Fault identification method based on intelligent model

Abstract

The invention relates to the technical field of fault recognition, in particular to a fault recognition method based on an intelligent model, and the method comprises the steps: collecting the comprehensive state information of equipment and the environment of the equipment through a multi-modal high-precision sensing element, and formulating a fusion strategy to form multi-dimensional tensor data, thereby providing a basis for subsequent analysis; the method comprises the following steps: constructing a semantic space, mining a fault embryo mode by using a generative adversarial network technology, and performing traceability analysis; based on the fault embryo mode and the traceability result, constructing a hidden order model reflecting the dynamic change of the equipment operation architecture; hidden correlation between fault embryos and features is searched by using an association rule recognition algorithm, and a reasoning prediction engine is constructed to perform fault prediction, so that the accuracy and timeliness of fault prediction are improved; and in combination with a reasoning prediction result, a self-adaptive defense mechanism is established, the fault state of the equipment operation architecture is identified and processed, and the safety and stability of equipment operation are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of fault identification, and in particular to a fault identification method based on an intelligent model. Background Art

[0002] There are still some challenges in current fault identification, including the following aspects: incomplete monitoring. Traditional fault identification methods may not be able to fully monitor the comprehensive status information of the equipment and its environment, resulting in limited accuracy and timeliness of fault identification; it is difficult to mine fault association rules. The association rules between faults and features are often hidden deep. Traditional fault identification methods are difficult to effectively mine these rules, affecting the accuracy of fault prediction; fault identification methods are not intelligent enough. Traditional fault identification methods often rely on manual experience and rules and lack intelligent processing methods, resulting in limited efficiency and accuracy of fault identification. To this end, the present invention proposes a fault identification method based on an intelligent model. Summary of the invention

[0003] The purpose of the present invention is to solve the problems in the background technology and to propose a fault identification method based on an intelligent model.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions: Fault identification methods based on intelligent models include: S1. Use multimodal high-precision sensing elements to build a multi-dimensional field sensing structure to collect comprehensive status information of the device and its environment; formulate corresponding fusion coding strategies for each acquired field data to obtain fused multi-dimensional tensor data; S2. Use natural language processing and knowledge graph technology to construct a semantic space and generate a semantic fusion feature data set; in the constructed semantic space, use the generative adversarial network technology to mine the fault embryo pattern; if the fault embryo pattern is identified, use the entity relationship network in the semantic space to perform traceability analysis; S3. Construct a hidden order model based on the identified fault embryo pattern and its traceability analysis results; S4. Based on the latent order model, the association rule recognition algorithm is used to further find the association rules between the identified fault embryo pattern and the semantic fusion feature data set; according to the analyzed association rules and the status of the current equipment operation architecture, an inference prediction engine based on latent order association is constructed, and inference prediction is performed; S5. Treat the normal state of the equipment operation architecture as itself and the fault state as not itself; combine the inference prediction results to establish an adaptive defense mechanism.

[0005] Furthermore, the multi-dimensional field specifically includes microscopic electromagnetic fields, weak gravitational fields, and macroscopic geographical environments and spatial positions; the multi-modal high-precision sensing elements specifically include near-field optical sensors, gravitational wave sensors, and satellite positioning and geographic information systems.

[0006] Furthermore, the process of S1 includes: Use near-field optical sensors to capture changes in the microscopic electromagnetic field distribution on the device surface; use gravitational wave sensors to monitor the weak gravitational field fluctuations caused by changes in mass distribution of large device clusters; combine satellite positioning with geographic information systems to obtain the macroscopic geographical environment and spatial location information of the device; obtain microscopic electromagnetic field data on the device surface, weak gravitational field fluctuation data of large device clusters, and macroscopic geographical environment and spatial location data of the device; Collect various text description data during equipment operation; Map the microscopic electromagnetic field data to the complex space, and use the real and imaginary parts of the complex numbers to represent the coupling relationship between different physical quantities; for the weak gravitational field fluctuation data, use the wavelet transform algorithm to encode and process it to form a data vector with high-dimensional characteristics; for the macroscopic geographic space data, convert the geographic information into a topological feature vector through topological mapping; For text description data, natural language processing technology is used to convert it into word vector representation; The tensor product operation is used to merge the feature vectors obtained by these different encoding methods into a unified multi-dimensional tensor data.

[0007] Furthermore, S2 includes: Extract equipment-related entities and relationships between entities from the fused multi-dimensional tensor data to build a fault knowledge graph; The various features in the fused multidimensional tensor data are semantically associated with the entities constructed in the knowledge graph, and semantic labels are assigned to each data feature, so as to form a semantic space that integrates multi-source data and domain knowledge, and generate a semantic fusion feature data set; among which, the various features in the fused multidimensional tensor data include microscopic electromagnetic field features, weak gravitational field features, macroscopic geographic space features, and text description data features; The generator is responsible for generating possible fault embryo pattern samples, and the discriminator distinguishes whether the fault embryo pattern samples are real fault embryos. When generating fault embryo pattern samples, the generator comprehensively uses the equipment operation status reflected by each field information in the generated semantic fusion feature data set. Through continuous adversarial training, the generator is used to learn the potential fault embryo feature patterns in the semantic fusion feature dataset; Along the relationship path between entities in the knowledge graph, the origin and development context of the fault embryo are traced back; the origin of the fault embryo can be traced back to multiple aspects, including the origin of the equipment itself, the origin of the external environment, and the origin of human operation; the development context of the fault embryo includes the initial evolution of the fault, the expansion of the impact range, and the changes in the equipment operation level; Through calculation and analysis, relevant quantitative values ​​of traceability analysis are obtained.

