Disclosure of Invention
The invention aims to solve the problems in the background technology, and provides a fault identification method based on an intelligent model.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
The fault identification method based on the intelligent model comprises the following steps:
S1, constructing a multidimensional field sensing structure by utilizing a multi-mode high-precision sensing element, and acquiring comprehensive state information of equipment and an environment thereof;
S2, constructing a semantic space by utilizing natural language processing and knowledge graph technology, and generating a semantic fusion characteristic data set; in the constructed semantic space, the fault embryo mode is mined by using a generation countermeasure network technology, and if the fault embryo mode is identified, the entity relationship network in the semantic space is utilized for tracing analysis;
S3, constructing a hidden order model based on the identified failure embryo mode and a traceability analysis result thereof;
S4, based on the hidden order model, using an association rule recognition algorithm to further find an association rule between the identified fault embryo mode and the semanteme fusion characteristic data set, constructing an inference prediction engine based on the hidden order association according to the association rule obtained by analysis and the state of the current equipment operation architecture, and performing inference prediction;
S5, regarding the normal state of the equipment operation architecture as self and regarding the fault state as non-self, and establishing a self-adaptive defense mechanism by combining an reasoning prediction result.
Further, the multi-dimensional field comprises a microscopic electromagnetic field, a weak gravitational field, a macroscopic geographic environment and a spatial position, and the multi-mode high-precision sensing element comprises a near-field optical sensor, a gravitational wave sensor and a satellite positioning and geographic information system.
Further, the process of S1 includes:
Capturing microscopic electromagnetic field distribution change of the surface of the equipment by adopting a near-field optical sensor, monitoring weak gravitational field fluctuation of a large-scale equipment cluster caused by mass distribution change by utilizing a gravitational wave sensor, acquiring macroscopic geographic environment and spatial position information of the equipment by combining a satellite positioning and geographic information system, acquiring microscopic electromagnetic field data of the surface of the equipment, weak gravitational field fluctuation data of the large-scale equipment cluster and macroscopic geographic environment and spatial position data of the equipment;
collecting various text description data in the running process of equipment;
The method comprises the steps of mapping microcosmic electromagnetic field data to a complex space, representing coupling relations between different physical quantities by utilizing real parts and imaginary parts of the complex, carrying out encoding processing on weak gravitational field fluctuation data by adopting a wavelet transformation algorithm to form a data vector with high-dimensional characteristics, and converting geographic information into a topological characteristic vector by topological mapping on macroscopic geographic space data;
For text description data, converting the text description data into word vector representation by using a natural language processing technology;
And fusing the eigenvectors obtained by the different coding modes into unified multidimensional tensor data by using tensor product operation.
Further, S2 includes:
Extracting the entity related to the equipment and the relation between the entities from the fused multidimensional tensor data, and constructing a fault knowledge graph;
Semantic association operation is carried out on various features in the multi-dimensional tensor data after fusion and constructed entities in the knowledge graph, semantic labels are given to each data feature, so that semantic space for fusing multi-source data and domain knowledge is formed, and a semantic fusion feature data set is generated;
The generator is responsible for generating a possible fault embryo mode sample, and the discriminator distinguishes whether the fault embryo mode sample is a real fault embryo, wherein when the generator generates the fault embryo mode sample, the generator comprehensively utilizes the running states of the equipment reflected by the field information in the generated semantic fusion characteristic data set;
learning a potential failure embryo characteristic mode in the semantically fused characteristic data set by using a generator through continuous countermeasure training;
Tracing the origins and development venues of the fault embryos along the relation paths among the entities in the knowledge graph, wherein the origins of the fault embryos are traced to a plurality of aspects including the origins of equipment, external environment origins and artificial operation origins;
and obtaining a relevant quantized value of the traceability analysis through calculation and analysis.
Furthermore, the hidden order model regards the whole equipment operation architecture as a dynamic compound system, wherein complex and hidden interaction relations exist among elements, the equipment operation architecture refers to a comprehensive equipment operation system formed by equipment components, a software management program, an operator and an operation environment, and the elements of the equipment operation architecture cover the equipment components, the software management program, the operator and the operation environment.
