A method for self-calibration and abnormal compensation of multi-source sensors of a power plant inspection robot

Abstract

This invention relates to the field of data processing and discloses a method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot. The method includes: selecting stable physical signals existing in the thermal power plant environment as calibration reference signals; constructing a hierarchical dynamic reference database; calculating calibration parameters in real time; comparing the calibration parameters with a preset threshold; if the calibration parameters exceed the preset threshold, determining that the real-time data from the multi-source sensors is abnormal and deleting it; otherwise, dynamically supplementing the real-time data from the multi-source sensors using the calibration parameters to obtain calibrated sensor data; interpolating the deleted abnormal data using a preset interpolation method and combining it with the calibrated sensor data to obtain the final multi-source sensor data for the thermal power plant inspection robot. This invention, combined with the online retrieval and self-evolutionary update mechanism of the hierarchical dynamic reference database, achieves real-time and accurate compensation for sensor drift.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot. Background Technology

[0002] Thermal power plants, as typical complex industrial environments, are characterized by high temperatures, high dust levels, strong electromagnetic interference, and dense equipment. Inspection robots need to be equipped with various sensors, such as visible light cameras, thermal imagers, microphone arrays, lidar, IMUs, magnetic compasses, and gas sensors, to perform tasks such as equipment status monitoring, leak detection, and temperature measurement. However, during long-term operation, sensors can experience measurement drift due to factors such as changes in ambient temperature, mechanical vibration, dust pollution, and electromagnetic interference, leading to decreased data accuracy and directly affecting the reliability of equipment fault diagnosis.

[0003] Currently, the calibration methods for multi-source sensors of inspection robots mainly have the following technical problems: Traditional sensor calibration relies on manual periodic calibration or external standard parts, which is not only time-consuming and labor-intensive, affecting the continuity of inspection, but also difficult to adapt to the time-varying characteristics of sensor drift in the complex environment of thermal power plants, and cannot achieve online self-calibration.

[0004] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a method for self-calibration and anomaly compensation of multi-source sensors in thermal power plant inspection robots. This method has the advantage of differentiated online self-calibration of multi-source sensors, thereby solving the problem of low efficiency caused by reliance on manual calibration in existing technologies.

[0006] (II) Technical Solution To achieve the advantages of differentiated online self-calibration of multi-source sensors mentioned above, the specific technical solution adopted in this invention is as follows: According to one aspect of the present invention, a method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot is provided, comprising the following steps: S1. Based on the inspection task of the thermal power plant inspection robot, select the stable physical signals that exist in the thermal power plant environment as the calibration reference signals. S2. Based on the multi-source sensors carried by the thermal power plant inspection robot, and according to the correlation characteristics between the sensed physical quantities and the calibration reference signals, the multi-source sensors are classified. Based on the classification results, the correlation mapping relationship between each type of sensor and the calibration reference signal is established, and a hierarchical dynamic reference database is constructed. S3. Collect real-time data from multiple sensors of the inspection robot in the thermal power plant during the inspection process, and use a hierarchical dynamic benchmark database to calibrate and compare the real-time data of multiple sensors. Based on the calibration and comparison results, calculate the calibration parameters in real time. S4. Compare the calibration parameters with the preset threshold. If the calibration parameters are greater than the preset threshold, determine that the real-time data of the multi-source sensor is abnormal and delete it. Otherwise, use the calibration parameters to dynamically supplement the real-time data of the multi-source sensor to obtain the calibrated sensor data. S5. The deleted abnormal data is interpolated using a preset interpolation method, and combined with the calibrated sensor data to obtain the final multi-source sensor data of the thermal power plant inspection robot.

[0007] Preferably, the inspection task based on the thermal power plant inspection robot, selecting a stable physical signal in the thermal power plant environment as the calibration reference signal includes the following steps: S11. Based on the inspection tasks of the thermal power plant inspection robot, analyze the inspection path of the thermal power plant inspection robot, and select physical signals that meet the signal stability indicators as candidate reference signals in combination with the preset signal stability indicators. S12. Conduct a measurability assessment of the candidate reference signals, and combine the spatial gradient of each candidate reference signal on the inspection path to conduct a correlation feasibility assessment of the candidate reference signals. S13. Based on the combined results of measurability assessment, spatial gradient analysis, and correlation feasibility assessment, select a calibration reference signal from the candidate reference signals.