[0008] Furthermore, the implicit order model regards the entire device operation architecture as a dynamic complex system, in which there are complex and hidden interactions between the various elements; among them, the device operation architecture refers to a comprehensive device operation system composed of device components, software management programs, operators and operating environment; the elements of the device operation architecture cover device components, software management programs, operators and operating environment.

[0009] Furthermore, S3 includes: Quantify each element in the equipment operation architecture and use nonlinear dynamic equations to analyze the dynamic relationship between the elements; During the construction process, the information of each element in the equipment operation architecture is added to the parameter setting of the hidden order model, and the relevant quantitative values ​​assigned by the fault embryo mode and its traceability analysis results are combined to associate it with the parameters of the hidden order model; Determine the normal state or potential fault state of the equipment operation architecture through the hidden order model: define the threshold of the normal state; compare and analyze the output value of the hidden order model with the defined threshold. If the output value of the hidden order model is within the normal state threshold range, that is, the output value is greater than or equal to the lower limit threshold of the normal state and less than or equal to the upper limit threshold of the normal state, then it is determined that the current state of the equipment operation architecture meets the normal operation standard and is considered to be in a normal state; if the output value of the hidden order model exceeds the normal state threshold range, that is, the output value is less than the lower limit threshold of the normal state or greater than the upper limit threshold of the normal state, it indicates that there is a problem with the equipment operation architecture and it is considered to be in a potential fault state; At the same time, it captures the order structure hidden within the equipment operation architecture that causes the transition from normal state to fault state.

[0010] Furthermore, S4 includes: Obtain the identified fault embryo pattern and semantic fusion feature dataset: the fault embryo pattern is stored in the form of a vector, each element in the fault embryo pattern vector represents its corresponding fault feature identifier and quantified value, and the semantic fusion feature dataset is stored in the form of a data table, each row represents a data sample, and each column represents a fusion feature; The Apriori algorithm is used to mine frequent item sets in the integrated fault embryo pattern vector and semantic fusion feature data set, and the minimum support and confidence thresholds are set to identify the feature combinations that frequently and simultaneously appear before the fault occurs. These feature combinations further constitute and filter out the association rules of the fault occurrence. Combine the acquired association rules with the state of the current equipment operation architecture to build an inference prediction engine based on implicit order association; use the inference prediction engine to make inference predictions about future faults based on the state of the current equipment operation architecture and the association rules: if the key monitoring parameters of the equipment operation architecture are found to meet the association rules, the inference prediction engine predicts the probability, time and type of fault occurrence; the key monitoring parameters are important indicators used in the equipment operation architecture to monitor and evaluate the equipment status, and are obtained in real time through the sensor network; At the same time, the Bayesian reasoning method is used to correct the prediction results.

[0011] Furthermore, the Apriori algorithm is used to mine frequent itemsets in the integrated fault embryo pattern vector and semantic fusion feature dataset, and the minimum support and confidence thresholds are set to identify the feature combinations that frequently and simultaneously appear before the fault occurs, including: The fault embryo pattern vector is integrated with the semantic fusion feature data set to ensure that each element in the fault embryo pattern vector matches and corresponds to the corresponding fusion feature in the semantic fusion feature data set; the integrated data set is converted into a transactional database processed by the Apriori algorithm, where each transaction represents a set of multiple features, and these features come from the fault embryo pattern vector and the semantic fusion feature data set; a minimum support threshold is set to filter frequent item sets, where the support represents the frequency of an item set appearing in all transactions; a minimum confidence threshold is set to filter association rules; and the iterative process of the Apriori algorithm is used to start from a single item and gradually generate frequent itemsets containing more items. Item sets, where the generated frequent item sets represent feature combinations that often appear at the same time before a fault occurs; in each iteration, candidate item sets are generated based on the frequent item sets generated in the previous round, and the support of each candidate item set is calculated; candidate item sets whose support is not less than the minimum support threshold are retained as new frequent item sets for the next round of iteration to generate more candidate item sets; the iterative process is repeated until new frequent item sets can no longer be generated; for each frequent item set, all possible association rules are generated, and the support and confidence of each association rule are calculated; association rules whose support and confidence are not less than the set threshold are retained as the final fault occurrence association rules; the association relationship between the feature combination represented by the generated association rules and the fault occurrence is expressed.

