Temperature sensor fault diagnosis method and system

Abstract

The invention relates to the technical field of fault diagnosis, and discloses a temperature sensor fault diagnosis method and system, and the method comprises the following steps: processing industrial furnace multi-source sensing data, and obtaining a standardized multi-physical field feature data set; processing the standardized multi-physics field feature data set to obtain a multi-physics field coupling feature vector; processing the multi-physical field coupling feature vector and the network topology structure of the temperature sensor to obtain a health state evaluation result of the temperature sensor; processing the health state evaluation result of the temperature sensor and historical fault data to obtain a sensor fault prediction report; and processing the fault sensor data and the surrounding health sensor data to obtain a data compensation result of the fault sensor. Under the monitoring of a temperature sensor network in a complex industrial environment, the reliability and the safety of the system can be remarkably improved, and the shutdown loss and the safety risk caused by sensor faults are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and more specifically, to a temperature sensor fault diagnosis method and system. Background Art

[0002] In some high-temperature industrial furnace system temperature monitoring scenarios, such as steel smelting, glass manufacturing or ceramic firing, the temperature sensor network is simultaneously affected by high-temperature thermal fields, molten metal fluid fields, electromagnetic induction fields and structural vibration fields. In such complex environments, the failure rate of temperature sensors is high and the failure mode is complex. The accuracy of fault diagnosis based solely on temperature data is low. Once the sensor fails, the lack of key temperature data may lead to product quality fluctuations or safety accidents.

[0003] In the prior art, fault diagnosis of temperature sensors in industrial environments mainly relies on single temperature data analysis or simple threshold detection. These methods have low accuracy under the interference of complex physical environments. Traditional fault diagnosis methods have the following main shortcomings: Dependence on a single data source: Most existing methods only use temperature data for fault diagnosis, ignoring the interaction of multiple physical fields, resulting in low diagnostic accuracy in complex physical environments.

[0004] Weak ability to identify physical environment interference: Unable to effectively distinguish between data anomalies caused by physical environment interference (such as electromagnetic interference, vibration) and actual sensor hardware failures, resulting in a high false alarm rate.

[0005] Lack of predictive analysis: Most methods only detect faults after they occur, lack the ability to conduct predictive analysis of the health status of the sensor, and cannot provide early warning of potential faults.

[0006] Low data compensation accuracy: After a sensor fails, the accuracy of traditional compensation methods is limited, especially under conditions of large temperature gradients and complex environments, where the compensation data deviates greatly from the actual value.

[0007] Poor model adaptability: Existing methods have weak generalization capabilities for different industrial scenarios and are difficult to adapt to changing working conditions and the special needs of different industries. Summary of the invention

[0008] The present invention provides a temperature sensor fault diagnosis method and system to solve the technical problems in the above-mentioned related technologies.

[0009] The present invention provides a temperature sensor fault diagnosis method, comprising: The multi-source sensor data of the industrial furnace is processed through a multi-physics hierarchical preprocessing network to obtain a standardized multi-physics feature data set; Processing the standardized multi-physics field feature data set through a multi-field coupling feature extraction network enhanced by physical knowledge to obtain a multi-physics field coupling feature vector; Processing the multi-physical field coupling feature vector and the temperature sensor network topology structure through a physical consistency graph neural network to obtain a temperature sensor health status assessment result; Processing the health status assessment result and historical fault data of the temperature sensor through a multi-physics field driven sensor fault prediction model to obtain a sensor fault prediction report; The faulty sensor data and surrounding healthy sensor data are processed through a sensor data compensation network assisted by a physical model to obtain the data compensation result of the faulty sensor.

[0010] Furthermore, in the step of processing the multi-source sensor data of the industrial furnace by a multi-physics field hierarchical preprocessing network, the mathematical expression of the multi-physics field hierarchical preprocessing network is: ; in: represents the raw multiphysics data; represents the data collection function; represents the noise filtering function; represents the feature extraction function; Represents the output normalized multiphysics feature dataset.

[0011] Furthermore, in the step of processing the standardized multi-physics field feature data set through the multi-field coupling feature extraction network enhanced by physical knowledge, the mathematical expression of the multi-field coupling feature extraction network enhanced by physical knowledge is: ; in: represents a standardized multiphysics feature dataset; represents the physical field-specific feature extraction function; Represents the physical field interaction characteristic analysis function; represents the feature concatenation function; Represents the multi-field fusion self-attention function; represents the multiphysics coupling eigenvector.

[0012] Furthermore, in the step of processing the multi-physical field coupling feature vector and the temperature sensor network topology structure by means of a physical consistency graph neural network, the mathematical expression of the physical consistency graph neural network is: ; in: Represents the multi-physics coupling eigenvector; Represents sensor location and topological relationship data; Represents the graph construction function; represents the graph convolution function; represents the health status assessment function; Indicates the temperature sensor health status assessment result.

[0013] Furthermore, the sensor network graph structure generated by the graph construction function in the physical consistency graph neural network is expressed as: ; in: Represents a node set consisting of N sensor nodes; represents a set of edges, connecting related sensor nodes; Represents the optimized adjacency matrix, defining the connection relationship and weight between nodes.

[0014] Furthermore, the graph convolution function calculation formula in the physical consistency graph neural network is: ; in: represents the adjacency matrix after adding the self-loop; represents the corresponding degree matrix, whose diagonal elements ; represents the N-dimensional identity matrix; represents the weight matrix of the lth layer; Represents the node feature matrix of the lth layer; represents a non-linear activation function.