[0008] Preferably, the preset signal stability indicators include signal amplitude fluctuation rate, phase drift, and signal repeatability coefficient.

[0009] Preferably, the process of acquiring the multi-source sensors carried by the thermal power plant inspection robot, classifying the multi-source sensors according to the correlation characteristics between the sensed physical quantities and calibration reference signals, establishing the correlation mapping relationship between each type of sensor and the calibration reference signal based on the classification results, and constructing a hierarchical dynamic reference database includes the following steps: S21. Based on the multi-source sensors carried by the thermal power plant inspection robot, determine the type of physical sensing quantity of each sensor. S22. Based on the physical sensing type of each sensor and combined with the correlation characteristics of the calibration reference signal, the multi-source sensors are classified to obtain a multi-source sensor classification system. S23. For each type of sensor, analyze the inherent physical mechanism between its physical sensing quantity and the corresponding calibration reference signal, and establish a correlation mapping relationship; S24. Based on the multi-source sensor classification system and the established corresponding mapping relationship, construct a hierarchical dynamic benchmark database.

[0010] Preferably, the process of classifying multi-source sensors based on the physical sensing type of each sensor and the correlation characteristics of the calibration reference signal to obtain a multi-source sensor classification system includes the following steps: S221. Based on the physical sensing type of each sensor, determine the measurement dimension corresponding to each sensor, and determine the correlation characteristics between each sensor and each calibration reference signal by extracting the physical attributes of the calibration reference signal. S222. Based on the strength of the correlation characteristics, the sensors are classified into single sensors, composite sensors, and independent sensors. S223. For the single sensor, composite sensor and independent sensor that have been classified, match the corresponding calibration reference signal and associated verification rules respectively, and after completing the calibration reference matching of all sensors, form the final multi-source sensor classification system.

[0011] Preferably, the matching of corresponding calibration reference signals and correlation verification rules for the divided single sensor, composite sensor, and independent sensor includes: Match a single sensor with a single calibration reference signal that is physically homologous to its physical sensing quantity, and set a direct comparison verification rule based on a preset deviation threshold; To match the composite sensor with multiple calibration reference signals that are related to its physical sensing quantity, and to set consistency verification rules based on multi-source information fusion; Match independent sensors with calibration reference signals indirectly associated with calibrated sensors, and set indirect verification rules based on a preset multi-sensor fusion mechanism.

[0012] Preferably, the process of collecting real-time data from multiple sensors during the inspection of a thermal power plant inspection robot, and using a hierarchical dynamic benchmark database to calibrate and compare the real-time data from the multiple sensors, and calculating the calibration parameters in real time based on the calibration and comparison results, includes the following steps: S31. Collect multi-source sensor data in real time during the robot's inspection task, and perform spatiotemporal alignment processing on the multi-source sensor data to obtain a spatiotemporally synchronized multimodal data stream. S32. Based on the spatiotemporally synchronized multimodal data stream and hierarchical dynamic benchmark database, and combined with the verification rules set for different sensor types, calculate the deviation and confidence level between the measured values ​​of each sensor and the benchmark signal respectively. S33. Based on the calculated deviation and confidence level, and combined with the correlation mapping relationship, the calibration parameters of each sensor are calculated in reverse using the least squares estimation method.

[0013] Preferably, the calculation of the deviation and confidence level between the measured values ​​of each sensor and the reference signal based on the spatiotemporally synchronized multimodal data stream and hierarchical dynamic benchmark database, combined with the verification rules set for different sensor types, includes the following steps: S321. Based on the established multi-source sensor classification system, identify the sensor type of the sensor to be calibrated. S322. Based on the identified sensor type and the corresponding verification rules, calculate the deviation between the multimodal data stream of the sensor's spatiotemporal synchronization and the reference signal. S323. Calculate the confidence level of the current comparison result based on the reliability of the data source, the type of verification rule, and the stability of historical calibration records used in the process of calculating the deviation.

[0014] Preferably, the step of back-calculating the calibration parameters of each sensor using the least squares estimation method based on the calculated deviation and confidence level, combined with the correlation mapping relationship, includes the following steps: S331. For each type of sensor, construct an observation equation with calibration parameters as unknowns, based on the type of its associated mapping relationship. S332. Using the deviation as the observed value and the corresponding confidence level as the weighting factor, construct a weighted least squares objective function, and obtain the optimal estimate of the calibration parameters by solving for the minimum value of the objective function. S333. Evaluate the confidence interval of the optimal estimated value of the calibration parameter obtained by solving. When the variance of the parameter estimate is less than the preset variance threshold, output the calibration parameter. When the variance of the parameter estimate is greater than or equal to the preset variance threshold, use the incremental data acquisition mechanism to increase the number of observation samples and re-execute step S332 until the estimated value of the calibration parameter meets the preset confidence requirements.

