Industrial instrument fault detection and self-calibration method based on intelligent sensing fusion

Abstract

The invention discloses an industrial instrument fault detection and self-calibration method based on intelligent sensing fusion, and relates to the technical field of instrument detection, and the method comprises the following steps: S1, carrying out the feature extraction of a multi-source signal; s2, performing consistency calibration to obtain a calibrated feature component; s3, performing weighted fusion to generate a fault indication quantity; s4, constructing a fault detection model; s5, using the fault severity calculation model to quantify the fault severity; s6, making corresponding calibration strategies according to different levels of fault severity; s7, performing dynamic evaluation on the effect of the calibration strategy; and S8, performing joint optimization on the constructed model and the calibration strategy. According to the industrial instrument fault detection and self-calibration method, the multi-source sensing signals and the multi-dimensional features are fused, comprehensive feature extraction and fault diagnosis are carried out, and the accuracy of fault detection is remarkably improved. Meanwhile, by introducing a calibration strategy, dynamic evaluation and joint optimization of a calibration effect are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of instrument detection, and in particular to an industrial instrument fault detection and self-calibration method based on intelligent sensor fusion. Background Art

[0002] With the continuous advancement of industrial automation and intelligence, industrial instrumentation, such as pressure sensors, temperature sensors, flow meters, and displacement sensors, is becoming increasingly widespread in production processes. These instruments are responsible for real-time monitoring and control of industrial process parameters, playing a vital role in ensuring production safety and product quality. However, due to the complex and ever-changing industrial field environment, industrial instruments are highly susceptible to factors such as temperature fluctuations, electromagnetic interference, mechanical vibration, and long-term wear and tear, resulting in signal drift, reduced sensitivity, zero offset, and even inaccuracy and failure.

[0003] Traditional instrument calibration methods rely heavily on manual operation and fixed calibration cycles, making them difficult to meet the real-time and high-reliability requirements of industrial sites. Furthermore, existing fault detection methods rely on single sensor signals or simple models, making them incapable of coping with multi-dimensional feature coupling and nonlinear variations. Furthermore, common calibration strategies rely on static calibration, ignoring the dynamic impact of operating environment and operating conditions on calibration accuracy. This results in suboptimal calibration results and insufficient stability. In practical applications, these deficiencies can lead to distorted equipment operating status monitoring, potentially causing significant safety hazards and economic losses.

[0004] In recent years, with the rapid development of intelligent sensing technology and data fusion algorithms, fault detection methods based on multi-sensor fusion and deep learning have gradually attracted attention. These methods improve the accuracy and stability of fault identification by fusing and analyzing multi-source signals. However, current research has primarily focused on fault detection, with less consideration given to fault self-calibration. Furthermore, the evaluation of calibration results and subsequent strategy optimization lack a systematic approach, making it difficult to maintain high accuracy and stability over the long term under complex operating conditions. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides an industrial instrument fault detection and self-calibration method based on intelligent sensor fusion to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides an industrial instrument fault detection and self-calibration method based on intelligent sensor fusion, comprising the following steps: S1. Extract multi-source signal features from sensor signals in industrial instrumentation systems; S2. performing consistency calibration on the extracted multi-source signal features to obtain calibrated feature components; S3. Perform weighted fusion on the calibrated characteristic components to generate a fault indicator; S4. Build a fault detection model based on the fault indicator to determine the fault type and its probability; S5. After outputting the fault category and probability, quantify the fault severity using the fault severity calculation model; S6. Based on the fault severity, formulate corresponding calibration strategies for different levels of fault severity; S7. Dynamically evaluate the effectiveness of the calibration strategy; S8. Jointly optimize the constructed model and calibration strategy.

[0007] To further optimize this technical solution, in step S1, the multi-source signal feature extraction includes time domain feature extraction, frequency domain feature extraction, and time-frequency feature extraction, extracting signal characteristics from different angles, and the expression of the comprehensive feature vector is: ; in, is a comprehensive feature vector, which is a high-dimensional feature vector used to characterize the sensor signal; Indicates the feature components.