[0012] Furthermore, S5 includes: Obtain the inference prediction results of the implicit order model and set the self-tolerance threshold set; when the probability value of the detected fault deviates from the range of the set self-tolerance threshold set, activate the adaptive defense mechanism to identify and process the fault state of the equipment operation architecture; generate an adaptive control strategy based on the type and severity of the fault.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: by constructing a multi-dimensional field perception structure through multimodal high-precision sensing elements, comprehensive monitoring of the equipment and its environment is achieved, and the accuracy and timeliness of fault identification are improved; by utilizing natural language processing and knowledge graph technology, a semantic space is constructed and fault embryo patterns are mined, thereby achieving early detection and source tracing analysis of faults, which helps to timely discover and deal with potential faults; by constructing a latent order model and an inference prediction engine, the association rules between faults and features can be further mined, thereby improving the accuracy of fault prediction; by introducing an adaptive defense mechanism, a control strategy can be generated according to the type and severity of the fault, effectively ensuring the safety and stability of equipment operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of the fault identification method based on intelligent model proposed in the present invention. 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 implementation regulations described 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0016] Reference Figure 1 , the fault identification method based on intelligent model includes: S1. Use multimodal high-precision sensing elements to build a multi-dimensional field sensing structure to collect comprehensive status information of the device and its environment; formulate corresponding fusion coding strategies for each acquired field data to obtain fused multi-dimensional tensor data; Among them, the multi-dimensional field includes microscopic electromagnetic fields, weak gravitational fields, and macroscopic geographical environments and spatial positions. The multi-modal high-precision sensing elements include near-field optical sensors (used to capture changes in the distribution of microscopic electromagnetic fields), gravitational wave sensors (used to monitor weak gravitational field fluctuations), and satellite positioning and geographic information systems (used to obtain macroscopic geographical environment and spatial position information). S2. Use natural language processing and knowledge graph technology to construct a semantic space and generate a semantic fusion feature data set; in the constructed semantic space, use the generative adversarial network technology to mine the fault embryo pattern; if the fault embryo pattern is identified, use the entity relationship network in the semantic space to perform source analysis to understand the origin and development context (possible causes and paths) of the fault; S3. Based on the identified fault embryo pattern and its traceability analysis results, a hidden order model is constructed; this model regards the entire equipment operation architecture as a dynamic composite system, in which there are complex and hidden interactions between the various elements; Among them, the equipment operation architecture refers to a comprehensive equipment operation system composed of equipment components, software management programs, operators and operating environment. For example, for an automated production line equipment, the equipment operation architecture includes all elements related to equipment operation, such as the mechanical devices of the production line, the program software that controls the operation of the production line, the operators and the factory environment where the production line is located; the elements of the equipment operation architecture include equipment components (such as motors, sensors, valves, etc.), software management programs, operators and operating environment (temperature, humidity, electromagnetic interference, etc.). These elements interact and influence each other and jointly determine the operating status of the equipment; in addition, the implicit order model is used to capture and describe these dynamic transformation behaviors hidden in the equipment operation architecture; S4. Based on the latent order model, the association rule recognition algorithm is used to further find the association rules between the identified fault embryo pattern and the semantic fusion feature data set; according to the analyzed association rules and the status of the current equipment operation architecture, an inference prediction engine based on latent order association is constructed, and inference prediction is performed to achieve accurate prediction of future faults; S5. Treat the normal state of the equipment operation architecture as itself and the fault state as not itself; combine the inference prediction results to establish an adaptive defense mechanism; Among them, by drawing on the self / non-self recognition in the biological immune system (that is, the ability of organisms to distinguish between their own cells and foreign invading pathogens), an adaptive defense mechanism is established, which can automatically adjust equipment operating parameters or take other preventive measures based on the reasoning and prediction results to avoid or mitigate the impact of failures.

[0017] It should be further explained that, in the specific implementation process, a multi-dimensional field perception structure is constructed using multi-modal high-precision sensing elements to collect comprehensive status information of the equipment and its environment; for each acquired field data, a corresponding fusion coding strategy is formulated to obtain the fused multi-dimensional tensor data in the following process: Near-field optical sensors are used to capture changes in the distribution of microscopic electromagnetic fields on the surface of the equipment: near-field optical sensors based on the principle of scanning near-field optical microscopy are selected, and the sensors are evenly distributed on the surface of the equipment. The sensor distribution density is 1 sensor per square centimeter. The sensors are calibrated using standard calibration blocks with a calibration accuracy of 0.1 microvolts per meter to ensure that accurate changes in the distribution of microscopic electromagnetic fields are captured; gravitational wave sensors are used to monitor weak gravitational field fluctuations caused by changes in mass distribution of large equipment clusters: the sensitivity range of gravitational wave sensors is adjusted according to the mass and distribution characteristics of large equipment clusters; other interference factors in the environment (such as the Earth's gravity) are eliminated by setting up shielding devices and using differential measurement methods. The influence of the slight fluctuation of the gravitational field of the body, the interference of the gravitational force of other nearby objects, etc. on the monitoring of the weak gravitational field fluctuations; combining satellite positioning with geographic information system to obtain the macroscopic geographical environment and spatial location information of the equipment: using the satellite positioning system of GPS and Beidou multi-system fusion, the data update frequency in the geographic information system is once an hour, and the data source is the fusion of official surveying and mapping data and real-time crowdsourcing data; obtaining the microscopic electromagnetic field data on the surface of the equipment, the weak gravitational field fluctuation data of large equipment clusters, and the macroscopic geographical environment and spatial location data of the equipment; it is understandable that these data come from different fields, contain detailed indications of the equipment's operating status, and fully reflect the comprehensive status of the equipment and its surrounding environment; Collect various text description data during the operation of the equipment, including log files, operation procedures, maintenance records, fault reports, etc., to reflect the background information of the equipment's operating status; Map the microscopic electromagnetic field data to the complex space, and use the real and imaginary parts of the complex numbers to represent the coupling relationship between different physical quantities: using a nonlinear mapping function (where a, b, and c are constants determined according to the characteristics of the equipment) Map the microscopic electromagnetic field data to the complex space; for the weak gravitational field fluctuation data, use the wavelet transform algorithm for encoding processing to form a data vector with high-dimensional characteristics: for example, use the Daubechies wavelet basis function for transformation; for the macroscopic geographic space data, convert the geographic information into a topological feature vector through topological mapping: map based on the mesh topology structure in graph theory, use the geographic coordinates in the geographic information as nodes, the connection relationship between geographic areas as edges, the node attributes as coordinate values, and the edge attributes as connection weights, and determine the weight size according to the distance; For text description data, natural language processing technology is used to convert it into word vector representation: the Word2Vec pre-trained model is used to process the text description data, and the specific abbreviations related to the device are standardized during the processing; The tensor product operation is used to fuse the feature vectors obtained by these different encoding methods into a unified multi-dimensional tensor data, fully retaining the key information of each field data: the principle of weight setting in the tensor product operation process is to set it according to the importance of each feature vector to the operating status of the equipment. For example, the weight of micro-electromagnetic field data is 0.3, the weight of weak gravitational field fluctuation data is 0.2, the weight of macro-geographic space data is 0.2, and the weight of text description data is 0.3. The fused multi-dimensional tensor data is normalized to the interval [0,1] by minimum-maximum normalization.