[0015] Furthermore, in the step of processing the temperature sensor health status assessment result and historical fault data by a multi-physical field driven sensor fault prediction model, the mathematical expression of the multi-physical field driven sensor fault prediction model is: ; in: Indicates the health status assessment result of the temperature sensor; Represents multi-physics environment parameters; Represents the multi-physics coupling eigenvector; represents the health feature extraction function; represents the fault prediction function; represents the fault analysis function; Indicates a sensor failure prediction report.

[0016] Furthermore, in the step of processing faulty sensor data and surrounding healthy sensor data through a sensor data compensation network assisted by a physical model, the mathematical expression of the sensor data compensation network assisted by a physical model is: ; in: Represents temperature sensor network data; Indicates fault sensor information; represents physical model parameters; represents the relevant sensor selection function; represents the data compensation function; Indicates the data compensation result of the faulty sensor.

[0017] Furthermore, the calculation formula of the physical constraint graph attention compensation function in the physical model-assisted sensor data compensation network is: ; in: represents the subgraph containing the faulty sensor and related sensors; Represents the timing characteristics of the relevant sensors; Represents the parameters of the heat conduction physical model; represents the physical constraint graph attention compensation function; Indicates the data compensation result of the faulty sensor.

[0018] A temperature sensor fault diagnosis system, comprising: A multi-physics hierarchical preprocessing network is used to process multi-source sensor data of industrial furnaces to obtain standardized multi-physics feature data sets; A multi-field coupling feature extraction network enhanced by physical knowledge, used to process the standardized multi-physics field feature data set to obtain a multi-physics field coupling feature vector; A physical consistency graph neural network, used to process the multi-physical field coupling feature vector and the temperature sensor network topology to obtain a temperature sensor health status assessment result; A multi-physics field driven sensor fault prediction model, used to process the health status assessment result and historical fault data of the temperature sensor to obtain a sensor fault prediction report; The sensor data compensation network assisted by the physical model is used to process the faulty sensor data and the surrounding healthy sensor data to obtain the data compensation result of the faulty sensor.

[0019] The beneficial effects of the present invention are: Improved fault prediction performance: Compared with traditional methods that only use temperature data, the fault prediction accuracy is improved and the average warning lead time is extended, providing a sufficient time window for preventive maintenance.

[0020] Improved data compensation accuracy: In the event of sensor failure, the average relative error between the compensated data and the actual value is reduced, and high accuracy can still be maintained in an environment with large temperature fluctuations, meeting the requirements of industrial process control.

[0021] Enhanced false fault identification capability: Through multi-physical field information fusion, the system can accurately distinguish between anomalies caused by physical interference and actual sensor hardware failures, reduce false alarm rates, and significantly reduce unnecessary maintenance costs.

[0022] Improved system robustness: In the presence of strong vibration, electromagnetic interference and extreme temperature gradient conditions, the system can still maintain a high fault prediction accuracy, demonstrating excellent adaptability to complex working conditions.

[0023] Computing efficiency optimization: Through edge deployment optimization, the system can run in real time on the edge server at the industrial site, reducing the average processing delay to meet real-time monitoring needs.

[0024] Enhanced maintenance decision support: Comprehensive failure prediction reports provide detailed analysis of failure risk, type and physical cause, providing maintenance personnel with actionable decision suggestions and shortening the mean time to troubleshoot. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a flow chart of a temperature sensor fault diagnosis method of the present invention. DETAILED DESCRIPTION

[0026] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0027] At least one embodiment of the present invention discloses a temperature sensor fault diagnosis method, such as Figure 1 As shown, including: Step 1: Process the multi-source sensor data of the industrial furnace through a multi-physics hierarchical preprocessing network to obtain a standardized multi-physics feature data set; The multi-physics hierarchical preprocessing network consists of a data acquisition module, a noise filtering module, and a feature extraction module. The raw data collected by the data acquisition module is transmitted to the noise filtering module for noise reduction, and the noise-reduced data is passed to the feature extraction module to generate standardized features.

[0028] The mathematical expression of the multi-physics hierarchical preprocessing network is: ; in: represents the raw multiphysics data; represents the data collection function; represents the noise filtering function; represents the feature extraction function; A standardized multiphysics characterization dataset representing the output; Step 1 includes the following sub-steps: Sub-step 1.1: Collect multi-physics field data of the industrial furnace system through a multi-scale adaptive sampling algorithm to obtain an original multi-physics field data set; Sub-step 1.1 uses a multi-scale adaptive sampling algorithm to process the data collection requirements of different physical fields in the industrial furnace system. The input is the real-time signals of various sensors, and the output is the original multi-physical field data set. The algorithm automatically adjusts the sampling frequency according to the change rate of different physical fields, using high-frequency sampling for fast-changing physical quantities and low-frequency sampling for slowly changing physical quantities to achieve optimal resource allocation.

[0029] The mathematical expression of the multi-scale adaptive sampling algorithm is: ; in: represents the sampling frequency of the i-th physical field at time t; Indicates the basic sampling frequency; Represents the basic sampling coefficient of the i-th physical field; represents the adaptive adjustment coefficient; Represents the change of the i-th physical field signal per unit time; Indicates the maximum allowed change in the signal.