Further, S3 includes:
quantifying each element in the equipment operation architecture, and analyzing the dynamic change relation among the elements by utilizing a nonlinear dynamic equation;
In the construction process, adding element information in the equipment operation architecture into parameter setting of a hidden order model, and associating the parameter setting with a hidden order model parameter by combining a fault embryo mode and a related quantitative value endowed by a traceable analysis result thereof;
Judging the normal state or potential fault state of the equipment operation framework by the hidden order model, namely, defining a threshold value of the normal state, comparing the output value of the hidden order model with the defined threshold value, and judging that the current state of the equipment operation framework accords with the standard of normal operation if the output value of the hidden order model is within the range of the threshold value of the normal state, namely, the output value is greater than or equal to the lower limit threshold value of the normal state and less than or equal to the upper limit threshold value of the normal state;
at the same time, the capture device runs an internally hidden orderly structure of the architecture that results in a transition from a normal state to a failed state.
Further, S4 includes:
The method comprises the steps of obtaining a data set of identified fault embryo modes and semantic fusion characteristics, wherein the fault embryo modes are stored in a vector form, each element in the fault embryo mode vector represents a corresponding fault characteristic identifier and a quantization value, the semantic fusion characteristic data set is stored in a data table form, each row represents a data sample, and each column represents a fusion characteristic;
the method comprises the steps of mining frequent item sets in integrated fault embryo mode vectors and semantic fusion feature data sets through an Apriori algorithm, setting minimum support and confidence threshold values to identify feature combinations which occur frequently and simultaneously before faults occur, and further forming and screening association rules of the faults;
The method comprises the steps of acquiring a critical monitoring parameter of a device operation architecture, combining the acquired association rule with the state of the current device operation architecture to construct an inference prediction engine based on hidden order association, and carrying out inference prediction on the occurrence of future faults by utilizing the inference prediction engine according to the state of the current device operation architecture and the association rule, wherein the inference prediction engine predicts the occurrence probability, time and type of the faults if monitoring that the critical monitoring parameter of the device operation architecture meets the association rule;
and meanwhile, correcting the prediction result by adopting a Bayesian inference method.
Further, mining frequent item sets in the integrated fault embryo mode vector and semantical fusion feature data set through an Apriori algorithm, and setting a minimum support and a confidence threshold to identify feature combinations which occur frequently and simultaneously before a fault occurs, wherein the feature combinations comprise:
The method comprises the steps of integrating a fault embryo mode vector with a semantical fusion feature data set, ensuring that each element in the fault embryo mode vector is matched and corresponds to a corresponding fusion feature in the semantical fusion feature data set, converting the integrated data set into a transactional database form processed by an Apriori algorithm, wherein each transaction represents a set containing a plurality of features, the features come from the fault embryo mode vector and the semantical fusion feature data set, setting a minimum support threshold value for screening frequent item sets, wherein the support value represents the occurrence frequency of one item set in all transactions, setting a minimum confidence threshold value for screening association rules, starting from a single item by using an iterative process of the Apriori algorithm, gradually generating frequent item sets containing more items, wherein the generated frequent item sets represent feature combinations which frequently occur simultaneously before the fault occurs, generating a candidate item set according to the frequent item set generated in the previous round in each iteration, calculating the support of each candidate item set, reserving the candidate item set with the support value not lower than the minimum support threshold value as a new frequent item set, setting a minimum support threshold value for screening frequent item sets, setting a confidence value for filtering the frequent item sets, repeatedly generating the candidate item sets, and repeatedly carrying out the iterative process until the association rules can not occur until the association rules are combined with the confidence rule and the association rules are set is generated, and the association rule is not capable of being set to be generated.
Further, S5 includes:
The method comprises the steps of obtaining an inference prediction result of a hidden order model, setting a self-tolerance threshold set, starting an adaptive defense mechanism to identify and process fault states of a device operation framework when a probability value of a detected fault deviates from a range of the set self-tolerance threshold set, and generating an adaptive regulation strategy according to the type and the severity of the fault.