[0018] It should be further explained that, in the specific implementation process, natural language processing and knowledge graph technology are used to construct a semantic space and generate a semantic fusion feature data set; in the constructed semantic space, the generative adversarial network technology is used to mine the fault embryo pattern; if the fault embryo pattern is identified, the process of tracing the source analysis using the entity relationship network in the semantic space is as follows: Extract equipment-related entities (equipment components, fault types, maintenance measures) and relationships between entities (causal relationships or association relationships) from the fused multidimensional tensor data to build a fault knowledge graph; understandably, the sources of these entities and relationships are relatively broad, including both the information presented by the text description in the equipment operation data and the potential connections implied by other field information in the fused multidimensional tensor data; Specifically, A1. Use a combination of rule-based and machine learning methods to extract equipment-related entities from the fused multidimensional tensor data: create a predefined dictionary of equipment components, fault types, and maintenance measures. For equipment components, the dictionary contains the names of common parts of the equipment and their synonyms; use a convolutional neural network-based entity recognition model to perform a preliminary scan on the fused multidimensional tensor data, and divide the fused multidimensional tensor data into windows of fixed length, for example, each window contains 100 data points, where the convolution kernel size of the convolutional neural network-based entity recognition model is set to 3, the step size is 1, and features are extracted through multiple convolutional layers and pooling layers; for the data part with matching items in the predefined dictionary, directly mark it as the corresponding entity, and for the data part that does not match but is identified as potential by the entity recognition model, For the entity part, manual review and confirmation are then carried out; A2. The relationship between entities (causal relationship or association relationship) is determined by methods based on logical rules and data statistical analysis: For causal relationships, logical rules are set. For example, if the temperature of a component of the equipment rises (reflected by sensor data) and then a device failure occurs (reflected by fault report data), a causal relationship from temperature rise to failure is established; for association relationships, the co-occurrence frequency between entities is calculated. Taking equipment components and fault types as examples, the number of times a specific fault type occurs when a component appears within a certain time window (such as 1 day) is counted. When the co-occurrence frequency exceeds the set threshold (such as 0.3), it is determined that there is an association relationship. At the same time, based on the timestamp sequence in the fused multidimensional tensor data, a preliminary judgment is made as to whether there may be a causal relationship tendency; The various features in the fused multidimensional tensor data are semantically associated with the entities constructed in the knowledge graph, and semantic labels are assigned to each data feature, so as to form a semantic space that integrates multi-source data and domain knowledge, and generate a semantic fusion feature data set; wherein the various features in the fused multidimensional tensor data include microscopic electromagnetic field features, weak gravitational field features, macroscopic geographic space features, and text description data features; it is understandable that in this semantic space, the data features not only retain the original numerical value or attribute information, but also establish a connection with the knowledge graph entity with the help of semantic labels, and finally generate a semantic fusion feature data set, and the semantic fusion feature data set contains the encoded feature information of the original multi-source data, as well as clear semantic meanings; Specifically, B1 uses a method based on semantic similarity calculation to semantically associate various features in the fused multidimensional tensor data with the entities constructed in the knowledge graph: use a pre-trained word vector model specially trained for the equipment field (such as a Word2Vec model pre-trained on the equipment field corpus); for each data feature, calculate its cosine similarity with the entity word vector in the knowledge graph, for example, for a feature vector representing the temperature of the device, calculate its cosine similarity with the entity word vector of "equipment overheating fault" in the knowledge graph; assign semantic labels according to the value of semantic similarity, if the cosine similarity is greater than 0. 8, a strong correlation semantic label is assigned; if it is between 0.5-0.8, a medium correlation semantic label is assigned; if it is less than 0.5, a weak correlation semantic label is assigned; B2, data features are integrated with knowledge graph entities according to semantic labels to generate a semantic fusion feature dataset: for data features with strong correlation semantic labels, they are directly merged with the corresponding entities as an element in the semantic fusion feature dataset; for data features with medium correlation and weak correlation, they are weighted and merged according to their relationship weight with the entity (calculated by the co-occurrence frequency of the correlation relationship) to form a semantic fusion feature dataset; The generator is responsible for generating possible fault embryo pattern samples, and the discriminator distinguishes whether the fault embryo pattern samples are real fault embryos. When generating fault embryo pattern samples, the generator comprehensively uses the equipment operation status reflected by each field information in the generated semantic fusion feature data set. Through continuous adversarial training, the generator is used to learn the potential fault embryo feature patterns in the semantic fusion feature dataset; it is understandable that these potential fault embryo feature patterns may exist before the fault is obviously manifested, and are early signs of fault development; Specifically, C1. The generator adopts a multi-layer perceptron structure: the number of input layer nodes is determined according to the dimension of the semantic fusion feature dataset. For example, if the dataset has 100 features, the number of input layer nodes is 100; the hidden layer is set to 3 layers, the number of nodes in the first layer is 200, using the ReLU activation function, the number of nodes in the second layer is 150, using the ReLU activation function, and the number of nodes in the third layer is 100, using the tanh activation function; the number of nodes in the output layer is determined according to the encoding dimension of the fault embryo pattern sample. If the fault embryo pattern sample is represented by a 50-dimensional vector, the output layer is 100. The number of nodes in the output layer is 50; C2, the discriminator adopts a probability-based discrimination method: for the input fault embryo pattern sample, calculate the probability that it belongs to a real fault embryo; input the fault embryo pattern sample into the discriminator composed of a multi-layer perceptron, and the number of nodes in the input layer of the discriminator is the same as the number of nodes in the output layer of the generator (such as 50); the hidden layer is set to 2 layers, the number of nodes in each layer is 80, and the ReLU activation function is used; the output layer node is 1, and the sigmoid activation function is used. The output value represents the probability that the sample is a real fault embryo. When the probability is greater than 0.5, it is judged to be true At the same time, the discrimination dimension of the severity of the fault embryo pattern is set: the severity is judged according to the numerical value of the relevant features in the fault embryo pattern sample (such as features related to the equipment components). For example, if the characteristic value representing the temperature of the equipment component exceeds 50% of the normal range, the severity of the fault embryo pattern is considered to be high; C3, the learning rate of adversarial training is set to 0.001, and the learning rate is adjusted according to the scale of the semantic fusion feature data set. If the scale of the semantic fusion feature data set is small (such as less than 1000 samples), the learning rate is increased to 0.005. If the scale of the semantic fusion feature data set is large, the learning rate is increased to 0.006. If the data set is large (e.g., greater than 5000 samples), the learning rate can be reduced to 0.0005; the batch size of the training is set to 32, and the batch size can be adjusted according to the hardware resources (e.g., GPU memory). If the memory is large, the batch size is increased to 64, and if the memory is small, it can be reduced to 16; the number of iterations is set to 1000. During the training process, the model is evaluated every 100 iterations. The loss function values ​​of the generator and the discriminator are used to determine whether the model has converged. If the loss function value changes by less than 0.01 in three consecutive evaluations, the model is considered to have converged and the training is stopped early; Along the relationship paths (causal relationships and association relationships) between entities in the knowledge graph, the origin and development context of the fault embryo are traced; the origin of the fault embryo can be traced back to multiple aspects, including the origin of the equipment itself, the origin of the external environment, and the origin of human operation; for example, for the origin of the equipment itself, if through in-depth semantic association analysis, it is found that the engine pistons produced in the current batch have defects in the manufacturing process, then it indicates that there are defects in the initial manufacturing of the equipment components, which constitutes a possible origin of the fault embryo. On the other hand, if the data in the semantic space shows that the gear set of the equipment has severe tooth surface wear after a long period of high-load operation, it indicates that the long-term operation of the equipment has caused wear and aging of the equipment components, which constitutes a possible origin of the fault embryo; for the external environment origin, the equipment is in a harsh macro-geographic environment of high humidity and high salt fog in coastal areas, which accelerates the corrosion process of the metal components of the equipment, thereby triggering the origin of the fault. In addition, through semantic association analysis, it is revealed that when the power grid voltage in the area where the equipment is running is unstable, it will cause an impact on the electrical system of the equipment, which will become another external environment origin of the fault embryo; In terms of human operation origin, if the operation records in the semantic space show that the operator did not start the equipment in the prescribed order, or made incorrect equipment parameter configuration during the operation of the equipment, it directly led to the generation of fault embryos; the development context of the fault embryo includes the initial evolution of the fault, the expansion of the impact range, and the changes in the equipment operation level; for example, for the initial evolution of the fault, the vibration parameters and temperature of the relevant components of the equipment in the semantic space are slightly abnormal, which indicates that the wear of the equipment components is in the initial stage, showing slight surface wear, and gradually intensifies to form a clear development context; for the expansion of the impact range, if the correlation between the components in the semantic space reflects that the fault embryo spreads from a single component to other related components, such as the engine piston failure affects the connecting rod force, and then affects the crankshaft operation, then the fault impact range is gradually expanding, forming a coherent development context; for the changes in the equipment operation level, if the fluctuation of the overall equipment operation indicators in the semantic space reflects that the fault embryo affects multiple components, resulting in a decrease in the equipment operation performance, such as reduced output power, increased energy consumption, etc., then it shows the development context of the fault from local to overall and gradual expansion; In the traceability analysis, the following aspects are specifically included: D1. If there are multiple possible relationship paths, the relationship path for priority tracing is selected according to the credibility of the relationship path: the credibility of the relationship path is determined by calculating the product of the strength of the relationship between entities on the relationship path. If there are three entity relationships on a relationship path, and the strength values ​​of the three entity relationships are 0.8, 0.7 and 0.6 respectively, then the credibility of the relationship path is 0.8x0.7x0.6=0.336; sort the relationship paths in descending order according to their credibility, and select the path for priority tracing; D2. Quantify the origin of the fault: For the origin of the equipment itself, quantify the degree of equipment aging based on the equipment usage time and equipment service life. Set the equipment service life to T, the usage time to t, and the degree of equipment aging is expressed as t / T; For the origin of the external environment, quantitative analysis is performed based on the degree of deviation between the current environment and the normal working environment range of the equipment. If the normal working environment range of the equipment is [h1,h2] and the current environment is h0, then the degree of environmental impact is expressed as |h0 -(h1+h2) / 2| / (h2-h1); For the origin of human operation, a quantitative analysis is performed based on the operation frequency and operation correctness. When the operation frequency is higher than 50% of the normal operation frequency, the operation frequency impact factor is 0.5. When the operation error rate is p, the operation correctness impact factor is 1-p. The comprehensive impact of the origin of human operation is expressed as: the product of the operation frequency impact factor and the operation correctness impact factor; D3. Quantitative description of the fault development context: For the initial evolution stage of the fault, a quantitative evaluation is performed based on the characteristic change speed of the fault embryo pattern, and the rate of change of the characteristic value of the fault embryo pattern per unit time is calculated to measure the fault. , if the rate of change is larger, it indicates that the initial evolution process of the fault is faster; for the expansion of the impact range, the number of equipment components or equipment functional modules involved is used as a quantitative indicator. If the initial fault only affects one equipment component, and as the fault continues to develop, its impact range expands to three components, then the expansion of the impact range is obtained by calculating (3-1) / 1, that is, it is expanded by 2 times; for changes in the equipment operation level, a quantitative analysis is performed based on the changes in the equipment operation parameters (i.e., efficiency, power). If the initial power of the equipment is g1, and as the fault evolves, its power drops to g2, then the degree of change in the equipment operation level is expressed as (g1-g2) / g1; In summary, after identifying the potential fault embryonic mode, it is necessary to further analyze the interaction between the fault embryonic mode and the entire equipment operation architecture in order to more accurately predict the occurrence of faults.