[0015] According to another aspect of the present invention, a multi-source sensor self-calibration and anomaly compensation system for a thermal power plant inspection robot is provided, the system comprising: The reference signal selection module is used to select stable physical signals in the thermal power plant environment as calibration reference signals for the inspection tasks of the thermal power plant inspection robot. The database construction module is used to acquire the multi-source sensors carried by the thermal power plant inspection robot, classify the multi-source sensors according to the correlation characteristics between the sensed physical quantities and the calibration reference signals, establish the correlation mapping relationship between each type of sensor and the calibration reference signal based on the classification results, and construct a hierarchical dynamic reference database. The parameter calculation module is used to collect real-time data from multiple sensors during the inspection process of the thermal power plant inspection robot, and to use a hierarchical dynamic benchmark database to calibrate and compare the real-time data from multiple sensors. Based on the calibration and comparison results, the calibration parameters are calculated in real time. The data calibration module is used to compare calibration parameters with preset thresholds. When the calibration parameters are greater than the preset thresholds, the real-time data of the multi-source sensors is determined to be abnormal data and deleted. Otherwise, the calibration parameters are used to dynamically supplement the real-time data of the multi-source sensors to obtain calibrated sensor data. The data interpolation module is used to interpolate the deleted abnormal data using a preset interpolation method, and combine it with the calibrated sensor data to obtain the final multi-source sensor data of the thermal power plant inspection robot.

[0016] (III) Beneficial Effects Compared with existing technologies, this invention provides a method for self-calibration and anomaly compensation of multi-source sensors in thermal power plant inspection robots, which has the following beneficial effects: (1) This invention achieves real-time and accurate compensation for sensor drift by finely classifying multi-source sensors and constructing physical model-driven or data-driven association mapping relationships in a differentiated manner, combined with the online retrieval and self-evolution update mechanism of the hierarchical dynamic benchmark database.

[0017] (2) This invention selects the final reference signal by comprehensively considering the three evaluation results, forming a complete closed loop from stability screening to engineering feasibility verification to spatial distinguishability evaluation, which significantly improves the selection quality and engineering applicability of calibration reference signals and effectively avoids the problem of calibration error accumulation or calibration failure caused by improper selection of reference signals. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a method for self-calibration and anomaly compensation of multi-source sensors for a thermal power plant inspection robot according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the self-calibration and anomaly compensation system for multi-source sensors of a thermal power plant inspection robot according to an embodiment of the present invention.

[0020] In the picture: 1. Reference signal selection module; 2. Database construction module; 3. Parameter calculation module; 4. Data calibration module; 5. Data interpolation module. Detailed Implementation

[0021] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0022] According to an embodiment of the present invention, a method for self-calibration and anomaly compensation of multi-source sensors of a thermal power plant inspection robot is provided.

[0023] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to an embodiment of the present invention, a method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot is provided, comprising the following steps: S1. Based on the inspection task of the thermal power plant inspection robot, select the stable physical signals that exist in the thermal power plant environment as the calibration reference signals. In a preferred embodiment, the inspection task based on the thermal power plant inspection robot, which selects a stable physical signal in the thermal power plant environment as a calibration reference signal, includes the following steps: S11. Based on the inspection tasks of the thermal power plant inspection robot, analyze the inspection path of the thermal power plant inspection robot, and select physical signals that meet the signal stability indicators as candidate reference signals in combination with the preset signal stability indicators. Specifically, stable physical signals in a thermal power plant environment include, but are not limited to: power frequency magnetic fields radiated by high-voltage buses and transformers; surface thermal field distributions formed by equipment under stable operating conditions; continuous environmental noise spectra generated by large equipment such as fans and steam turbines; stable atmospheric pressure field distributions that vary with altitude; and stable reflectivity of fixed material surfaces to lidar. These signals share common characteristics: spatial distribution differences along the inspection path; long-term stability or predictable variation patterns over time; and the ability to be detected by existing sensors mounted on the robot.