[0030] The acquired raw multiphysics datasets include: Temperature field data: contains multi-point temperature values ​​and their timestamps; Fluid field data: including flow velocity, flow direction, and pressure parameters; Electromagnetic field data: including electric field strength, magnetic field strength, and electromagnetic interference spectrum parameters; Vibration field data: including acceleration, amplitude, and frequency parameters; Environmental parameter data: including humidity, air pressure, and ambient temperature parameters; Sub-step 1.2: Processing the original multi-physics data set by a multi-physics adaptive noise filtering algorithm to obtain a denoised multi-physics data set; Sub-step 1.2 uses a multi-physics field adaptive noise filtering algorithm to process the original multi-physics field data set obtained in sub-step 1.1, and outputs a denoised multi-physics field data set. The algorithm combines wavelet transform and Kalman filter to perform adaptive denoising processing based on the noise characteristics of different physical fields.

[0031] The mathematical expression of the multi-physics adaptive noise filtering algorithm is: ; in: Represents the original data of the i-th physical field; represents the wavelet transform function; represents the inverse wavelet transform function; represents the threshold function, is the threshold parameter of the i-th physical field; represents the Kalman filter function; represents the denoised data of the i-th physical field; For different physical fields, the threshold parameter Adaptively calculated using the following formula: ; in: represents the standard deviation estimate of the i-th physical field noise; Indicates the number of data points; Sub-step 1.3: Processing the denoised multi-physics data set through a physical knowledge fusion feature extraction network to obtain a standardized multi-physics feature data set; Sub-step 1.3 uses a physical knowledge fusion feature extraction network to process the denoised multi-physics data set obtained in sub-step 1.2 and output a standardized multi-physics feature data set. The network combines physical model knowledge and deep learning technology to extract key features related to temperature sensor failure from multi-source heterogeneous data.

[0032] The physical knowledge fusion feature extraction network consists of a feature extraction module, a physical knowledge constraint module, and a feature standardization module. The initial features extracted by the feature extraction module are passed to the physical knowledge constraint module for screening and optimization, and the optimized features are passed through the feature standardization module to generate the final standardized features.

[0033] The mathematical expression of the network is: ; in: represents the multi-physics data after noise reduction; represents the feature extraction function; Represents physical knowledge constraint function; represents the feature normalization function; represents a standardized multiphysics feature dataset; The feature extraction function is implemented based on a multi-layer perceptron: ; in: Represents the activation function ReLU; denote the first and second weight matrices respectively; Represents the bias vector.

[0034] The physical knowledge constraint function adjusts the features through physical rules: ; in: Represents element-wise multiplication; represents the physical correlation mask matrix; represents the physical consistency correction function; Represents the physical knowledge correction coefficient.

[0035] The standardized multiphysics feature dataset contains the following feature categories: Temperature field characteristics: temperature distribution, gradient, and rate of change; Fluid field characteristics: velocity distribution, turbulence intensity, pressure gradient; Electromagnetic field characteristics: electromagnetic interference intensity, spectrum characteristics; Vibration field characteristics: vibration mode and frequency characteristics; Cross-coupling features: features that describe the interaction between different physical fields; Step 2: Process the standardized multi-physics field feature data set through the multi-field coupling feature extraction network enhanced by physical knowledge to obtain the multi-physics field coupling feature vector; Step 2 uses a multi-field coupling feature extraction network enhanced by physical knowledge to process the standardized multi-physics field feature data set obtained in step 1 and output a multi-physics field coupling feature vector. The network consists of a physical field-specific feature extraction module, a physical field interaction feature analysis module, and a multi-field fusion self-attention module. The single-field features extracted by the physical field-specific feature extraction module are passed to the physical field interaction feature analysis module for cross-field correlation analysis. The outputs of the two modules are jointly input into the multi-field fusion self-attention module to generate the final multi-physics field coupling feature vector.

[0036] The mathematical expression of the network is: ; in: represents a standardized multiphysics feature dataset; represents the physical field-specific feature extraction function; Represents the physical field interaction characteristic analysis function; represents the feature concatenation function; Represents the multi-field fusion self-attention function; represents the multiphysics coupling eigenvector.

[0037] Step 2 includes the following sub-steps: Sub-step 2.1: Processing the standardized multi-physics field feature data set through a physics-specific feature extraction network to obtain a single-field feature vector set; Sub-step 2.1 uses a physical field-specific feature extraction network to process the standardized multi-physical field feature data set obtained in step 1, and outputs a single-field feature vector set. The network contains four parallel feature extraction branches, each branch specializing in processing the features of a physical field, and uses a targeted network structure to capture the characteristic patterns of different physical fields.

[0038] The mathematical expression of the physics-specific feature extraction network is: ; ; in: , , , They represent the standardized characteristic data of temperature field, electromagnetic field, vibration field and fluid field respectively; , , , They represent the feature extraction functions of temperature field, electromagnetic field, vibration field and fluid field respectively; , , , They represent the extracted characteristic vectors of temperature field, electromagnetic field, vibration field and fluid field respectively; Represents a single-field eigenvector set, which contains the eigenvectors of four physical fields.

[0039] The feature extraction function for the temperature field uses a time and frequency dual-path feature extractor: ; in: represents a time domain convolutional neural network; represents a frequency domain convolutional neural network; represents fast Fourier transform; Represents the adaptive weight coefficient, which is learned through the attention mechanism.

[0040] The feature extraction function of the electromagnetic field adopts the method of combining wavelet packet transform and CNN: ; in: represents wavelet packet transform; Represents a convolutional neural network processing wavelet coefficients.

[0041] The improved long short-term memory network is used for the feature extraction function of the vibration field: ; in: Represents a bidirectional long short-term memory network.

[0042] The feature extraction function for the fluid field adopts a recursive neural network based on phase space reconstruction: ; in: represents the phase space reconstruction function; Represents a gated recurrent unit network.