[0019] It should be further explained that, in the specific implementation process, based on the identified fault embryo pattern and its traceability analysis results, the process of constructing the implicit order model is as follows: Quantify each element in the equipment operation architecture, and use nonlinear dynamic equations to analyze the dynamic change relationship between elements; select corresponding quantitative indicators for different types of elements: for equipment components, quantify according to their performance indicators; for software management programs, quantify according to their code complexity, execution efficiency and other indicators; for operators, quantify according to operation proficiency, operation frequency and other indicators; for operating environment, quantify according to environmental parameters; after completing the quantification of elements, select the relevant nonlinear dynamic equation type according to the nature and interaction relationship of the elements. When there are different relationships such as competition and cooperation, energy transmission, and information interaction between elements, different types of nonlinear dynamic equations are corresponding respectively; for example, when different equipment components compete and work together for resources, the deformation of the Lotka-Volterra equation is considered; if it is an energy transmission or information interaction relationship, the Hamiltonian equation or the dynamic equation related to information entropy is referred to; During the construction process, the information of each element in the equipment operation architecture is added to the parameter setting of the hidden order model (that is, when the fault embryo pattern shows that the failure probability of any element in the equipment operation architecture is high, the element information is converted into the coefficient of the nonlinear dynamic equation), and the relevant quantitative values ​​assigned by the fault embryo pattern and its traceability analysis results are combined to associate it with the parameters of the hidden order model; Determine the normal state or potential fault state of the equipment operation architecture through the hidden order model: define the threshold of the normal state; compare and analyze the output value of the hidden order model with the defined threshold. If the output value of the hidden order model is within the normal state threshold range, that is, the output value is greater than or equal to the lower limit threshold of the normal state and less than or equal to the upper limit threshold of the normal state, then it is determined that the current state of the equipment operation architecture meets the normal operation standard and is considered to be in a normal state. At this time, the equipment is likely to be able to operate stably and reliably, and various performance indicators are within the expected reasonable range; if the output value of the hidden order model exceeds the normal state threshold range, that is, the output value is less than the lower limit threshold of the normal state or greater than the upper limit threshold of the normal state, it indicates that there may be potential problems in the equipment operation architecture, which is considered to be a potential fault state. Once this happens, further in-depth analysis is required; At the same time, the order structure hidden inside the equipment operation architecture that causes the transition from the normal state to the fault state is captured: the hidden order structure is abstracted from the element relationship of the equipment operation architecture, including the energy flow order structure, software interaction order structure, environmental action order structure and human-machine operation order structure: for the energy transmission between equipment components, a mathematical model is established based on the energy flow in the circuit and the energy transfer in the mechanical transmission, and then the implicit order structure is found; for the interaction between the software management program and the equipment components, the hidden order structure is analyzed from the aspects of instruction flow and data flow; for the impact of the operating environment on the equipment components, the hidden order structure between the operating environment and the equipment components is revealed by monitoring the environmental parameters and analyzing their effects on the various components of the equipment; for the interaction between the operator and the equipment components, the hidden interactive order structure between the operator and the equipment components is revealed by analyzing the process of the operator receiving the equipment status information and making control inputs accordingly; Specifically, after establishing the implicit order model that describes the dynamic transformation behavior of the equipment operation architecture, the model needs further verification and optimization to ensure that it can accurately predict the occurrence of failures.