[0024] S12. Conduct a measurability assessment of the candidate reference signals, and combine the spatial gradient of each candidate reference signal on the inspection path to conduct a correlation feasibility assessment of the candidate reference signals. The measurability assessment aims to verify whether candidate reference signals are detectable within the measurement range of existing sensors mounted on the robot. Specifically, for each candidate reference signal, its amplitude range, frequency characteristics, spatial distribution range, and other parameters are analyzed and matched with the measurement range, sensitivity, resolution, and other indicators of the sensors mounted on the robot. For example, the power frequency magnetic field strength is typically in the range of 0.1 μT to 10 μT. Spatial gradient analysis aims to ensure that candidate reference signals have distinguishable characteristic differences at different spatial locations, thereby providing a valid reference for the robot's spatial positioning and sensor calibration. Specifically, for each candidate reference signal, its spatial gradient, i.e., the amount of signal change per unit distance, is calculated along the inspection path. The correlation feasibility assessment aims to determine whether there is a clear physical mapping relationship between the candidate reference signal and the corresponding sensor output.

[0025] S13. Based on the combined results of measurability assessment, spatial gradient analysis, and correlation feasibility assessment, select a calibration reference signal from the candidate reference signals.

[0026] In a preferred embodiment, the preset signal stability indicators include signal amplitude fluctuation rate, phase drift, and signal repeatability coefficient.

[0027] S2. Based on the multi-source sensors carried by the thermal power plant inspection robot, and according to the correlation characteristics between the sensed physical quantities and the calibration reference signals, the multi-source sensors are classified. Based on the classification results, the correlation mapping relationship between each type of sensor and the calibration reference signal is established, and a hierarchical dynamic reference database is constructed. In a preferred embodiment, the process of acquiring the multi-source sensors carried by the thermal power plant inspection robot, classifying the multi-source sensors according to the correlation characteristics between the sensed physical quantities and calibration reference signals, establishing the correlation mapping relationship between each type of sensor and the calibration reference signal based on the classification results, and constructing a hierarchical dynamic reference database includes the following steps: S21. Based on the multi-source sensors carried by the thermal power plant inspection robot, determine the type of physical sensing quantity of each sensor. The types of physical sensing quantities include, but are not limited to, the following categories: attitude and position, vision and imaging, acoustics and vibration, lidar, and environmental parameters.

[0028] S22. Based on the physical sensing type of each sensor and combined with the correlation characteristics of the calibration reference signal, the multi-source sensors are classified to obtain a multi-source sensor classification system. In a preferred embodiment, the process of classifying multi-source sensors based on the type of physical sensing quantity of each sensor and the correlation characteristics of the calibration reference signal to obtain a multi-source sensor classification system includes the following steps: S221. Based on the physical sensing type of each sensor, determine the measurement dimension corresponding to each sensor, and determine the correlation characteristics between each sensor and each calibration reference signal by extracting the physical attributes of the calibration reference signal. S222. Based on the strength of the correlation characteristics, the sensors are classified into single sensors, composite sensors, and independent sensors. In this context, a single sensor refers to a sensor that has a strong correlation with one and only one calibration reference signal. A composite sensor refers to a sensor that has varying degrees of correlation with multiple calibration reference signals. An independent sensor refers to a sensor that has no direct physical correlation or a very weak correlation with any of the selected calibration reference signals.

[0029] S223. For the single sensor, composite sensor and independent sensor that have been classified, match the corresponding calibration reference signal and associated verification rules respectively, and after completing the calibration reference matching of all sensors, form the final multi-source sensor classification system.

[0030] In a preferred embodiment, the step of matching the corresponding calibration reference signal and correlation verification rules for the divided single sensor, composite sensor, and independent sensor respectively includes: Match a single sensor with a single calibration reference signal that is physically homologous to its physical sensing quantity, and set a direct comparison verification rule based on a preset deviation threshold; To match the composite sensor with multiple calibration reference signals that are related to its physical sensing quantity, and to set consistency verification rules based on multi-source information fusion; Match independent sensors with calibration reference signals indirectly associated with calibrated sensors, and set indirect verification rules based on a preset multi-sensor fusion mechanism.