[0043] Sub-step 2.2: Process the single-field feature vector set through the cross-field correlation analysis algorithm constrained by physical laws to obtain the physical field interaction feature matrix; Sub-step 2.2 uses a cross-field correlation analysis algorithm constrained by physical laws to process the single-field feature vector set obtained in sub-step 2.1 and output a physical field interaction feature matrix. This algorithm is based on physical domain knowledge, analyzes the coupling relationship and interaction between different physical fields, and generates a feature matrix that characterizes the interaction effect between physical fields.

[0044] The mathematical expression of the cross-field correlation analysis algorithm constrained by physical laws is: ; ; in: Represents the eigenvector of the i-th physical field; represents the interaction feature extraction function between the i-th and j-th physical fields; Represents the interaction characteristics between the i-th and j-th physical fields; represents the physics interaction characteristic matrix.

[0045] The interactive feature extraction function of the temperature field and the electromagnetic field is designed based on the Joule thermal effect and the electromagnetic induction principle: ; in: represents the time derivative of the temperature characteristic; Represents material-dependent conductivity parameters.

[0046] For the interactive feature extraction function of temperature field and vibration field, it is designed based on thermal expansion and thermoelastic effects: ; in: Represents the spatial gradient of temperature characteristics; Represents the coefficient of thermal expansion.

[0047] Similarly, the interaction feature extraction functions between other physical field pairs are also designed based on the corresponding physical laws, and finally form a complete physical field interaction feature matrix R.

[0048] Sub-step 2.3: Process the single-field feature vector set and the physical field interaction feature matrix through the multi-field fusion self-attention network to obtain the multi-physical field coupling feature vector; Sub-step 2.3 uses a multi-field fusion self-attention network to process the single-field feature vector set obtained in sub-step 2.1 and the physical field interaction feature matrix obtained in sub-step 2.2, and outputs a multi-physical field coupling feature vector. The network uses the self-attention mechanism to capture the complex relationship between different features and ensures that the features conform to physical laws through physical consistency constraints.

[0049] The multi-field fusion self-attention network consists of a feature initial fusion layer, a multi-head self-attention layer, and a physical consistency constraint layer. The feature initial fusion layer fuses the single-field feature vector and the interactive feature matrix into a unified feature, the multi-head self-attention layer captures the relationship between different features, and the physical consistency constraint layer ensures that the feature representation conforms to physical laws.

[0050] The mathematical expression of the network is: ; in: Represents the output of the last self-attention layer; Represents the physical law constraint function; represents the self-attention balance parameter; represents the multiphysics coupling eigenvector.

[0051] The initial feature fusion process is expressed as: ; in: Represents a vector concatenation operation; Represents the operation of converting a matrix to a vector.

[0052] The calculation formula of the multi-head self-attention layer is: ; in: Represents the characteristics of the i-th node in the l-th layer; and Represent the weight matrix and bias vector of the lth layer respectively; represents the attention weight from node j to node i in layer l, which is calculated as: ; ; in: represents the learnable self-attention vector; Represents a vector concatenation operation; Represents the ReLU activation function.

[0053] The physical consistency constraint function includes the physical model of the first law of thermodynamics, the heat conduction equation, and the law of electromagnetic induction to ensure that the feature representation conforms to the laws of physics: ; in: represents the kth physical constraint function; represents the weight of the kth physical constraint; Represents the total number of physical constraints.

[0054] The final output multi-physics coupling feature vector It is a d-dimensional vector, where d is usually set to 128 or 256, containing rich information about the complex interactions between physical fields, which will be used for subsequent sensor health status assessment.

[0055] Step 3: Process the multi-physics field coupling feature vector and the temperature sensor network topology through the physical consistency graph neural network to obtain the temperature sensor health status assessment result; Step 3 uses a physical consistency graph neural network to process the multi-physics field coupling feature vectors and the topological structure information of the temperature sensor network obtained in step 2, and outputs the health status assessment result of the temperature sensor. The network models the sensors as nodes in the graph, and the physical and data relationships between sensors as edges, and achieves high-precision health status assessment through message passing mechanisms and physical constraints.

[0056] The physical consistency graph neural network consists of a graph construction module, a graph convolution module, and a health status assessment module. The sensor network graph structure generated by the graph construction module is passed to the graph convolution module for feature propagation and update, and the output of the graph convolution module is passed to the health status assessment module to generate the final health status assessment result.

[0057] The mathematical expression of the network is: ; in: Represents the multi-physics coupling eigenvector; Represents sensor location and topological relationship data; Represents the graph construction function; represents the graph convolution function; represents the health status assessment function; Indicates the temperature sensor health status assessment result.

[0058] Step 3 includes the following sub-steps: Sub-step 3.1: Process the sensor spatial layout information and physical association relationship through a dynamic adaptive graph construction algorithm to obtain the sensor network graph structure; Sub-step 3.1 uses a dynamic adaptive graph construction algorithm to process the sensor spatial layout information and physical association relationships, and outputs the sensor network graph structure. The algorithm comprehensively considers the physical distance, data correlation and physical environment factors of the sensors to construct a graph structure that can accurately reflect the complex relationship between sensors.

[0059] The mathematical expression of the dynamic adaptive graph construction algorithm is: ; in: Represents a node set consisting of N sensor nodes; represents a set of edges, connecting related sensor nodes; Represents the optimized adjacency matrix, defining the connection relationship and weight between nodes.