The method has the advantages that a multi-dimensional field sensing structure is built through the multi-mode high-precision sensing element, comprehensive monitoring of equipment and environments of the equipment is achieved, accuracy and timeliness of fault identification are improved, natural language processing and knowledge graph technology is utilized, semantic space is built, fault embryo modes are mined, early detection and traceability analysis of faults are achieved, potential faults can be found and processed in time, association rules between the faults and features can be further mined through building a hidden order model and an inference prediction engine, accuracy of fault prediction is improved, a regulation strategy can be generated according to fault types and severity through introduction of a self-adaptive defense mechanism, and safety and stability of equipment operation are effectively guaranteed.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, the fault recognition method based on the intelligent model includes:
S1, constructing a multidimensional field sensing structure by utilizing a multi-mode high-precision sensing element, and acquiring comprehensive state information of equipment and an environment thereof;
The multi-mode high-precision sensing element comprises a near-field optical sensor (used for capturing microscopic electromagnetic field distribution change), a gravitational wave sensor (used for monitoring weak gravitational field fluctuation) and a satellite positioning and geographic information system (used for acquiring macroscopic geographic environment and spatial position information);
s2, constructing a semantic space by utilizing natural language processing and knowledge graph technology, and generating a semantic fusion characteristic data set, mining a fault embryo mode by utilizing a generation countermeasure network technology in the constructed semantic space, and performing traceability analysis by utilizing an entity relation network in the semantic space if the fault embryo mode is identified so as to understand the origin and development venation (possible reasons and paths) of fault occurrence;
S3, constructing a hidden order model based on the identified failure embryo mode and a traceability analysis result thereof, wherein the model regards the whole equipment operation architecture as a dynamic compound system, and complex and hidden interaction relations exist among all elements;
The equipment operation framework refers to a comprehensive equipment operation system formed by equipment components, a software management program, an operator and an operation environment, for example, for an automatic production line equipment, the equipment operation framework comprises mechanical devices of the production line, program software for controlling the operation of the production line, operators, factory building environments where the production line is located and other elements related to the operation of the equipment, wherein the elements of the equipment operation framework cover equipment components (such as motors, sensors, valves and the like), the software management program, the operators and the operation environment (such as temperature and humidity, electromagnetic interference and the like), and the interaction and the mutual influence of the elements jointly determine the operation state of the equipment;
S4, based on the hidden order model, an association rule recognition algorithm is used for further searching association rules between the identified fault embryo mode and the semanteme fusion characteristic data set, an inference prediction engine based on the hidden order association is constructed according to the association rules obtained by analysis and the state of the current equipment operation architecture, and inference prediction is performed to realize accurate prediction of future fault occurrence;
S5, regarding the normal state of the equipment operation architecture as self and regarding the fault state as non-self, and establishing a self-adaptive defense mechanism by combining an reasoning prediction result;
wherein, by reference to self/non-recognized (i.e. the organism can distinguish self cells from external invading pathogens) in the biological immune system, an adaptive defense mechanism is established, and the defense mechanism can automatically adjust the operation parameters of the equipment or take other preventive measures according to the reasoning prediction result so as to avoid or mitigate the influence of faults.