[0031] It should be noted that the multi-sensor fusion mechanism is constructed using a gray-box modeling method that combines physical mechanisms with data-driven approaches. Specifically, based on the spatial colocation relationship or physical coupling mechanism between the sensor and the calibrated device, a semi-parametric framework with clear physical meaning is constructed. Known physical equations, such as gas diffusion equations, heat conduction equations, and vibration transfer functions, serve as the main structure of the multi-sensor fusion mechanism. Complex factors that are difficult to model accurately are compensated for through sparse parameterization or Gaussian process residual terms. All input variables of the multi-sensor fusion mechanism come from interpretable physical quantities output by the calibrated sensors, such as temperature, vibration amplitude, and frequency characteristics. The output of the multi-sensor fusion mechanism is the confidence interval of the expected value of each independent sensor, rather than a single value. During the calibration process, the input, output, and intermediate calculation processes are recorded simultaneously to form a traceable calibration evidence chain. When the indirect verification rule triggers anomaly judgment, the state or parameter changes of the specific input sensor can be located by tracing back the evidence chain.

[0032] S23. For each type of sensor, analyze the inherent physical mechanism between its physical sensing quantity and the corresponding calibration reference signal, and establish a correlation mapping relationship; S24. Based on the multi-source sensor classification system and the established corresponding mapping relationship, construct a hierarchical dynamic benchmark database.

[0033] It should be noted that the hierarchical dynamic benchmark database includes: a benchmark signal layer, a sensor feature layer, and a mapping relationship layer. The benchmark signal layer stores benchmark signal values ​​at various spatial locations and under different operating conditions. The sensor feature layer stores historical measurement characteristics of each sensor at its corresponding location, providing comparative data for the dynamic updating of the benchmark database. The mapping relationship layer stores the association mapping parameters and confidence levels between each type of sensor and its corresponding benchmark signal.

[0034] S3. Collect real-time data from multiple sensors of the inspection robot in the thermal power plant during the inspection process, and use a hierarchical dynamic benchmark database to calibrate and compare the real-time data of multiple sensors. Based on the calibration and comparison results, calculate the calibration parameters in real time. In a preferred embodiment, the process of collecting real-time data from multiple sensors of the power plant inspection robot during the inspection process, and using a hierarchical dynamic benchmark database to calibrate and compare the real-time data from the multiple sensors, and calculating the calibration parameters in real time based on the calibration and comparison results, includes the following steps: S31. Collect multi-source sensor data in real time during the robot's inspection task, and perform spatiotemporal alignment processing on the multi-source sensor data to obtain a spatiotemporally synchronized multimodal data stream. S32. Based on the spatiotemporally synchronized multimodal data stream and hierarchical dynamic benchmark database, and combined with the verification rules set for different sensor types, calculate the deviation and confidence level between the measured values ​​of each sensor and the benchmark signal respectively. In a preferred embodiment, the calculation of the deviation and confidence level between the measured values ​​of each sensor and the reference signal based on the spatiotemporally synchronized multimodal data stream and hierarchical dynamic benchmark database, combined with the verification rules set for different sensor types, includes the following steps: S321. Based on the established multi-source sensor classification system, identify the sensor type of the sensor to be calibrated. S322. Based on the identified sensor type and the corresponding verification rules, calculate the deviation between the multimodal data stream of the sensor's spatiotemporal synchronization and the reference signal. It should be noted that, for a single sensor, a matching single calibration reference signal is retrieved from the hierarchical dynamic reference database, and the sensor's measured value is directly compared with the theoretical value of the reference signal to calculate the absolute or relative deviation as the deviation amount; for a composite sensor, multiple matching calibration reference signals are retrieved from the hierarchical dynamic reference database, the preliminary deviation between the sensor's measured value and each reference signal is calculated, and the comprehensive deviation is calculated using a weighted fusion method as the deviation amount; for an independent sensor, calibration reference signals indirectly associated with calibrated sensors are retrieved from the hierarchical dynamic reference database, and the output of the calibrated sensor is converted into the expected value of the independent sensor using a preset multi-sensor fusion mechanism, and the deviation between the measured value and the expected value is calculated as the deviation amount.

[0035] S323. Calculate the confidence level of the current comparison result based on the reliability of the data source, the type of verification rule, and the stability of historical calibration records used in the process of calculating the deviation.

[0036] It should be noted that, for a single sensor, the confidence level is calculated using a product fusion method based on the spatial distribution confidence level of the reference signal, the historical calibration stability coefficient of the sensor, and the signal-to-noise ratio of the measured data; for a composite sensor, the overall confidence level is calculated based on the individual confidence levels of each reference signal and the consistency index in the weighted fusion process; for an independent sensor, the confidence level of indirect calibration is obtained by passing the confidence level of the calibrated sensor it relies on and the uncertainty quantification index.