[0008] To further optimize this technical solution, in step S2, the maximum and minimum normalization is used to normalize each feature component Calibrate and limit all characteristic components to the interval [0,1] to obtain the calibrated characteristic components .

[0009] To further optimize this technical solution, in step S3, a weighted fusion formula is used to generate fault indicators by optimizing weighting coefficients to achieve a comprehensive evaluation of multi-source features; The weighted fusion formula is as follows: ; in, is the fault indication quantity; For the The weighting coefficients of the feature components are calculated by the feature variance: ; Where, For the The variance of the calibrated feature components; For the The calibrated value of the characteristic components, The sum of the variances after calibration for all feature components; The weighted fusion formula is used to ensure that the weights of feature components are proportional to the amplitude of signal fluctuations.

[0010] To further optimize this technical solution, in step S4, the fault detection model includes: Model input: fault indicator Time series , is the time window length; Model output: probability distribution of various types of failures , For the The probability of class failure.

[0011] To further optimize this technical solution, the fault detection model has a built-in loss function and optimization algorithm to train and optimize the model; The loss function uses the cross entropy loss function to measure the deviation between the predicted value and the true value; The optimization algorithm uses the Adam optimization algorithm to adjust the model weights and evaluates the generalization performance of the model through cross-validation to prevent overfitting.

[0012] To further optimize this technical solution, in step S5, the fault severity calculation model includes: Construct severity index, select the index that is representative of the fault, and the membership function of each index Indicates the indicator's contribution to severity: ; in, For the indicator values, is the membership function; Construct severity evaluation formula: ; in, For the The risk level value corresponding to each indicator; The fault severity is divided into different levels, including minor faults, general faults, serious faults and critical faults.

[0013] To further optimize this technical solution, in step S6, the calibration strategy includes: Zero offset calibration; Parameter compensation calibration; Model reconstruction and calibration.

[0014] To further optimize the technical solution, in step S7, the dynamic evaluation of the calibration strategy includes: Calibration effect verification: The calibration error change value is used for evaluation, that is, the error change before and after calibration is compared. If the error is greatly reduced, it indicates that the calibration strategy has a significant effect. The calibration error change value is: , and are the error values before and after calibration respectively; Stability analysis: The stability coefficient is used to measure the consistency of calibration results in long-term operation and to effectively judge the reliability of the calibration strategy.

[0015] To further optimize this technical solution, in step S8, the goal of the joint optimization is to find the best balance between accuracy and stability, taking into account the accuracy of the model and the stability of the instrument signal after calibration. In actual operation, the model parameters are tuned and dynamically updated through a genetic algorithm.

[0016] In a second aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of an industrial instrument fault detection and self-calibration method based on intelligent sensor fusion as described in the first aspect of the present invention are implemented.

[0017] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of an industrial instrument fault detection and self-calibration method based on intelligent sensor fusion as described in the first aspect of the present invention are implemented.

[0018] Compared with the existing technology, the present invention provides an industrial instrument fault detection and self-calibration method based on intelligent sensor fusion, which belongs to the technical field of supervisory control and data acquisition systems and has the following beneficial effects: This intelligent sensor fusion-based industrial instrument fault detection and self-calibration method significantly improves fault detection accuracy by fusing multi-source sensor signals and multidimensional features for comprehensive feature extraction and fault diagnosis. Furthermore, by introducing a calibration strategy, it enables dynamic evaluation and joint optimization of calibration results, effectively overcoming the issues of unstable calibration results and insufficient fault detection accuracy in existing technologies. Through a multi-objective optimization strategy and dynamic threshold adjustment mechanism, it ensures the long-term reliability and adaptability of the instrument in complex industrial environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 This is a flow chart of an industrial instrument fault detection and self-calibration method based on intelligent sensor fusion proposed by the present invention; Figure 2 This is a classification diagram of the calibration strategy in the industrial instrument fault detection and self-calibration method based on intelligent sensor fusion proposed in the present invention. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it constitute a separate or selective embodiment that is mutually exclusive with other embodiments.