It should be further described that, in the specific implementation process, the multi-mode high-precision sensing element is utilized to construct a multi-dimensional field sensing structure, and the comprehensive state information of the equipment and the environment is acquired:
Selecting a near field optical sensor based on a scanning near field optical microscope principle, arranging the sensors on the surface of the equipment in a uniform distribution mode, calibrating the sensors by using a standard calibration block, wherein the sensor distribution density is 1 sensor per square centimeter, and the calibration accuracy reaches 0.1 microvolts per meter so as to ensure that accurate microscopic electromagnetic field distribution change is captured; the method comprises the steps of monitoring weak gravitational field fluctuation caused by mass distribution change of a large-scale equipment cluster by utilizing a gravitational wave sensor, adjusting the sensitivity range of the gravitational wave sensor according to the mass and distribution characteristics of the large-scale equipment cluster, eliminating the influence of other interference factors (such as the small fluctuation of the gravitational field of the earth, the interference of the gravitational force of other nearby objects and the like) in the environment to the weak gravitational field fluctuation by arranging a shielding device and adopting a differential measurement method, acquiring macroscopic geographic environment and spatial position information of equipment by combining satellite positioning and a geographic information system, wherein the data updating frequency in the geographic information system is once per hour by adopting a GPS and Beidou multi-system fusion satellite positioning system, and acquiring microscopic electromagnetic field data of the equipment surface, the weak gravitational field fluctuation data of the large-scale equipment cluster and the macroscopic geographic environment and spatial position data of the equipment, wherein the data come from different fields and contain detailed indication of the running state of the equipment and comprehensively reflect the comprehensive state of the equipment and the surrounding environment;
various text description data including log files, operation flows, maintenance records, fault reports and the like in the running process of the equipment are collected and used for reflecting the running state background information of the equipment;
Mapping microscopic electromagnetic field data into complex space, using real part and imaginary part of complex number to represent coupling relation between different physical quantities by using non-linear mapping functionMapping microscopic electromagnetic field data to complex space, encoding weak gravitational field fluctuation data by using wavelet transformation algorithm to form data vector with high-dimensional characteristic, converting geographic information into topological characteristic vector by topology mapping for macroscopic geographic space data, mapping based on net topological structure in graph theory, taking geographic coordinates in geographic information as nodes, connection relation between geographic areas as edges, node attribute as coordinate value, edge attribute as connection weight, determining weight according to distance;
The text description data is converted into Word vector representation by using a natural language processing technology, wherein Word2Vec pre-training models are adopted to process the text description data, and specific abbreviations related to equipment are standardized in the processing process;
The feature vectors obtained by the different coding modes are fused into unified multidimensional tensor data by tensor operation, and key information of each field data is comprehensively reserved, wherein the weight setting principle in the tensor operation process is to set the importance degree of the running state of the equipment according to each feature vector, for example, the weight of microcosmic electromagnetic field data is 0.3, the weight of weak gravitational field fluctuation data is 0.2, the weight of macroscopic geospatial data is 0.2, the weight of text description data is 0.3, and the fused multidimensional tensor data is subjected to minimum-maximum normalization processing to normalize the data to a [0,1] interval.
It is further described that, in the specific implementation process, a semantic space is constructed by utilizing natural language processing and knowledge graph technology, and a semantically fused characteristic data set is generated, in the constructed semantic space, a fault embryo mode is mined by utilizing a generated countermeasure network technology, and if the fault embryo mode is identified, the process of tracing and analyzing by utilizing an entity relation network in the semantic space is as follows:
The method comprises the steps of extracting the related entity (equipment component, fault type and maintenance measure) of the equipment and the relation (causal relation or association relation) between the entities from the fused multidimensional tensor data to construct a fault knowledge graph, wherein the sources of the entity and the relation are wide, and the information presented by text description in the equipment operation data and the potential relation implied by other field information in the fused multidimensional tensor data are contained;