[0060] The algorithm first calculates the physical distance matrix based on the spatial position of the sensor: ; in: and Represent the three-dimensional spatial coordinates of sensors i and j respectively; represents the Euclidean distance between two sensors; represents the Gaussian kernel width parameter; Indicates the distance threshold.

[0061] The correlation matrix is ​​then calculated based on the sensor data: ; in: and Represent the temperature time series data of sensors i and j respectively; Represents the Pearson correlation coefficient calculation function.

[0062] The physical distance matrix and the correlation matrix are fused to generate the initial adjacency matrix: ; in: represents a balancing parameter that adjusts the relative importance of physical distance and data relevance.

[0063] Finally, an adaptive threshold is applied to optimize the adjacency matrix: ; in: Represents the adaptive threshold, which is determined by the graph connectivity optimization algorithm.

[0064] Sub-step 3.2: Process the sensor network graph structure and multi-physics field coupling feature vectors through the physical constraint graph convolutional network to obtain the node health feature matrix.

[0065] Sub-step 3.2 uses a physical constraint graph convolutional network to process the sensor network graph structure obtained in sub-step 3.1 and the multi-physics field coupling feature vector obtained in step 2, and outputs a node health feature matrix. The network implements feature propagation and update between nodes through graph convolution operations, and introduces physical constraints to ensure that feature evolution complies with physical laws such as heat conduction.

[0066] The physical constraint graph convolution network consists of a node feature initialization module, a multi-layer graph convolution module, and a physical constraint module. The node feature initialization module assigns multi-physical field coupling features to corresponding sensor nodes, the multi-layer graph convolution module performs message transmission on the graph, and the physical constraint module introduces heat conduction physical constraints.

[0067] The mathematical expression of the network is: ; in: Represents the node feature matrix of the lth layer; Represents the feature update amount of the lth layer; represents a nonlinear activation function; Represents the node health feature matrix output by the last layer.

[0068] The node feature initialization process is: ; in: Represents the multi-physics coupling eigenvector; and Represent the initialized weight matrix and the initialized bias vector respectively; represents the initial features of the i-th sensor node.

[0069] The calculation formula of the graph convolution layer is: ; in: represents the adjacency matrix after adding the self-loop; represents the corresponding degree matrix, whose diagonal elements ; represents the N-dimensional identity matrix; represents the weight matrix of the lth layer.

[0070] The heat conduction physical constraint is introduced through the following loss function: ; in: represents the discrete Laplace operator; represents the virtual time step; represents the thermal diffusion coefficient; represents the Frobenius norm.

[0071] Sub-step 3.3: Process the node health feature matrix through a multi-level threshold health status classifier to obtain the temperature sensor health status assessment result; Sub-step 3.3 uses a multi-level threshold health status classifier to process the node health feature matrix obtained in sub-step 3.2 and output the temperature sensor health status assessment result. The classifier uses a multi-level threshold judgment mechanism, combined with time continuity test and interference elimination strategy, to achieve accurate assessment of the sensor health status.

[0072] The multi-level threshold health status classifier consists of a health score calculation module, a dynamic threshold generation module, and a state judgment module. The health score calculation module converts the node health characteristics into a numerical health score, the dynamic threshold generation module calculates the judgment threshold based on historical data and current working conditions, and the state judgment module determines the health status of the sensor based on the health score and threshold.

[0073] The mathematical expression of the classifier is: ; in: represents the node health feature matrix; represents the health score calculation function; represents a dynamic threshold set; represents the state determination function; Indicates the temperature sensor health status assessment result.

[0074] The health score calculation formula is: ; in: represents the health feature vector of the i-th sensor node; represents a multi-layer perceptron; represents the health score of the ith sensor, 1 represents completely healthy and 0 represents completely faulty.

[0075] The dynamic threshold calculation formula is: ; in: and represent the mean and standard deviation of the historical health scores, respectively; represents the working condition adjustment function; Indicates the current working condition parameters.

[0076] The status determination rules are: ; in: and Indicates the minimum duration thresholds for the fault state and warning state respectively.

[0077] Interference elimination judgment is achieved by analyzing the correlation between health score anomalies and physical field disturbances: ; in: Represents the time series correlation analysis function; A time series representing a disturbance in a particular physical field; Represents the correlation threshold.

[0078] The final output of the temperature sensor health status assessment result is: ; in: Indicates the sensor index; Indicates health status (healthy, warning, or faulty); Indicates the confidence level of the state judgment.

[0079] Step 4: Process the temperature sensor health status assessment results and historical fault data through the multi-physics field driven sensor fault prediction model to obtain a sensor fault prediction report; Step 4 uses a multi-physics field driven sensor fault prediction model to process the temperature sensor health status assessment results and historical fault data obtained in step 3, and outputs a sensor fault prediction report. The model combines the sensor health status evolution characteristics and multi-physics field environmental parameters, predicts the possible future sensor failure risks through deep learning technology, and analyzes the possible failure types and causes.

[0080] The multi-physics field driven sensor fault prediction model consists of a health feature extraction module, a fault prediction module and a fault analysis module. The health evolution features extracted by the health feature extraction module are input into the fault prediction module for risk prediction. The prediction results of the fault prediction module and the multi-physics field coupling features are input into the fault analysis module to generate the final fault prediction report.

[0081] The mathematical expression of the model is: ; in: Indicates the health status assessment result of the temperature sensor; Represents multi-physics environment parameters; Represents the multi-physics coupling eigenvector; represents the health feature extraction function; represents the fault prediction function; represents the fault analysis function; Indicates a sensor failure prediction report.