[0020] It should be further explained that, in the specific implementation process, based on the latent order model, the association rule recognition algorithm is used to further find the association rules between the identified fault embryo pattern and the semantic fusion feature data set; according to the analyzed association rules and the status of the current equipment operation architecture, an inference prediction engine based on latent order association is constructed, and the process of inference prediction is as follows: Obtain the identified fault embryo pattern and semantic fusion feature dataset: the fault embryo pattern is stored in the form of a vector, each element in the fault embryo pattern vector represents its corresponding fault feature identifier and quantified value, and the semantic fusion feature dataset is stored in the form of a data table, each row represents a data sample, and each column represents a fusion feature; The Apriori algorithm is used to mine frequent itemsets in the integrated fault embryo pattern vector and semantic fusion feature data set, and the minimum support and confidence thresholds are set to identify feature combinations that frequently and simultaneously appear before the fault occurs. These feature combinations further constitute and filter out the association rules for the occurrence of faults: the fault embryo pattern vector is integrated with the semantic fusion feature data set to ensure that each element in the fault embryo pattern vector matches and corresponds to the corresponding fusion feature in the semantic fusion feature data set; the integrated data set is converted into the form of a transactional database processed by the Apriori algorithm, where each transaction represents a set of multiple features, and these features come from the fault embryo pattern vector and the semantic fusion feature data set; the minimum support threshold is set to filter frequent itemsets, where the support represents the frequency of an item set appearing in all transactions; the maximum support threshold is set. A small confidence threshold is used to filter association rules; using the iterative process of the Apriori algorithm, starting from a single item, a frequent item set containing more items is gradually generated, where the generated frequent item set represents the feature combination that often appears at the same time before the fault occurs; in each iteration, a candidate item set is generated based on the frequent item set generated in the previous round, and the support of each candidate item set is calculated; the candidate item set with a support not less than the minimum support threshold is retained as a new frequent item set for the next round of iteration to generate more candidate item sets; the iterative process is repeated until no new frequent item sets can be generated; for each frequent item set, all possible association rules are generated, and the support and confidence of each association rule are calculated; the association rules with a support and confidence not less than the set threshold are retained as the final fault occurrence association rules; the association relationship between the feature combination represented by the generated association rule and the fault occurrence is expressed; Combine the acquired association rules with the status of the current equipment operation architecture to build an inference prediction engine based on implicit order association; use the inference prediction engine to make inference predictions about future faults based on the status of the current equipment operation architecture and the association rules: if the key monitoring parameters of the equipment operation architecture are found to meet the association rules, the inference prediction engine predicts the probability, time and type of faults; the key monitoring parameters are important indicators used in the equipment operation architecture to monitor and evaluate the equipment status, and are obtained in real time through the sensor network; according to the feature combination in the association rules and the historical fault data, use machine learning algorithms, time series prediction models and classification algorithms to make comprehensive predictions about the probability, time and type of faults; At the same time, the Bayesian reasoning method is used to correct the prediction results to improve the accuracy and reliability of the prediction: the prior probability is obtained based on the statistics of historical fault data, and the prior probability is classified; the likelihood function is determined based on the relationship between the currently monitored key monitoring parameters and the occurrence of faults, and the posterior probability is calculated according to the Bayesian formula, and the posterior probability is converted into a correction value for the probability, time and type of fault occurrence.