[0037] S33. Based on the calculated deviation and confidence level, and combined with the correlation mapping relationship, the calibration parameters of each sensor are calculated in reverse using the least squares estimation method.

[0038] In a preferred embodiment, the step of back-calculating the calibration parameters of each sensor using the least squares estimation method based on the calculated deviation and confidence level, combined with the correlation mapping relationship, includes the following steps: S331. For each type of sensor, construct an observation equation with calibration parameters as unknowns, based on the type of its associated mapping relationship. It should be noted that for physics-model-driven sensors, the observation equations are constructed based on their physical models. For example, for a thermal imager, its temperature response model can be expressed as: ,in, This is the proportionality coefficient. To achieve zero bias, the calibration parameter is... and At a certain inspection point, based on the theoretical temperature in the benchmark database... and sensor measured temperature Observation equations can be constructed. ,in, This represents the residual. For multiple inspection points, multiple observation equations can be formed, constituting a system of equations.

[0039] S332. Using the deviation as the observed value and the corresponding confidence level as the weighting factor, construct a weighted least squares objective function, and obtain the optimal estimate of the calibration parameters by solving for the minimum value of the objective function.

[0040] S333. Evaluate the confidence interval of the optimal estimated value of the calibration parameter obtained by solving. When the variance of the parameter estimate is less than the preset variance threshold, output the calibration parameter. When the variance of the parameter estimate is greater than or equal to the preset variance threshold, use the incremental data acquisition mechanism to increase the number of observation samples and re-execute step S332 until the estimated value of the calibration parameter meets the preset confidence requirements.

[0041] Specifically, the variance threshold is determined based on the repeatability accuracy index in the sensor's technical specifications. For example, for the temperature zero bias parameter of a thermal imager, the variance threshold is set to (0.1℃). 2 For the azimuth correction parameter of the magnetic compass, the variance threshold is set to (0.5°). 2 .

[0042] S4. Compare the calibration parameters with the preset threshold. If the calibration parameters are greater than the preset threshold, determine that the real-time data of the multi-source sensor is abnormal and delete it. Otherwise, use the calibration parameters to dynamically supplement the real-time data of the multi-source sensor to obtain the calibrated sensor data. Specifically, the calibration parameters are mainly the zero offset compensation amount or the proportional coefficient correction value. The preset threshold is set according to the maximum permissible error of the sensor calibrated at the factory. For example, for a thermal imager, the threshold for the zero offset compensation amount is set to ±2℃; for a magnetic compass, the threshold for the azimuth correction amount is set to ±1°; and for a barometric altimeter, the threshold for the zero altitude offset is set to ±0.5m.

[0043] S5. The deleted abnormal data is interpolated using a preset interpolation method, and combined with the calibrated sensor data to obtain the final multi-source sensor data of the thermal power plant inspection robot.

[0044] It should be noted that the preset interpolation methods include linear interpolation, cubic spline interpolation, or time series prediction-based methods, etc. The interpolated data and the calibrated sensor data are combined to obtain the final multi-source sensor data of the thermal power plant inspection robot.

[0045] like Figure 2 As shown, according to another embodiment of the present invention, a multi-source sensor self-calibration and anomaly compensation system for a thermal power plant inspection robot is provided. The system includes: The reference signal selection module 1 is used to select stable physical signals in the thermal power plant environment as calibration reference signals based on the inspection tasks of the thermal power plant inspection robot. Database construction module 2 is used to acquire the multi-source sensors carried by the thermal power plant inspection robot, classify the multi-source sensors according to the correlation characteristics between the sensed physical quantities and the calibration reference signals, establish the correlation mapping relationship between each type of sensor and the calibration reference signal based on the classification results, and construct a hierarchical dynamic reference database. The parameter calculation module 3 is used to collect real-time data from multiple sensors during the inspection process of the thermal power plant inspection robot, and to use a hierarchical dynamic benchmark database to calibrate and compare the real-time data from multiple sensors. Based on the calibration and comparison results, the calibration parameters are calculated in real time. Data calibration module 4 is used to compare calibration parameters with preset thresholds. When the calibration parameters are greater than the preset thresholds, the real-time data of the multi-source sensor is determined to be abnormal data and deleted. Otherwise, the calibration parameters are used to dynamically supplement the real-time data of the multi-source sensor to obtain calibrated sensor data. The data interpolation module 5 is used to interpolate the deleted abnormal data using a preset interpolation method, and combine it with the calibrated sensor data to obtain the final multi-source sensor data of the thermal power plant inspection robot.