Specifically, A1, a method based on combination of rules and machine learning is adopted to extract equipment related entities from fused multi-dimensional tensor data, a predefined dictionary of equipment components, fault types and maintenance measures is created, for the equipment components, the dictionary comprises common part names of the equipment and synonyms thereof, a method based on entity identification models of convolutional neural networks is used to initially scan the fused multi-dimensional tensor data, the fused multi-dimensional tensor data is divided according to fixed length windows, for example, each window comprises 100 data points, wherein the size of a convolution kernel of the entity identification models based on the convolutional neural networks is set to be 3, the step size is 1, characteristics are extracted through a plurality of convolution layers and pooling layers, for the data part with matched items in the predefined dictionary, the part which is not matched but is identified as potential entity by the entity identification models is subjected to manual verification, A2, a method based on logic rules and data statistical analysis is adopted to determine the relation (causal relation or association relation) between the entities, for the relation, for example, if the temperature of a certain component of the equipment is increased (by a sensor) is set to be 3, the frequency of occurrence of the fault types is set to be 0, the fact that the fault types are calculated when the frequency of the fault types is set to be 0, and the occurrence of the fault types is calculated to be the certain, and the fault types are calculated to be the frequency to be the same, if the fault types are set to be the occurrence in the certain time is set to be the certain, and the fault types are set to be the same, meanwhile, according to the time stamp sequence in the multi-dimensional tensor data after fusion, whether causal relation tendency exists or not is primarily judged;
Semantic association operation is carried out on various features in the multi-dimensional tensor data after fusion and constructed entities in the knowledge graph, semantic labels are given to each data feature, so that semantic space for fusing multi-source data and domain knowledge is formed, and a semantic fusion feature data set is generated, wherein the various features in the multi-dimensional tensor data after fusion comprise microscopic electromagnetic field features, weak gravitational field features, macroscopic geographic space features and text description data features;
Specifically, B1, performing semantic association on various features in the fused multidimensional tensor data and constructed entities in the knowledge graph by adopting a semantic similarity calculation method, namely, using a pre-trained Word vector model (such as a Word2Vec model pre-trained by a device domain corpus) for special training in the device domain, calculating cosine similarity between each data feature and entity Word vectors in the knowledge graph, for example, for a feature vector representing the temperature of the device, calculating cosine similarity between each feature and entity Word vectors in the knowledge graph, giving a semantic tag according to the value of the semantic similarity, giving a strong association semantic tag if the cosine similarity is greater than 0.8, giving a medium association semantic tag if the cosine similarity is less than 0.5, and giving a weak association semantic tag if the cosine similarity is less than 0.8;
The generator is responsible for generating a possible fault embryo mode sample, and the discriminator distinguishes whether the fault embryo mode sample is a real fault embryo, wherein when the generator generates the fault embryo mode sample, the generator comprehensively utilizes the running states of the equipment reflected by the field information in the generated semantic fusion characteristic data set;
by constantly countertraining, learning potential failure embryo feature patterns in the semantically fused feature dataset with the generator, it is understood that these potential failure embryo feature patterns may exist before the failure has not been significantly manifested, as early signs of failure development;
Specifically, the generator adopts a multi-layer perceptron structure, wherein the number of input layer nodes is determined according to the dimension of the semantically fused characteristic dataset, for example, if the dataset has 100 characteristics, the number of the input layer nodes is 100; the hidden layer is set to be 3 layers, the number of nodes of the first layer is 200, a ReLU activation function is used, the number of nodes of the second layer is 150, a ReLU activation function is used, the number of nodes of the third layer is 100, and a tanh activation function is used; the number of output layer nodes is determined according to the coding dimension of a fault embryo mode sample, if the fault embryo mode sample is represented by a 50-dimensional vector, the output layer nodes are 50, C2, the discriminator adopts a probability-based discriminating method, the probability that the fault embryo mode sample belongs to a real fault embryo is calculated for the input fault embryo mode sample, the fault embryo mode sample is input into the discriminator formed by a multi-layer perceptron, the number of the input layer nodes of the discriminator is the same as the number of the output layer nodes of a generator (such as 50), the number of hidden layers is set to be 2, the number of each layer node is 80, a ReLU activation function is used, the number of the output layer nodes is 1, the sigmoid activation function is used, the output value represents the probability that the sample is a real fault embryo, when the probability is larger than 0.5, the real fault embryo is discriminated, meanwhile, the discriminating dimension of the severity of the fault embryo mode is set according to the value of relevant characteristics (such as characteristics related to equipment components) in the fault embryo mode sample is set, for example, if the characteristic values representing the temperature of the equipment components exceeds 50% of a normal range, the severity of the fault embryo mode is considered to be higher, C3, the contrast training is set to be 0.001, the learning data is adjusted according to the semantic fusion data scale, if the semantical fusion characteristic data set is smaller in scale (for example, less than 1000 samples), the learning rate is increased to 0.005, if the semantical fusion characteristic data set is larger in scale (for example, more than 5000 samples), the learning rate can be reduced to 0.0005, the batch size of training is set to 32, the batch size can be adjusted according to the condition of hardware resources (for example, GPU video memory), if the video memory is larger, the batch size is increased to 64, if the video memory is smaller, the iteration times can be reduced to 16;