[0082] Step 4 includes the following sub-steps: Sub-step 4.1: Process the time series of the temperature sensor health status assessment results through the time series health feature extraction algorithm to obtain the sensor health evolution feature vector; Sub-step 4.1 uses a time series health feature extraction algorithm to process the time series of the temperature sensor health status assessment results obtained in step 3, and outputs a sensor health evolution feature vector. The algorithm extracts features that reflect the sensor performance degradation trend and pattern from the historical data of the health status assessment results through a time series analysis method.

[0083] The mathematical expression of the time series health feature extraction algorithm is: ; in: Represents trend characteristics and describes the long-term trend of health status; It represents the fluctuation characteristics and describes the short-term fluctuation pattern of health status; Represents periodic characteristics and describes the periodic changes in health status; Represents event characteristics, describing the pattern in which health status is affected by a specific event; Represents the sensor health evolution feature vector.

[0084] The algorithm first performs time series preprocessing, including standardization and denoising: ; in: represents the health score of the i-th sensor at time t; and denote the mean and standard deviation of the sensor health score, respectively; represents the standardized health score.

[0085] Then the improved seasonal decomposition algorithm is applied to extract trend features: ; in: Indicates trend items, reflecting long-term change trends; represents the seasonal term, reflecting periodic changes; Represents the residual term, reflecting random fluctuations.

[0086] The trend characteristic calculation formula is: ; in: Indicates the rate of change of a trend; Indicates the acceleration of the trend.

[0087] The fluctuation characteristic calculation formula is: ; in: represents the standard deviation of the residual term; It represents the skewness of the residual term; Represents the kurtosis of the residual term.

[0088] The periodic features and event features are extracted through wavelet analysis and autocorrelation function, and finally all the features are integrated into the sensor health evolution feature vector.

[0089] Sub-step 4.2: Process the sensor health evolution feature vector and multi-physics field environment parameters through the attention-enhanced long short-term memory prediction model to obtain the sensor failure risk prediction result; Sub-step 4.2 uses the attention-enhanced long short-term memory prediction model to process the sensor health evolution feature vector and multi-physical field environmental parameters obtained in sub-step 4.1, and outputs the sensor failure risk prediction result. This model combines the attention mechanism and the long short-term memory network to capture the failure risk patterns on different time scales and consider the influence of physical environmental factors.

[0090] The mathematical expression of the attention-enhanced long-term short-term memory prediction model is: ; in: represents the sensor health evolution feature vector; Represents multi-physics environment parameters; Long short-term memory network representing enhanced attention; Represents the sensor failure risk prediction result.

[0091] The calculation formula of the physical field attention mechanism is: ; in: represents the jth physical field environment parameter; Represents a vector concatenation operation; and Represents the weights and biases of the attention network; Represents the attention weight of the j-th physical field environment parameter.

[0092] The update formula of the long short-term memory network is: ; in: and Respectively represent the hidden state and cell state of the LSTM unit at time t; and They represent the hidden state and cell state of the LSTM unit at time t;1 respectively.

[0093] The final failure risk prediction results include failure risk probabilities and uncertainty estimates within different time frames: ; in: Indicates that sensors i in the future The risk probability of failure within a certain period of time; Represents the uncertainty estimate of the risk prediction; Indicates different forecast time ranges (such as 24 hours, 72 hours, 168 hours).

[0094] Sub-step 4.3: Processing the sensor fault risk prediction results and the multi-physics field coupling feature vectors through the multi-modal fault analysis engine to obtain a sensor fault prediction report; Sub-step 4.3 uses a multimodal fault analysis engine to process the sensor fault risk prediction results obtained in sub-step 4.2 and the multi-physics field coupling feature vector obtained in step 2, and outputs a sensor fault prediction report. The engine analyzes potential fault types, physical causes, and key evolution time points through a combination of machine learning and rule reasoning to generate a comprehensive fault prediction report.

[0095] The mathematical expression of the multimodal fault analysis engine is: ; in: represents the sensor failure risk prediction result; Represents the multi-physics coupling eigenvector; Indicates the health status assessment result of the temperature sensor; represents the multi-modal fault analysis function; Indicates a sensor failure prediction report.

[0096] The fault type classifier is implemented based on support vector machine: ; in: represents the fault risk vector of sensor i; Represents a vector concatenation operation; represents a support vector machine classifier; Indicates the predicted failure type.

[0097] The physical field contribution analysis is achieved through gradient calculation: ; in: Represents the failure risk versus physical field parameters The partial derivative of Represents the contribution of the j-th physical field to the failure risk.

[0098] The calculation formula for key time point marking is: ; in: represents the health score of sensor i at time t; Indicates the change threshold.

[0099] The resulting sensor failure prediction report contains the following: Basic information of the sensor (ID, location, type); The probability and uncertainty of failure risk in each future time period; Predicted fault types and probability ranking; Analysis of the contribution of each physical field to the failure risk; Marking of key time points in the evolution of health status; Failure prevention recommendations and optimal maintenance time windows.

[0100] Step 5: Process the faulty sensor data and surrounding healthy sensor data through the sensor data compensation network assisted by the physical model to obtain the data compensation result of the faulty sensor.

[0101] Step 5 uses a physical model-assisted sensor data compensation network to process the faulty sensor data and surrounding healthy sensor data, and outputs the data compensation result of the faulty sensor. The network combines the physical heat transfer model and the graph attention mechanism, and can infer the actual value of the failed sensor based on physical laws and adjacent sensor data in the case of sensor failure, thus achieving high-precision data compensation.