[0021] It should be further explained that, in the specific implementation process, the normal state of the equipment operation architecture is regarded as itself, and the fault state is regarded as not itself; combined with the inference prediction results, the process of establishing an adaptive defense mechanism is as follows: Obtain the inference prediction results of the implicit order model and set the self-tolerance threshold set; when the probability value of the detected fault deviates from the range of the set self-tolerance threshold set, start the adaptive defense mechanism to identify and process the fault state of the equipment operation architecture; generate an adaptive control strategy according to the type and severity of the fault; the control strategy includes measures such as maintenance and replacement of equipment components, dynamic adjustment of equipment operation parameters, optimization of work processes and reallocation of resources; for example, when it is predicted that a key component is about to fail, the equipment's operating mode is automatically adjusted to reduce the load of the component, and maintenance personnel are arranged to prepare to replace the component in advance to ensure that the equipment can smoothly transition when a fault occurs, reducing downtime and losses.

[0022] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0023] It should be understood that determining B based on A does not mean determining B only based on A. B can also be determined based on A and / or other information.

[0024] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0025] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A fault identification method based on an intelligent model, characterized in that: S1. Use multimodal high-precision sensing elements to build a multi-dimensional field sensing structure to collect comprehensive status information of the device and its environment; formulate corresponding fusion coding strategies for each acquired field data to obtain fused multi-dimensional tensor data; S2. Use natural language processing and knowledge graph technology to build a semantic space and generate a semantic fusion feature dataset; In the constructed semantic space, the generative adversarial network technology is used to mine the fault embryo pattern; if the fault embryo pattern is identified, the entity relationship network in the semantic space is used for traceability analysis; S3. Construct a hidden order model based on the identified fault embryo pattern and its traceability analysis results; S4. Based on the latent order model, the association rule recognition algorithm is used to further find the association rules between the identified fault embryo pattern and the semantic fusion feature data set; according to the analyzed association rules and the status of the current equipment operation architecture, an inference prediction engine based on latent order association is constructed, and inference prediction is performed; S5. Treat the normal state of the equipment operation architecture as itself and the fault state as not itself; combine the inference prediction results to establish an adaptive defense mechanism.

2. The fault identification method based on intelligent model according to claim 1, characterized in that: In S1, the multidimensional field specifically includes microscopic electromagnetic fields, weak gravitational fields, and macroscopic geographical environments and spatial positions; the multimodal high-precision sensing elements specifically include near-field optical sensors, gravitational wave sensors, and satellite positioning and geographic information systems.

3. The fault identification method based on intelligent model according to claim 1, characterized in that: The S1 process includes: Use near-field optical sensors to capture changes in the microscopic electromagnetic field distribution on the device surface; use gravitational wave sensors to monitor the weak gravitational field fluctuations caused by changes in mass distribution of large device clusters; combine satellite positioning with geographic information systems to obtain the macroscopic geographical environment and spatial location information of the device; obtain microscopic electromagnetic field data on the device surface, weak gravitational field fluctuation data of large device clusters, and macroscopic geographical environment and spatial location data of the device; Collect various text description data during equipment operation; Map the microscopic electromagnetic field data to the complex space, and use the real and imaginary parts of the complex numbers to represent the coupling relationship between different physical quantities; for the weak gravitational field fluctuation data, use the wavelet transform algorithm to encode and process it to form a data vector with high-dimensional characteristics; for the macroscopic geographic space data, convert the geographic information into a topological feature vector through topological mapping; For text description data, natural language processing technology is used to convert it into word vector representation; The tensor product operation is used to merge the feature vectors obtained by these different encoding methods into a unified multi-dimensional tensor data.