Tracing the origin and development venation of the fault embryo along the relation path (causality and association relation) between the entities in the knowledge graph; wherein the origin of the faulty embryo is traced back to a number of aspects, including the origin of the device itself, the origin of the external environment and the origin of the manual operation; for example, if the engine pistons produced in the current batch are found to have defects in the manufacturing process through deep semantic association analysis, the engine pistons are predicted to have defects in the initial manufacturing of equipment components, so that one possible origin of a fault embryo is formed, on the other hand, if data in a semantic space shows that the gear set of the equipment is seriously worn out after long-time high-load operation, the equipment components are predicted to be worn out and aged due to long-term operation, so that the possible origin of the fault embryo is formed, the equipment is in a severe macroscopic geographical environment in a coastal area with high humidity and high salt mist for the external environment origin, the corrosion process of the metal components of the equipment is accelerated, the fault origin is further caused, in addition, the impact is caused to an electrical system of the equipment when the power grid voltage in the area where the equipment is operated is unstable is revealed through semantic association analysis, so that the other external environment of the fault embryo is formed, if an operation record in the semantic space shows that an operator does not start the equipment according to a specified sequence or performs wrong equipment parameter configuration during the equipment operation, the initial development of the fault embryo is directly caused, the initial development of the fault embryo comprises the initial development of the fault embryo, the initial development of the fault embryo has a slight development range and the fault is expanded, for example, the fault temperature range is abnormal, the fault condition occurs in the relevant conditions occur in the equipment is abnormal conditions such conditions, for the expansion of the influence range, if the incidence relation among all components in the semantic space shows that a fault embryo is diffused from a single component to other related components, such as an engine piston fault influences the connecting rod stress so as to influence the operation of a crankshaft, the fault influence range is gradually expanded to form a coherent development context;
d1, if a plurality of possible relationship paths exist, selecting a relationship path which is traced preferentially according to the reliability of the relationship path, wherein the reliability of the relationship path is determined by calculating the strength product of the relationship among entities on the relationship path, and if three entity relationships exist on one relationship path, the strength values of the three entity relationships are respectively 0.8, 0.7 and 0.6, the reliability of the relationship path is 0.8x0.7x0.6=0.336; the method comprises the steps of arranging in descending order according to the reliability of a relation path, selecting a path with priority tracing, quantifying the fault origin, namely quantifying the equipment aging degree according to the equipment service time and the equipment service life, setting the equipment service life as T and the equipment aging degree as T according to the equipment service time and the equipment service life, quantifying the equipment aging degree as T/T according to the deviation degree of the current environment and the normal working environment range of the equipment for external environment origin, quantifying the external environment origin according to the deviation degree of the current environment and the normal working environment range of the equipment, wherein the current environment is h0, the environment influence degree is represented as |h0- (h1+h2)/2|/(h 2-h 1), quantifying the artificial operation origin according to the operation frequency and the operation correctness, quantifying the operation frequency influence factor as 0.5 when the operation frequency is higher than 50% of the normal operation frequency, the operation correctness influence factor as p, synthesizing the artificial operation origin influence degree as the product of the operation frequency influence factor and the operation correctness influence factor, quantifying the initial development of the fault, quantifying the fault development according to the characteristic of the analysis of the initial development stage, the method comprises the steps of calculating the change rate measurement of a characteristic value of a fault embryo mode in unit time, if the change rate is larger, indicating that the initial evolution process of a fault is quicker, for the expansion of an influence range, adopting the number of related equipment components or the number of equipment functional modules as quantization indexes, if the initial fault only affects one equipment component, and the influence range is expanded to three components along with the continuous development of the fault, obtaining the expansion degree of the influence range, namely, expanding by 2 times through calculating (3-1)/1;
In summary, after identifying a potential faulty embryo pattern, it is desirable to further analyze the interaction between the faulty embryo pattern and the overall plant operating architecture in order to more accurately predict the occurrence of a fault.
