration feature descriptors, perceptual audio feature descriptors, and statistical audio feature descriptors. These feature descriptors describe the vibration and sound signal characteristics of electromechanical equipment from different perspectives, providing rich information for comprehensive assessment of the health status of the equipment.

[0094] Among them, the time domain vibration feature descriptors include mean, standard deviation, root mean square value, peak value, skewness, kurtosis, pulse factor, waveform factor and margin factor, which quantitatively describe the time domain characteristics of electromechanical equipment vibration from the aspects of vibration signal offset, fluctuation degree, energy size, amplitude range, peak degree, distribution skewness, distribution sharpness, impact component, waveform sharpness and fault degree, so as to capture abnormal changes in equipment vibration and detect potential mechanical failures early; the time domain audio feature descriptors include short-time energy, short-time average amplitude, short-time zero-crossing rate and pre-emphasis, which describe the time domain characteristics of electromechanical equipment vibration from the time domain perspective of sound signals. The time domain characteristics of the signal, such as energy change, amplitude change, frequency change and high-frequency attenuation compensation, are described, so as to evaluate the loudness, frequency and dynamic change of electromechanical equipment noise, and monitor the abnormal noise and working status of the equipment; the frequency domain vibration feature descriptor includes frequency mean, frequency center of gravity, root mean square frequency, standard deviation frequency, power spectrum density, spectrum mean and spectrum variance, etc. From the frequency domain perspective of the vibration signal, the frequency domain characteristics such as amplitude distribution, energy distribution, spectrum concentration degree and spectrum dispersion degree of different frequency components are described, so as to identify the characteristic frequency components in the equipment vibration and judge the fault type and severity; the frequency domain audio feature descriptor Descriptors include spectrum centroid, spectrum bandwidth, spectrum roll-off, spectrum flux and spectrum entropy, etc., which describe the frequency domain characteristics of the spectrum, such as the center of gravity position of the spectrum, the degree of spectrum expansion, the distribution cutoff point of spectrum energy, the dynamic change of the spectrum and the uniformity of the spectrum from the frequency domain perspective of the sound signal, so as to analyze the frequency characteristics of electromechanical equipment noise, identify different types of abnormal audio, and diagnose the type of equipment failure; time-frequency domain vibration feature descriptors include wavelet transform coefficients, Wigner-Ville distribution and Chirp features, etc., which describe the joint distribution characteristics of the signal in the two dimensions of time and frequency, so as to analyze the frequency characteristics of the vibration signal. Non-stationary characteristics and instantaneous frequency changes can accurately diagnose complex mechanical faults; perceptual audio feature descriptors include Mel-frequency cepstral coefficients, Chroma features, etc., which simulate the human ear's perception of sound and describe the timbre, pitch and other perceptual attributes of sound signals, thereby evaluating the sound quality and timbre changes of electromechanical equipment noise from the perspective of human ear perception, and better detecting abnormal conditions of equipment; statistical audio feature descriptors include energy entropy and spectral extension, etc., which describe the time domain distribution and frequency domain distribution characteristics of sound signals, thereby evaluating the time continuity and spectral concentration of electromechanical equipment noise and comprehensively characterizing the sound characteristics of the equipment.

[0095] By introducing various types of vibration and audio feature descriptors, the state characteristics of electromechanical equipment are characterized from multiple angles such as time domain, frequency domain, time-frequency domain, perception and statistics. This can comprehensively evaluate the health status of equipment and accurately detect various types of faults, providing a reliable basis for intelligent maintenance and fault prevention of equipment.

[0096] The electromechanical equipment fault monitoring method combined with adjacent sample analysis provided by the embodiment of the present invention has at least the following technical effects:

[0097] Interact with electromechanical equipment, receive electromechanical equipment operation control parameters and electromechanical equipment monitoring feature matrix, characterize the real-time performance of multiple monitoring indicators of the equipment, and provide original data support for subsequent fault diagnosis and analysis. Perform multi-level adjacency search in non-fault state according to the electromechanical equipment operation control parameters to obtain the adjacency sample tree, effectively expand the number and diversity of normal state samples, and lay a data foundation for building more comprehensive and reliable standard features. Recursively aggregate the adjacency sample tree to generate the electromechanical equipment standard feature matrix, which reflects the standard performance value range of each monitoring indicator of the electromechanical equipment during the entire normal operation process, and provides a reference for fault assessment. Compare the electromechanical equipment monitoring feature matrix and the electromechanical equipment standard feature matrix, construct the electromechanical equipment feature deviation vector matrix, calculate the degree of deviation of each monitoring indicator from the standard value range, and establish a deviation vector matrix that characterizes the degree of deviation, characterize the distance of each dimension of the current state of the electromechanical equipment from the normal state, and provide a direct basis for fault diagnosis and early warning. The electromechanical equipment characteristic deviation vector matrix is ​​input into the electromechanical equipment failure probability prediction model to generate the electromechanical equipment failure probability, and the multi-dimensional deviation is mapped into the quantitative probability of failure risk, which is convenient for intuitively evaluating the health level of the current equipment status. When the electromechanical equipment failure probability is greater than or equal to the failure probability threshold, the electromechanical equipment maintenance instruction is generated and sent to the electromechanical equipment operation and maintenance management end to eliminate the fault in the early stage, improve the efficiency and accuracy of fault monitoring, and ensure the safe and stable operation of electromechanical equipment.