[0024] Example 1: Reference Figures 1 and 2 , which is the first embodiment of the present invention, provides an industrial instrument fault detection and self-calibration method based on intelligent sensor fusion, comprising the following steps: S1. Extract multi-source signal features from sensor signals in industrial instrumentation systems.

[0025] Acquiring multi-source sensor signal features is a prerequisite for subsequent fusion and fault detection. Industrial instruments utilize a variety of sensors (such as pressure, temperature, and vibration sensors) to acquire signals. Because different sensors have varying ranges, accuracies, and sampling rates, directly fusion using raw data can lead to information distortion or noise amplification. Therefore, feature extraction algorithms are used to decompose and extract the signals, preserving valid information while removing redundancy and noise.

[0026] Multi-source signal feature extraction includes time domain feature extraction, frequency domain feature extraction and time-frequency feature extraction, extracting signal characteristics from different angles.

[0027] In this embodiment, Time Domain Feature Extraction: Calculate the signal's mean, variance, peak value, skewness, and kurtosis coefficient to form basic time domain feature components. These features can reflect the overall volatility and trend of the signal. For example, the peak feature of a pressure signal can indicate a system's instantaneous overload, while the mean feature can indicate a system's stable operating state.

[0028] Frequency Domain Feature Extraction: Using the Fast Fourier Transform (FFT) to extract the signal's spectral characteristics, the primary frequency components and harmonic amplitudes are extracted. Frequency domain features such as primary frequency components and harmonic amplitudes can reflect periodic fluctuations or mechanical vibration characteristics during instrument operation.

[0029] Time-frequency feature extraction: Wavelet transform is used to extract instantaneous energy characteristics and signal mutation characteristics. This can effectively characterize signal mutations and instability, and is particularly suitable for detecting short-term or intermittent faults.

[0030] The expression of the comprehensive eigenvector is: ; in, is a comprehensive feature vector, which is a high-dimensional feature vector used to characterize the sensor signal; Indicates the feature components.

[0031] S2. Perform consistency calibration on the extracted multi-source signal features to obtain calibrated feature components.

[0032] The extracted features may have scale differences or offsets. Feature consistency calibration ensures that the feature quantities are comparable, laying the foundation for subsequent fusion. Since multi-source sensors have differentiated characteristics (such as range and response speed), feature vectors need to be uniformly calibrated. Calibrate and limit all characteristic components to the interval [0,1] to obtain the calibrated characteristic components .

[0033] In this embodiment, the maximum and minimum normalization formula is as follows: ; in, and are the minimum and maximum values in the feature set, respectively.

[0034] S3. Perform weighted fusion on the calibrated feature components to generate a fault indicator.

[0035] By using the weighted fusion formula and optimizing the weighted coefficients, fault indicators are generated to achieve a comprehensive evaluation of multi-source features. The weighted fusion formula is as follows: ; in, is the fault indication quantity; For the The weighting coefficients of the feature components are calculated by the feature variance: ; Where, For the The variance of the calibrated feature components; For the The calibrated value of the characteristic components, The sum of the variances after calibration for all feature components; The weighted fusion formula is used to ensure that the weights of feature components are proportional to the amplitude of signal fluctuations, so that features that are sensitive to changes have higher weights.

[0036] In this embodiment, a preset threshold is designed , the calculated fault indication quantity With preset threshold Compare: like , it is considered that there is a fault and enters the fault detection phase. like , the system is in normal state.

[0037] Through this weighted fusion method, the feature sensitivity and fusion robustness are effectively improved, and a seamless connection from the feature layer to the indicator layer is achieved.

[0038] S4. Build a fault detection model based on the fault indicator to determine the fault type and its probability.