[0046] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot, characterized in that, The method includes the following steps: S1. Based on the inspection task of the thermal power plant inspection robot, select the stable physical signals that exist in the thermal power plant environment as the calibration reference signals. S2. Based on the multi-source sensors carried by the thermal power plant inspection robot, and according to the correlation characteristics between the sensed physical quantities and the calibration reference signals, the multi-source sensors are classified. Based on the classification results, the correlation mapping relationship between each type of sensor and the calibration reference signal is established, and a hierarchical dynamic reference database is constructed. S3. Collect real-time data from multiple sensors of the inspection robot in the thermal power plant during the inspection process, and use a hierarchical dynamic benchmark database to calibrate and compare the real-time data of multiple sensors. Based on the calibration and comparison results, calculate the calibration parameters in real time. S4. Compare the calibration parameters with the preset threshold. If the calibration parameters are greater than the preset threshold, determine that the real-time data of the multi-source sensor is abnormal and delete it. Otherwise, use the calibration parameters to dynamically supplement the real-time data of the multi-source sensor to obtain the calibrated sensor data. S5. The deleted abnormal data is interpolated using a preset interpolation method, and combined with the calibrated sensor data to obtain the final multi-source sensor data of the thermal power plant inspection robot.

2. The method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot according to claim 1, characterized in that, The inspection task based on the thermal power plant inspection robot, which selects stable physical signals existing in the thermal power plant environment as calibration reference signals, includes the following steps: S11. Based on the inspection tasks of the thermal power plant inspection robot, analyze the inspection path of the thermal power plant inspection robot, and select physical signals that meet the signal stability indicators as candidate reference signals in combination with the preset signal stability indicators. S12. Conduct a measurability assessment of the candidate reference signals, and combine the spatial gradient of each candidate reference signal on the inspection path to conduct a correlation feasibility assessment of the candidate reference signals. S13. Based on the combined results of measurability assessment, spatial gradient analysis, and correlation feasibility assessment, select a calibration reference signal from the candidate reference signals.

3. The method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot according to claim 2, characterized in that, The preset signal stability indicators include signal amplitude fluctuation rate, phase drift, and signal repeatability coefficient.

4. The method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot according to claim 1, characterized in that, The process of acquiring the multi-source sensors carried by the thermal power plant inspection robot, classifying the multi-source sensors according to the correlation characteristics between the sensed physical quantities and calibration reference signals, establishing the correlation mapping relationship between each type of sensor and the calibration reference signal based on the classification results, and constructing a hierarchical dynamic reference database includes the following steps: S21. Based on the multi-source sensors carried by the thermal power plant inspection robot, determine the type of physical sensing quantity of each sensor. S22. Based on the physical sensing type of each sensor and combined with the correlation characteristics of the calibration reference signal, the multi-source sensors are classified to obtain a multi-source sensor classification system. S23. For each type of sensor, analyze the inherent physical mechanism between its physical sensing quantity and the corresponding calibration reference signal, and establish a correlation mapping relationship; S24. Based on the multi-source sensor classification system and the established corresponding mapping relationship, construct a hierarchical dynamic benchmark database.

5. The method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot according to claim 4, characterized in that, The process of classifying multi-source sensors based on the type of physical sensing quantity of each sensor and the correlation characteristics of the calibration reference signal to obtain a multi-source sensor classification system includes the following steps: S221. Based on the physical sensing type of each sensor, determine the measurement dimension corresponding to each sensor, and determine the correlation characteristics between each sensor and each calibration reference signal by extracting the physical attributes of the calibration reference signal. S222. Based on the strength of the correlation characteristics, the sensors are classified into single sensors, composite sensors, and independent sensors. S223. For the single sensor, composite sensor and independent sensor that have been classified, match the corresponding calibration reference signal and associated verification rules respectively, and after completing the calibration reference matching of all sensors, form the final multi-source sensor classification system.

6. The method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot according to claim 5, characterized in that, The process of matching corresponding calibration reference signals and associated verification rules for the divided single sensor, composite sensor, and independent sensor includes: Match a single sensor with a single calibration reference signal that is physically homologous to its physical sensing quantity, and set a direct comparison verification rule based on a preset deviation threshold; To match the composite sensor with multiple calibration reference signals that are related to its physical sensing quantity, and to set consistency verification rules based on multi-source information fusion; Match independent sensors with calibration reference signals indirectly associated with calibrated sensors, and set indirect verification rules based on a preset multi-sensor fusion mechanism.