It should be further noted that, in the specific implementation process, the process of constructing the hidden order model based on the identified failure embryo mode and the traceability analysis result thereof is as follows:
For different types of elements, corresponding quantization indexes are selected, wherein for equipment components, the equipment components are quantized according to performance indexes of the equipment components, software management programs are quantized according to indexes such as code complexity, execution efficiency and the like of the software management programs, operators are quantized according to operation proficiency, operation frequency and the like, operation environments are quantized according to environment parameters, after element quantization is completed, relevant nonlinear dynamical equation types are selected according to the properties and interaction relation of the elements, when different relations such as competition and cooperation exist between the elements, energy transmission, information interaction and the like, the different types of nonlinear dynamical equations respectively correspond, for example, the competition and cooperation work entropy of different equipment components on resources is considered, the deformation of Lotka-Volterra equations is considered, and the Hamiltonian equation or information-related dynamical equation is referred to if the relation of energy transmission or information interaction is the relation;
in the construction process, adding element information in the equipment operation architecture into parameter setting of a hidden order model (namely, converting the element information into a coefficient of a nonlinear dynamic equation when the fault probability of any element in the equipment operation architecture is high, and associating the element information with hidden order model parameters by combining a fault embryo mode and a related quantized value endowed by a traceability analysis result thereof;
The method comprises the steps of judging the normal state or potential fault state of an equipment operation framework through a hidden order model, comparing an output value of the hidden order model with a defined threshold value for analysis, judging that the current state of the equipment operation framework accords with the standard of normal operation if the output value of the hidden order model is within the range of the normal state threshold value, namely, the output value is greater than or equal to a lower limit threshold value of the normal state and less than or equal to an upper limit threshold value of the normal state, and judging that the equipment operation framework can stably and reliably operate in the normal state if the current state of the equipment operation framework accords with the standard of the normal operation, wherein each performance index is in an expected reasonable range;
The method comprises the steps of capturing an internal hidden order structure of a device operation framework, abstracting the hidden order structure from element relation of the device operation framework, wherein the internal hidden order structure comprises an energy flow order structure, a software interaction order structure, an environment action order structure and a man-machine operation order structure, establishing a mathematical model for energy transmission among device components according to energy flow in a circuit and energy transmission in mechanical transmission, further finding out the hidden order structure, analyzing the hidden order structure from the aspects of instruction flow and data flow for interaction relation between a software management program and the device components, revealing the hidden order structure between the operation environment and the device components by monitoring environment parameters and analyzing the effect of the operation environment on the device components, and revealing the hidden interaction order structure between the operator and the device components by analyzing the interaction of the operator and the device components and receiving device state information and controlling input according to the information;
In particular, after establishing a hidden order model describing the dynamic transition behavior of the device operating architecture, the model requires further verification and optimization to ensure that it can accurately predict the occurrence of faults.