[0039] This step uses a long short-term memory (LSTM) network as the core model for fault detection. LSTM networks are highly capable of modeling time series and effectively capture the temporal characteristics of changes in indicated quantities. Given the non-stationary and time-varying nature of industrial instrument data, LSTM networks can effectively address these characteristics and accurately identify fault modes.

[0040] The fault detection model is constructed by: Model input: fault indicator Time series , is the time window length; due to the differences in the working conditions of different instruments, the sliding window technology is used to enhance the sample size and time sensitivity.

[0041] Model output: probability distribution of various types of failures , For the The probability of class failure.

[0042] The fault detection model has a built-in loss function and optimization algorithm to train and optimize the model.

[0043] The loss function uses the cross entropy loss function to measure the deviation between the predicted value and the true value.

[0044] In this embodiment, the loss function is as follows: ; in, is the actual label, is the predicted value, is the sample size.

[0045] The optimization algorithm uses the existing Adam optimization algorithm to adjust the model weights to ensure convergence speed and stability. The generalization performance of the model is evaluated through cross-validation to prevent overfitting.

[0046] Select the one with the highest probability , which is the maximum value among the probability values calculated for all possible fault types, represents the most confident prediction category of the model. In practical applications, if the highest probability If the probability of the prediction is higher than the set threshold, the system can consider the prediction result to be highly credible and directly use the fault type as the final judgment; if the highest probability If the value falls below the threshold, the model prediction is considered unreliable, triggering manual review or backup strategies. This can effectively reduce the probability of false positives and missed detections, and improve the security and stability of system operations.

[0047] S5. After outputting the fault category and probability, the fault severity calculation model is used to quantify the fault severity.

[0048] Since different faults have different impacts on system stability and security, the fault severity is quantified. A comprehensive severity score is obtained by weighted evaluation of multiple feature indicators. , indicating the extent of the impact of the fault on the instrument or system.

[0049] Fault severity calculation model, including: Construct severity index, select the index that is representative of the fault, and the membership function of each index Indicates the indicator's contribution to severity: ; in, For the indicator values, is the membership function; Construct severity evaluation formula: ; in, For the The risk level value corresponding to each indicator; The fault severity is divided into different levels, including minor faults, general faults, serious faults and critical faults.

[0050] In this embodiment, the levels are divided into: Minor faults: ; General faults: ; Serious failure: ; Critical failure: .

[0051] S6. Based on the fault severity, formulate corresponding calibration strategies for different levels of fault severity.

[0052] Calibration strategies include: Zero offset calibration is suitable for minor faults, such as chronic drift faults. This calibration is usually inexpensive but relies on reference redundant channels or historical baseline values to eliminate zero drift by adjusting the sensor reference point.

[0053] Parameter compensation calibration: for general and serious faults, adjust the instrument sensitivity and linearity through compensation coefficient.

[0054] Model reconstruction and calibration: In the event of a critical fault, direct calibration is difficult to recover. A model reconstruction strategy is adopted to reconstruct the instrument response model through historical data and real-time data, and update the calibration parameters and characteristic curves.

[0055] Because different calibration strategies have significant differences in resource consumption (such as time, manpower, and replacement components) and accuracy recovery capabilities, this step does not adopt a static strategy matching mechanism. Instead, it introduces a strategy selection mechanism based on an optimization goal. The optimization goal is to balance calibration accuracy and calibration cost: ; in, : Overall optimization objective function, representing the expected calibration cost or system cost; :Corresponding to The probability of determining the level of failure (can be set by empirical data, historical statistics or operational estimation); : No. The calibration strategy cost corresponding to each fault level (can also be set by empirical data, historical statistics or operational estimation).