7. The method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot according to claim 1, characterized in that, The process of collecting real-time data from multiple sensors during the inspection of a thermal power plant by an inspection robot, and using a hierarchical dynamic benchmark database to calibrate and compare the real-time data from the multiple sensors, and then calculating the calibration parameters in real time based on the calibration and comparison results, includes the following steps: S31. Collect multi-source sensor data in real time during the robot's inspection task, and perform spatiotemporal alignment processing on the multi-source sensor data to obtain a spatiotemporally synchronized multimodal data stream. S32. Based on the spatiotemporally synchronized multimodal data stream and hierarchical dynamic benchmark database, and combined with the verification rules set for different sensor types, calculate the deviation and confidence level between the measured values ​​of each sensor and the benchmark signal respectively. S33. Based on the calculated deviation and confidence level, and combined with the correlation mapping relationship, the calibration parameters of each sensor are calculated in reverse using the least squares estimation method.

8. The method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot according to claim 7, characterized in that, The calculation of the deviation and confidence level between the measured values ​​of each sensor and the reference signal, based on the spatiotemporally synchronized multimodal data stream and hierarchical dynamic benchmark database, and combined with the verification rules set for different sensor types, includes the following steps: S321. Based on the established multi-source sensor classification system, identify the sensor type of the sensor to be calibrated. S322. Based on the identified sensor type and the corresponding verification rules, calculate the deviation between the multimodal data stream of the sensor's spatiotemporal synchronization and the reference signal. S323. Calculate the confidence level of the current comparison result based on the reliability of the data source, the type of verification rule, and the stability of historical calibration records used in the process of calculating the deviation.

9. A method for self-calibration and anomaly compensation of multi-source sensors in a thermal power plant inspection robot according to claim 8, characterized in that, The process of calculating the calibration parameters of each sensor using the least squares estimation method based on the calculated deviation and confidence level, combined with the correlation mapping relationship, includes the following steps: S331. For each type of sensor, construct an observation equation with calibration parameters as unknowns, based on the type of its associated mapping relationship. S332. Using the deviation as the observed value and the corresponding confidence level as the weighting factor, construct a weighted least squares objective function, and obtain the optimal estimate of the calibration parameters by solving for the minimum value of the objective function. S333. Evaluate the confidence interval of the optimal estimated value of the calibration parameter obtained by solving. When the variance of the parameter estimate is less than the preset variance threshold, output the calibration parameter. When the variance of the parameter estimate is greater than or equal to the preset variance threshold, use the incremental data acquisition mechanism to increase the number of observation samples and re-execute step S332 until the estimated value of the calibration parameter meets the preset confidence requirements.

10. A self-calibration and anomaly compensation system for multi-source sensors of a thermal power plant inspection robot, used to implement the self-calibration and anomaly compensation method for multi-source sensors of a thermal power plant inspection robot as described in any one of claims 1-9, characterized in that, The system includes: The reference signal selection module is used to select stable physical signals in the thermal power plant environment as calibration reference signals for the inspection tasks of the thermal power plant inspection robot. The database construction module is used to acquire the multi-source sensors carried by the thermal power plant inspection robot, classify the multi-source sensors according to the correlation characteristics between the sensed physical quantities and the calibration reference signals, establish the correlation mapping relationship between each type of sensor and the calibration reference signal based on the classification results, and construct a hierarchical dynamic reference database. The parameter calculation module is used to collect real-time data from multiple sensors during the inspection process of the thermal power plant inspection robot, and to use a hierarchical dynamic benchmark database to calibrate and compare the real-time data from multiple sensors. Based on the calibration and comparison results, the calibration parameters are calculated in real time. The data calibration module is used to compare calibration parameters with preset thresholds. When the calibration parameters are greater than the preset thresholds, the real-time data of the multi-source sensors is determined to be abnormal data and deleted. Otherwise, the calibration parameters are used to dynamically supplement the real-time data of the multi-source sensors to obtain calibrated sensor data. The data interpolation module is used to interpolate the deleted abnormal data using a preset interpolation method, and combine it with the calibrated sensor data to obtain the final multi-source sensor data of the thermal power plant inspection robot.

Images