It is further described that, in the specific implementation process, based on the hidden order model, an association rule recognition algorithm is applied to further find the association rule between the identified fault embryo mode and the semantically fused characteristic data set, and according to the association rule obtained by analysis and the state of the current equipment operation architecture, an inference prediction engine based on the hidden order association is constructed, and the inference prediction process is as follows:
The method comprises the steps of obtaining a data set of identified fault embryo modes and semantic fusion characteristics, wherein the fault embryo modes are stored in a vector form, each element in the fault embryo mode vector represents a corresponding fault characteristic identifier and a quantization value, the semantic fusion characteristic data set is stored in a data table form, each row represents a data sample, and each column represents a fusion characteristic;
The method comprises the steps of mining frequent item sets in an integrated fault embryo pattern vector and semantically fused feature data set through an Apriori algorithm, setting minimum support and confidence threshold values to identify feature combinations which occur frequently and simultaneously before faults occur, further constructing and screening association rules for faults occurring, integrating the fault embryo pattern vector and the semantically fused feature data set to ensure that each element in the fault embryo pattern vector is matched and corresponds to the corresponding fusion feature in the semantically fused feature data set, converting the integrated data set into a form of a transactional data base processed by the Apriori algorithm, each transaction represents a set containing a plurality of features, and the features come from the fault embryo pattern vector and the semantically fused feature data set, setting the minimum support threshold values to screen the frequency of occurrence of one item set in all the transactions, setting the minimum confidence threshold values to screen the association rules, gradually generating frequent item sets containing more items from a single item, wherein the generated frequent item sets represent the corresponding fusion feature sets in the semantically fused feature data set, generating a set before the fault embryo pattern vector and the semantically fused feature sets, generating a candidate item set repeatedly at the same time according to the minimum support threshold values, generating the candidate item sets repeatedly until the candidate item sets can not occur frequently in the iteration sets are generated repeatedly at the same time for each iteration threshold value, the method comprises the steps of calculating the support degree and the confidence degree of each association rule, reserving the association rule with the support degree and the confidence degree not lower than a set threshold as a final fault occurrence association rule, and utilizing the association relationship between the feature combination represented by the generated association rule and fault occurrence;
The method comprises the steps of acquiring a critical monitoring parameter of a device operation architecture, combining the acquired association rule with the state of the current device operation architecture to construct an inference prediction engine based on hidden order association, and carrying out inference prediction on the occurrence of future faults by utilizing the inference prediction engine according to the state of the current device operation architecture and the association rule, wherein the inference prediction engine predicts the occurrence probability, time and type of the faults if monitoring that the critical monitoring parameter of the device operation architecture meets the association rule, wherein the critical monitoring parameter is an important index for monitoring and evaluating the state of the device in the device operation architecture and is acquired in real time through a sensor network;
and meanwhile, correcting the prediction result by adopting a Bayesian inference method to improve the accuracy and reliability of prediction, namely obtaining prior probability based on historical fault data statistics, classifying the prior probability, determining a likelihood function based on the relation between the currently monitored key monitoring parameters and fault occurrence, carrying out posterior probability calculation according to a Bayesian formula, and converting the posterior probability into correction values for the fault occurrence probability, time and fault type.
It should be further described that, in the specific implementation process, the normal state of the device operation architecture is regarded as self, the fault state is regarded as non-self, and the process of establishing the self-adaptive defense mechanism by combining the reasoning prediction result is as follows:
The method comprises the steps of obtaining an inference prediction result of a hidden order model, setting a self-tolerance threshold set, starting an adaptive defense mechanism to identify and process a fault state of a device operation framework when a probability value of a detected fault deviates from a range of the set self-tolerance threshold set, and generating an adaptive regulation strategy according to the type and severity of the fault, wherein the regulation strategy comprises measures such as maintenance, replacement of a device component, dynamic adjustment of device operation parameters, optimization of a workflow and resource reassignment, for example, when a certain key component is predicted to be about to be faulty, the operation mode of the device is automatically adjusted, the load of the component is reduced, maintenance personnel is arranged to prepare the replacement component in advance, the device can be stably transited when the fault happens, and downtime and loss are reduced.
It should be understood that, in various embodiments of the present application, the sequence numbers of the foregoing processes do not mean the order of execution, and the order of execution of the processes should be determined by the functions and internal logic thereof, and should not constitute any limitation on the implementation process of the embodiments of the present application.
It should be understood that determining B from a does not mean determining B from a alone, but can also determine B from a and/or other information.
The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Finally, the foregoing description of the preferred embodiment of the invention is provided for the purpose of illustration only, and is not intended to limit the invention to the particular form disclosed, but on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.