[0056] This optimization model is essentially an expected cost minimization model, combined with Bayesian thinking. That is, given the probability of each fault level occurring, Under these circumstances, the optimal calibration strategy is selected for each fault level to minimize the overall expected cost. For example: If a higher-cost reconstruction calibration strategy corresponds to Lower, the model will tend not to trigger first; If a medium-cost strategy corresponds to If it is higher, the strategy will receive a higher weight in the optimization model; The system can be updated dynamically Or based on feedback mechanism correction , to achieve online policy scheduling.

[0057] S7. Dynamically evaluate the effectiveness of the calibration strategy.

[0058] Dynamic evaluation of calibration strategies, including: Calibration effect verification: The calibration error change value is used for evaluation, that is, the error change before and after calibration is compared. If the error is greatly reduced, it indicates that the calibration strategy has a significant effect. The calibration error change value is: , and To ensure the accuracy of the evaluation, it is recommended to perform multiple measurements after calibration and take the average value as the final post-calibration error value to avoid accidental factors affecting the evaluation results.

[0059] Stability analysis: The stability coefficient is used to measure the consistency of the calibration results in long-term operation, which is used to effectively judge the reliability of the calibration strategy. In actual operation, a time window is selected. Perform stability monitoring and calculate the deviation rate between the calibrated signal and the true signal to obtain the stability coefficient. If the stability coefficient is lower than the set threshold, the calibration effect is considered poor and the calibration strategy or parameters need to be readjusted.

[0060] Because industrial instrumentation operates in complex and frequently changing environments, even after calibration, instrument performance can be disrupted by external factors. To address this, the system monitors calibration results in real time, collecting and analyzing instrument output signals to dynamically determine whether calibration is maintaining effectiveness. If the system detects a gradual decline in calibration effectiveness, it automatically issues an alarm and prompts the operator to perform manual calibration or initiate a backup strategy.

[0061] S8. Jointly optimize the constructed model and calibration strategy.

[0062] The first task of joint optimization is to clarify the optimization goal, that is, to find the best balance between accuracy and stability. To achieve this goal, the accuracy of the calibration model and the stability of the instrument signal after calibration are comprehensively considered.

[0063] To prevent the model from falling into a local optimum, a genetic algorithm is used to fine-tune model parameters. Crossover and mutation operations are introduced during the optimization process to enhance population diversity and global search capabilities, thereby improving the algorithm's convergence accuracy and robustness. Each generation of optimization results is screened and eliminated to ensure that the final model performs well under complex working conditions.

[0064] In genetic algorithms, crossover swaps genes from two individuals with high fitness to create a new individual that inherits the favorable traits of its parent. Mutation, on the other hand, introduces new possible solutions by perturbing genetic loci within a small range. The ratio of crossover to mutation needs to be dynamically adjusted and automatically optimized based on the algorithm's convergence speed and fitness distribution to avoid premature convergence and population degradation.

[0065] To ensure optimization effectiveness, the fitness of individuals in the population is evaluated after each generation of the algorithm. Evaluation metrics include model detection accuracy, calibration stability, and system response speed. The evaluation results are ranked, and the top-performing individuals are selected as the basis for the next generation of genetic operations, gradually approaching the global optimal solution. In practical applications, overfitting and degradation of generalization performance must also be prevented, especially in noisy industrial environments. Cross-validation is used to further verify model robustness.

[0066] Due to the complexity of industrial instrumentation applications, model and strategy parameters cannot remain static. Therefore, this step proposes an adaptive parameter adjustment mechanism to dynamically update model weights and optimization algorithm parameters based on real-time operating status and calibration results. By periodically reviewing historical calibration results, optimization strategy parameters, such as the crossover and mutation probabilities in the genetic algorithm, are automatically adjusted to ensure the model's adaptability and stability over the long term.

[0067] Example 2: This embodiment also provides a computer device, which is suitable for an industrial instrument fault detection and self-calibration method based on intelligent sensor fusion, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement an industrial instrument fault detection and self-calibration method based on intelligent sensor fusion as proposed in the above embodiment.

[0068] This embodiment further provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion proposed in the above embodiment is implemented.