[0098] Embodiment 2:

[0099] See also Figure 2 , Figure 2 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 2 As shown, an electronic device 100 provided by an embodiment of the present invention includes a memory 110, a processor 120, and a first computer program 111 stored in the memory 110 and executable on the processor 120. When the processor 120 executes the computer program 111, a method for monitoring electromechanical equipment faults combined with adjacent sample analysis is implemented.

[0100] Embodiment three:

[0101] See also Figure 3 , Figure 3 A schematic diagram of an embodiment of a computer-readable storage medium provided in an embodiment of the present invention. Figure 3 As shown, this embodiment provides a computer-readable storage medium 200 on which a second computer program 211 is stored. When the second computer program 211 is executed by a processor, the electromechanical equipment fault monitoring method combined with adjacent sample analysis is implemented.

[0102] 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.

[0103] 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.

[0104] 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 device that provides the functions specified in a block or multiple blocks.

[0105] 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.

[0106] 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 1A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0107] 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.

[0108] 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 monitoring electromechanical equipment faults in combination with adjacent sample analysis, characterized in that: A system for monitoring faults of electromechanical equipment in combination with adjacent sample analysis, the system being communicatively connected to the electromechanical equipment, comprising: Interact with electromechanical equipment, receive electromechanical equipment operation control parameters and electromechanical equipment monitoring feature matrix; Perform a multi-level adjacency search in a non-fault state according to the electromechanical equipment operation control parameters to obtain an adjacency sample tree; Recursively aggregating the adjacent sample trees to generate a standard feature matrix of electromechanical equipment; Comparing the electromechanical equipment monitoring characteristic matrix with the electromechanical equipment standard characteristic matrix, and constructing an electromechanical equipment characteristic deviation vector matrix; Inputting the electromechanical equipment characteristic deviation vector matrix into an electromechanical equipment failure probability prediction model to generate an electromechanical equipment failure probability; When the failure probability of the electromechanical equipment is greater than or equal to the failure probability threshold, a maintenance instruction for the electromechanical equipment is generated and sent to the electromechanical equipment operation and maintenance management terminal; Among them, performing a non-fault state multi-level adjacency search according to the electromechanical equipment operation control parameters to obtain an adjacency sample tree includes: Obtaining a static tag of the electromechanical equipment, wherein the static tag of the electromechanical equipment includes a model of the electromechanical equipment and a service time of the electromechanical equipment; Performing a non-fault state primary adjacency search according to the electromechanical equipment model, the service time of the electromechanical equipment, and the operation control parameters of the electromechanical equipment to obtain a primary adjacency sample, wherein the primary adjacency sample has the electromechanical equipment primary sample operation control parameters; Based on the electromechanical equipment primary sample operation control parameter, traverse the primary adjacent sample to perform a non-fault state secondary adjacent search to obtain a secondary adjacent sample, wherein the secondary adjacent sample has the electromechanical equipment secondary sample operation control parameter; Based on the operation control parameters of the electromechanical equipment secondary sample, traverse the secondary adjacent samples to perform a non-fault state third-level adjacent search to obtain a third-level adjacent sample; The root node is configured to be empty, the first-level adjacent sample is set as a first-level leaf node, the second-level adjacent sample is set as a second-level leaf node, and the third-level adjacent sample is set as a third-level leaf node, thereby generating the adjacent sample tree; The adjacent sample tree is recursively aggregated to generate a standard feature matrix of electromechanical equipment, including: Step 1: extracting a first record feature matrix of a first adjacent sample node, a second record feature matrix of a second adjacent sample node, and an Mth record feature matrix of an Mth adjacent sample node according to the adjacent sample tree, wherein the first adjacent sample node to the Mth adjacent sample node are bottom-level leaf nodes and have the same parent node; Step 2: Perform frequent fitting on the first record feature matrix, the second record feature matrix, and up to the Mth record feature matrix to generate a first fusion feature matrix; Step 3: Prune the first adjacent sample node to the Mth adjacent sample node, and add the first fusion feature matrix to the parent node; Repeat steps 1 to 3, and when only the root node remains in the adjacent sample tree, output the electromechanical equipment standard feature matrix of the root node; The step of performing frequent fitting on the first record feature matrix, the second record feature matrix, and the Mth record feature matrix to generate a first fusion feature matrix includes: Extracting a first feature attribute record value set to a Qth feature attribute record value set according to the first record feature matrix, the second record feature matrix to the Mth record feature matrix; Performing outlier removal processing on the first characteristic attribute record value set to obtain a first characteristic attribute concentrated record value, and counting the distribution interval of the first characteristic attribute concentrated record value to set it as a first characteristic attribute fusion value; Until the Qth feature attribute record value set is processed to remove outliers, a Qth feature attribute set record value is obtained, and a distribution interval of the Qth feature attribute set record value is counted and set as the Qth feature attribute fusion value; constructing the first fusion feature matrix according to the first feature attribute fusion value up to the Qth feature attribute fusion value; The step of performing outlier removal processing on the first characteristic attribute record value set to obtain the first characteristic attribute set record values ​​includes: Obtaining a first characteristic attribute record value of the first characteristic attribute record value set; Deleting the first characteristic attribute record value from the first characteristic attribute record value set to obtain a first characteristic attribute update record value set; Calculate the average of the first characteristic attribute record values ​​of the first characteristic attribute update record value set; Traversing the first characteristic attribute update record value set and the first characteristic attribute record value mean to find the deviation modulus value, obtaining the first characteristic attribute update record value deviation modulus value set, calculating the mean of the first characteristic attribute deviation modulus value set, and generating the first characteristic attribute average deviation modulus value; Calculate the deviation modulus of the first characteristic attribute record value and the mean of the first characteristic attribute record value to obtain the deviation modulus of the first characteristic attribute record value; When the deviation modulus of the first characteristic attribute record value is greater than 4 times the average deviation modulus of the first characteristic attribute, the first characteristic attribute record value is deleted; When the first characteristic attribute record value set is traversed, the record values ​​in the first characteristic attribute set are output.