[0102] The physical model-assisted sensor data compensation network consists of a fault detection module, a related sensor selection module, and a data compensation module. The fault signal detected by the fault detection module is transmitted to the related sensor selection module, and the related sensor data selected by the module is input into the data compensation module to generate the final data compensation result.

[0103] The mathematical expression of the network is: ; in: Represents temperature sensor network data; Indicates fault sensor information; represents physical model parameters; represents the relevant sensor selection function; represents the data compensation function; Indicates the data compensation result of the faulty sensor.

[0104] Step 5 includes the following sub-steps: Sub-step 5.1: Process the sensor real-time data and fault prediction report through the real-time anomaly detection trigger to obtain a data compensation start signal; Sub-step 5.1 uses a real-time anomaly detection trigger to process the real-time sensor data and the fault prediction report obtained in step 4, and outputs a data compensation start signal. This trigger quickly detects sensor data anomalies or interruptions by combining statistical methods and prediction information, and triggers the data compensation process.

[0105] The mathematical expression of the real-time anomaly detection trigger is: ; in: Represents real-time sensor data; Indicates a failure prediction report; represents the anomaly detection scoring function; Indicates triggering decision function; Indicates the data compensation start signal.

[0106] The anomaly detection scoring function is calculated based on a combination of statistical deviation and predicted risk: ; in: Indicates the Z score of the sensor value; represents the failure risk probability of sensor i; and Represents the first and second scoring weight coefficients.

[0107] The trigger decision function determines whether to start data compensation through threshold judgment: ; in: Indicates the trigger threshold; Indicates a situation where sensor data is interrupted.

[0108] The data compensation start signal contains the following information: fault sensor ID, fault detection timestamp, fault type indication, and anomaly score value.

[0109] Sub-step 5.2: Process the sensor network topology and fault sensor information through a physically-oriented relevant sensor selection algorithm to obtain an optimal set of relevant sensors; Sub-step 5.2 uses a physically guided correlation sensor selection algorithm to process the sensor network topology and the faulty sensor information obtained in sub-step 5.1, and outputs the optimal set of correlation sensors. Based on the physical heat transfer model and data correlation analysis, the algorithm selects a subset of sensors with high correlation with the faulty sensor from the sensor network to provide reliable input data for subsequent data compensation.

[0110] The mathematical expression of the physical-oriented sensor selection algorithm is: ; in: Represents the sensor network graph structure; Indicates the fault sensor index; Represents the parameters of the heat conduction physical model; represents the sensor ranking function; represents the function of selecting the first K elements; represents the optimal set of related sensors.

[0111] The sensor ranking function is calculated based on the combined relevance score: ; in: represents the physical distance between sensors i and j; represents the time delay mutual correlation coefficient between sensors i and j; represents the time delay corresponding to the maximum cross-correlation; represents the correlation score based on the heat flow propagation path; , and represents the first, second, and third correlation weight coefficients; Represents the distance constant.

[0112] The heat flow propagation path analysis is based on the heat conduction equation: ; in: represents the thermal diffusion coefficient; represents the Laplace operator; Represents a heat source term.

[0113] The optimal correlated sensor set finally selected contains K sensors with the highest correlation with the faulty sensor, and K is usually 5-10.

[0114] Sub-step 5.3: Process the optimal relevant sensor set data through the physical constraint graph attention compensation network to obtain the data compensation result of the faulty sensor; Sub-step 5.3 uses the physical constraint graph attention compensation network to process the optimal relevant sensor set data obtained in sub-step 5.2 and output the data compensation result of the faulty sensor. The network combines the graph attention mechanism and physical constraints to capture the complex relationship between different sensors and ensure that the generated compensation value conforms to the laws of physics.

[0115] The mathematical expression of the physical constraint graph attention compensation network is: ; in: represents the subgraph containing the faulty sensor and related sensors; Represents the timing characteristics of the relevant sensors; Represents the parameters of the heat conduction physical model; represents the physical constraint graph attention compensation function; represents the data compensation result of the i-th faulty sensor.

[0116] The timing characteristics of the relevant sensors are expressed as: ; in: represents the temperature value of sensor j at time t; k△t; Indicates the length of the time window; Represents the time step.

[0117] The calculation formula of the graph attention layer is: ; ; in: and Represents the characteristics of nodes i and j; represents the neighbor set of node i; represents the third weight matrix; Represents the graph attention vector; Represents a vector concatenation operation; represents the activation function; represents the attention weight from node j to node i.

[0118] The physical constraint layer introduces the heat conduction law through the following loss function: ; in: represents the prediction function based on the physical model; represents the spatial distance between nodes i and j; represents the time delay between nodes i and j.

[0119] The final data compensation result includes the compensation value and its uncertainty estimate: ; in: represents the compensation value of faulty sensor i at time t; represents the standard deviation estimate of the compensation value; Represents the confidence interval for the compensation value.

[0120] This implementation is suitable for temperature sensor network monitoring in complex industrial environments such as steel smelting, glass manufacturing, and power generation. It can significantly improve system reliability and safety and reduce downtime losses and safety risks caused by sensor failures.

[0121] A temperature sensor fault diagnosis system, comprising: A multi-physics hierarchical preprocessing network is used to process multi-source sensor data of industrial furnaces to obtain standardized multi-physics feature data sets; A multi-field coupling feature extraction network enhanced by physical knowledge, used to process the standardized multi-physics field feature data set to obtain a multi-physics field coupling feature vector; A physical consistency graph neural network, used to process the multi-physical field coupling feature vector and the temperature sensor network topology to obtain a temperature sensor health status assessment result; A multi-physics field driven sensor fault prediction model, used to process the health status assessment result and historical fault data of the temperature sensor to obtain a sensor fault prediction report; The sensor data compensation network assisted by the physical model is used to process the faulty sensor data and the surrounding healthy sensor data to obtain the data compensation result of the faulty sensor.