4. The fault identification method based on intelligent model according to claim 1, characterized in that: The S2 includes: Extract equipment-related entities and relationships between entities from the fused multi-dimensional tensor data to build a fault knowledge graph; The various features in the fused multidimensional tensor data are semantically associated with the entities constructed in the knowledge graph, and semantic labels are assigned to each data feature, so as to form a semantic space that integrates multi-source data and domain knowledge, and generate a semantic fusion feature data set; among which, the various features in the fused multidimensional tensor data include microscopic electromagnetic field features, weak gravitational field features, macroscopic geographic space features, and text description data features; The generator is responsible for generating possible fault embryo pattern samples, and the discriminator distinguishes whether the fault embryo pattern samples are real fault embryos. When generating fault embryo pattern samples, the generator comprehensively uses the equipment operation status reflected by each field information in the generated semantic fusion feature data set. Through continuous adversarial training, the generator is used to learn the potential fault embryo feature patterns in the semantic fusion feature dataset; Along the relationship path between entities in the knowledge graph, the origin and development context of the fault embryo are traced back; the origin of the fault embryo can be traced back to multiple aspects, including the origin of the equipment itself, the origin of the external environment, and the origin of human operation; the development context of the fault embryo includes the initial evolution of the fault, the expansion of the impact range, and the changes in the equipment operation level; Through calculation and analysis, relevant quantitative values ​​of traceability analysis are obtained.

5. The fault identification method based on intelligent model according to claim 1, characterized in that: In S3, the implicit order model regards the entire device operation architecture as a dynamic complex system, in which there are complex and hidden interactions between the various elements; wherein the device operation architecture refers to a comprehensive device operation system composed of device components, software management programs, operators and operating environment; the elements of the device operation architecture include device components, software management programs, operators and operating environment.

6. The fault identification method based on intelligent model according to claim 1, characterized in that: The S3 includes: Quantify each element in the equipment operation architecture and use nonlinear dynamic equations to analyze the dynamic relationship between the elements; During the construction process, the information of each element in the equipment operation architecture is added to the parameter setting of the hidden order model, and the relevant quantitative values ​​assigned by the fault embryo mode and its traceability analysis results are combined to associate it with the parameters of the hidden order model; Determine the normal state or potential fault state of the equipment operation architecture through the hidden order model: define the threshold of the normal state; compare and analyze the output value of the hidden order model with the defined threshold. If the output value of the hidden order model is within the normal state threshold range, that is, the output value is greater than or equal to the lower limit threshold of the normal state and less than or equal to the upper limit threshold of the normal state, then it is determined that the current state of the equipment operation architecture meets the normal operation standard and is considered to be in a normal state; if the output value of the hidden order model exceeds the normal state threshold range, that is, the output value is less than the lower limit threshold of the normal state or greater than the upper limit threshold of the normal state, it indicates that there is a problem with the equipment operation architecture and it is considered to be in a potential fault state; At the same time, it captures the order structure hidden within the equipment operation architecture that causes the transition from normal state to fault state.

7. The fault identification method based on intelligent model according to claim 1, characterized in that: The S4 includes: Obtain the identified fault embryo pattern and semantic fusion feature dataset: the fault embryo pattern is stored in the form of a vector, each element in the fault embryo pattern vector represents its corresponding fault feature identifier and quantified value, and the semantic fusion feature dataset is stored in the form of a data table, each row represents a data sample, and each column represents a fusion feature; The Apriori algorithm is used to mine frequent item sets in the integrated fault embryo pattern vector and semantic fusion feature data set, and the minimum support and confidence thresholds are set to identify the feature combinations that frequently and simultaneously appear before the fault occurs. These feature combinations further constitute and filter the association rules of the fault occurrence. Combine the acquired association rules with the state of the current equipment operation architecture to build an inference prediction engine based on implicit order association; use the inference prediction engine to make inference predictions about future faults based on the state of the current equipment operation architecture and the association rules: if the key monitoring parameters of the equipment operation architecture are found to meet the association rules, the inference prediction engine predicts the probability, time and type of fault occurrence; the key monitoring parameters are important indicators used in the equipment operation architecture to monitor and evaluate the equipment status, and are obtained in real time through the sensor network; At the same time, the Bayesian reasoning method is used to correct the prediction results.

8. The fault identification method based on intelligent model according to claim 7 is characterized in that: The Apriori algorithm is used to mine frequent item sets in the integrated fault embryo pattern vector and semantic fusion feature data set, and the minimum support and confidence thresholds are set to identify feature combinations that frequently and simultaneously appear before the fault occurs, including: The fault embryo pattern vector is integrated with the semantic fusion feature data set to ensure that each element in the fault embryo pattern vector matches and corresponds to the corresponding fusion feature in the semantic fusion feature data set; the integrated data set is converted into a transactional database processed by the Apriori algorithm, where each transaction represents a set of multiple features, and these features come from the fault embryo pattern vector and the semantic fusion feature data set; a minimum support threshold is set to filter frequent item sets, where the support represents the frequency of an item set appearing in all transactions; a minimum confidence threshold is set to filter association rules; and the iterative process of the Apriori algorithm is used to start from a single item and gradually generate frequent itemsets containing more items. Item sets, where the generated frequent item sets represent feature combinations that often appear at the same time before a fault occurs; in each iteration, candidate item sets are generated based on the frequent item sets generated in the previous round, and the support of each candidate item set is calculated; candidate item sets whose support is not less than the minimum support threshold are retained as new frequent item sets for the next round of iteration to generate more candidate item sets; the iterative process is repeated until new frequent item sets can no longer be generated; for each frequent item set, all possible association rules are generated, and the support and confidence of each association rule are calculated; association rules whose support and confidence are not less than the set threshold are retained as the final fault occurrence association rules; the association relationship between the feature combination represented by the generated association rules and the fault occurrence is expressed.

9. The fault identification method based on intelligent model according to claim 1, characterized in that: The S5 includes: Obtain the inference prediction results of the implicit order model and set the self-tolerance threshold set; when the probability value of the detected fault deviates from the range of the set self-tolerance threshold set, activate the adaptive defense mechanism to identify and process the fault state of the equipment operation architecture; generate an adaptive control strategy based on the type and severity of the fault.

Images