2. The method according to claim 1, characterized in that A non-fault state primary adjacency search is performed according to the electromechanical equipment model, the service time of the electromechanical equipment, and the operation control parameters of the electromechanical equipment to obtain a primary adjacency sample, including: According to the electromechanical device model, a first adjacency constraint condition is constructed, and when the sample electromechanical device model is the same as the electromechanical device model, the first adjacency constraint condition is considered to be satisfied; According to the service time of the electromechanical equipment, based on the service time deviation threshold, a second adjacency constraint condition is constructed, and when the service time deviation between the service time of the sample electromechanical equipment and the service time of the electromechanical equipment is less than or equal to the service time deviation threshold, the second adjacency constraint condition is deemed to be satisfied; According to the electromechanical equipment operation control parameters, based on the control parameter attribute deviation threshold set, a third adjacency constraint condition is constructed, and when the control parameter deviation set of the sample electromechanical equipment operation control parameters and the electromechanical equipment operation control parameters is less than or equal to the one-to-one corresponding control parameter attribute deviation threshold set, the third adjacency constraint condition is deemed to be satisfied; Taking the characteristic attributes of electromechanical equipment monitoring as a target, at least 50 first-level adjacency samples that simultaneously satisfy the first adjacency constraint condition, the second adjacency constraint condition and the third adjacency constraint condition are collected.

3. The method according to claim 1, characterized in that Inputting the electromechanical equipment characteristic deviation vector matrix into the electromechanical equipment failure probability prediction model to generate the electromechanical equipment failure probability, including: Configuring a plurality of electromechanical device characteristic deviation vector record matrices of a preset electromechanical device model; Traversing the plurality of electromechanical equipment characteristic deviation vector record matrices, and collecting a plurality of groups of equipment detection results of the preset electromechanical equipment models; Traversing the multiple groups of equipment detection results, counting the fault detection ratios, and obtaining the true values ​​of the fault probabilities of multiple electromechanical equipment; The electromechanical equipment failure probability prediction model is trained using the multiple electromechanical equipment failure probability true values ​​as supervision data and the multiple electromechanical equipment characteristic deviation vector record matrices as input data.

4. The method according to claim 1, characterized in that The characteristic attributes of electromechanical equipment monitoring include at least time domain vibration feature descriptor, time domain audio feature descriptor, frequency domain vibration feature descriptor, frequency domain audio feature descriptor, time-frequency domain vibration feature descriptor, perceptual audio feature descriptor and statistical audio feature descriptor.

5. An electronic device, characterized in that: include: Memory for storing computer software programs; A processor is used to read and execute the computer software program, thereby implementing the electromechanical equipment fault monitoring method combined with adjacent sample analysis as described in any one of claims 1 to 4.

6. A non-transitory computer-readable storage medium, characterized in that: The storage medium stores a computer software program, and when the computer software program is executed by the processor, the method for monitoring electromechanical equipment faults in combination with adjacent sample analysis according to any one of claims 1 to 4 is implemented.

Images