[0122] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A temperature sensor fault diagnosis method, characterized in that: The following steps are involved: The multi-source sensor data of the industrial furnace is processed through a multi-physics hierarchical preprocessing network to obtain a standardized multi-physics feature data set; Processing the standardized multi-physics field feature data set through a multi-field coupling feature extraction network enhanced by physical knowledge to obtain a multi-physics field coupling feature vector; Processing the multi-physical field coupling feature vector and the temperature sensor network topology structure through a physical consistency graph neural network to obtain a temperature sensor health status assessment result; Processing the health status assessment result and historical fault data of the temperature sensor through a multi-physics field driven sensor fault prediction model to obtain a sensor fault prediction report; The faulty sensor data and surrounding healthy sensor data are processed through a sensor data compensation network assisted by a physical model to obtain the data compensation result of the faulty sensor.

2. The temperature sensor fault diagnosis method according to claim 1, characterized in that: In the step of processing the multi-source sensor data of the industrial furnace by a multi-physics field hierarchical preprocessing network, the mathematical expression of the multi-physics field hierarchical preprocessing network is: ； in: represents the raw multiphysics data; represents the data collection function; represents the noise filtering function; represents the feature extraction function; Represents the output normalized multiphysics feature dataset.

3. The temperature sensor fault diagnosis method according to claim 2, characterized in that: In the step of processing the standardized multi-physics field feature data set through the multi-field coupling feature extraction network enhanced by physical knowledge, the mathematical expression of the multi-field coupling feature extraction network enhanced by physical knowledge is: ； in: represents a standardized multiphysics feature dataset; represents the physical field-specific feature extraction function; Represents the physical field interaction characteristic analysis function; represents the feature concatenation function; Represents the multi-field fusion self-attention function; represents the multiphysics coupling eigenvector.

4. The temperature sensor fault diagnosis method according to claim 3, characterized in that: In the step of processing the multi-physical field coupling feature vector and the temperature sensor network topology structure by means of a physical consistency graph neural network, the mathematical expression of the physical consistency graph neural network is: ； in: Represents the multi-physics coupling eigenvector; Represents sensor location and topological relationship data; Represents the graph construction function; represents the graph convolution function; represents the health status assessment function; Indicates the temperature sensor health status assessment result.

5. The temperature sensor fault diagnosis method according to claim 4, characterized in that: The sensor network graph structure generated by the graph construction function in the physical consistency graph neural network is expressed as: ； in: Represents a node set consisting of N sensor nodes; represents a set of edges, connecting related sensor nodes; Represents the optimized adjacency matrix, defining the connection relationship and weight between nodes.

6. The temperature sensor fault diagnosis method according to claim 5, characterized in that: The calculation formula of the graph convolution function in the physical consistency graph neural network is: ； in: represents the adjacency matrix after adding the self-loop; represents the corresponding degree matrix, whose diagonal elements ; represents the N-dimensional identity matrix; represents the weight matrix of the lth layer; Represents the node feature matrix of the lth layer; represents a non-linear activation function.

7. The temperature sensor fault diagnosis method according to claim 6, characterized in that: In the step of processing the temperature sensor health status assessment result and historical fault data by a multi-physics field driven sensor fault prediction model, the mathematical expression of the multi-physics field driven sensor fault prediction model is: ； in: Indicates the health status assessment result of the temperature sensor; Represents multi-physics environment parameters; Represents the multi-physics coupling eigenvector; represents the health feature extraction function; represents the fault prediction function; represents the fault analysis function; Indicates a sensor failure prediction report.

8. The temperature sensor fault diagnosis method according to claim 7, characterized in that: In the step of processing faulty sensor data and surrounding healthy sensor data through a sensor data compensation network assisted by a physical model, the mathematical expression of the sensor data compensation network assisted by a physical model is: ； in: Represents temperature sensor network data; Indicates fault sensor information; represents physical model parameters; represents the relevant sensor selection function; represents the data compensation function; Indicates the data compensation result of the faulty sensor.

9. The temperature sensor fault diagnosis method according to claim 8, characterized in that: The calculation formula of the physical constraint graph attention compensation function in the physical model-assisted sensor data compensation network is: ； in: represents the subgraph containing the faulty sensor and related sensors; Represents the timing characteristics of the relevant sensors; Represents the parameters of the heat conduction physical model; represents the physical constraint graph attention compensation function; represents the data compensation result of the i-th faulty sensor.

10. A temperature sensor fault diagnosis system, characterized in that: include: A multi-physics hierarchical preprocessing network is used to process multi-source sensor data of industrial furnaces to obtain standardized multi-physics feature data sets; A multi-field coupling feature extraction network enhanced by physical knowledge, used to process the standardized multi-physics field feature data set to obtain a multi-physics field coupling feature vector; A physical consistency graph neural network, used to process the multi-physical field coupling feature vector and the temperature sensor network topology to obtain a temperature sensor health status assessment result; A multi-physics field driven sensor fault prediction model, used to process the health status assessment result and historical fault data of the temperature sensor to obtain a sensor fault prediction report; The sensor data compensation network assisted by the physical model is used to process the faulty sensor data and the surrounding healthy sensor data to obtain the data compensation result of the faulty sensor.

Images