[0069] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0070] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0071] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0072] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0073] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0074] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion, characterized in that: The following steps are involved: S1. Extract multi-source signal features from sensor signals in industrial instrumentation systems; S2. performing consistency calibration on the extracted multi-source signal features to obtain calibrated feature components; S3. Perform weighted fusion on the calibrated characteristic components to generate a fault indicator; S4. Build a fault detection model based on the fault indicator to determine the fault type and its probability; S5. After outputting the fault category and probability, quantify the fault severity using the fault severity calculation model; S6. Based on the fault severity, formulate corresponding calibration strategies for different levels of fault severity; S7. Dynamically evaluate the effectiveness of the calibration strategy; S8. Jointly optimize the constructed model and calibration strategy.

2. The method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion according to claim 1 is characterized in that: In step S1, multi-source signal feature extraction includes time domain feature extraction, frequency domain feature extraction and time-frequency feature extraction, extracting signal characteristics from different angles, and the expression of the comprehensive feature vector is: ； in, is a comprehensive feature vector, which is a high-dimensional feature vector used to characterize the sensor signal; Indicates the feature components.

3. The method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion according to claim 1 is characterized in that: In step S2, each feature component is normalized using the maximum and minimum normalization. Calibrate and limit all characteristic components to the interval [0,1] to obtain the calibrated characteristic components .

4. The method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion according to claim 1 is characterized in that: In step S3, a weighted fusion formula is used to generate fault indicators by optimizing weighting coefficients to achieve a comprehensive evaluation of multi-source features; The weighted fusion formula is as follows: ； in, is the fault indication quantity; For the The weighting coefficients of the feature components are calculated by the feature variance: ； Where, For the The variance of the calibrated feature components; For the The calibrated value of the characteristic components, The sum of the variances after calibration for all feature components; The weighted fusion formula is used to ensure that the weights of feature components are proportional to the amplitude of signal fluctuations.

5. The method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion according to claim 1 is characterized in that: In step S4, the fault detection model includes: Model input: fault indicator Time series , is the time window length; Model output: probability distribution of various types of failures , For the The probability of class failure.

6. The method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion according to claim 5 is characterized in that: The fault detection model has a built-in loss function and optimization algorithm to train and optimize the model; The loss function uses the cross entropy loss function to measure the deviation between the predicted value and the true value; The optimization algorithm uses the Adam optimization algorithm to adjust the model weights and evaluates the generalization performance of the model through cross-validation to prevent overfitting.

7. The method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion according to claim 1 is characterized in that: In step S5, the fault severity calculation model includes: Construct severity index, select the index that is representative of the fault, and the membership function of each index Indicates the indicator's contribution to severity: ； in, For the indicator values, is the membership function; Construct severity evaluation formula: ； in, For the The risk level value corresponding to each indicator; The fault severity is divided into different levels, including minor faults, general faults, serious faults and critical faults.

8. The method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion according to claim 1 is characterized in that: In step S6, the calibration strategy includes: Zero offset calibration; Parameter compensation calibration; Model reconstruction and calibration.

9. The method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion according to claim 1, characterized in that: In step S7, the dynamic evaluation of the calibration strategy includes: Calibration effect verification: The calibration error change value is used for evaluation, that is, the error change before and after calibration is compared. If the error is greatly reduced, it indicates that the calibration strategy has a significant effect. The calibration error change value is: , and are the error values before and after calibration respectively; Stability analysis: The stability coefficient is used to measure the consistency of calibration results in long-term operation and to effectively judge the reliability of the calibration strategy.

10. The method for industrial instrument fault detection and self-calibration based on intelligent sensor fusion according to claim 1, characterized in that: In step S8, the goal of the joint optimization is to find the best balance between accuracy and stability, taking into account the accuracy of the model and the stability of the instrument signal after calibration. In actual operation, the model parameters are tuned and dynamically updated through a genetic algorithm.

Images