Intelligent gas turbine leakage inspection method based on multi-sensor data fusion

Abstract

The invention belongs to the technical field of intelligence, and discloses an intelligent gas turbine leakage inspection method based on multi-sensor data fusion. Comprising the steps of deploying M sensors in each of T ranges, performing mutual verification of cross-sensor data by adopting a multi-level regression model, obtaining a verification result, and automatically triggering an automatic calibration mechanism of the sensors according to the verification result for calibration; using an SVR model to predict leakage risk data for each range; aiming at the leakage range, using an infrared thermal imaging camera to inspect, and obtaining an abnormal area; for the abnormal area, triggering a multispectral camera to inspect the abnormal area, obtaining waveband comprehensive data by using a dynamic fusion method, and judging whether the abnormal area leaks or not according to the waveband comprehensive data; if leakage occurs, early warning is triggered immediately; if not, triggering an infrared thermal imaging camera and a multispectral camera to carry out high-frequency inspection on the abnormal area; the accuracy and response speed of gas turbine leakage inspection can be improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent technologies, and specifically to an intelligent inspection method for gas turbine leakage with multi-sensor data fusion. Background Art

[0002] In existing intelligent inspection systems for gas turbine leakage with multi-sensor data fusion, there are many defects:

[0003] Firstly, in the operating environment of the gas turbine, there are harsh environments such as high temperature, vibration, electromagnetic interference or dust, which lead to a decline in sensor performance or abnormal data. The existing system lacks dynamic automatic calibration of sensors, resulting in false alarms or missed alarms;

[0004] Secondly, the existing gas turbine leakage inspection uses sensors for inspection, fuses the data of the sensors to judge whether there is an abnormality, and only conducts global inspection, that is, collects data in various ranges of the gas turbine to judge whether there is leakage or abnormality; However, this method is prone to ignoring local abnormalities, especially small-scale abnormalities or tiny leaks that are difficult to detect through global data; Due to relying only on global data, the existing method often fails to achieve a detailed inspection of the abnormal area;

[0005] In addition, when the existing method detects an abnormality, it cannot effectively distinguish the abnormality caused by gas turbine leakage from the abnormality caused by equipment failure or other environmental factors, and directly defaults to the abnormality caused by leakage. This rough judgment will lead to incorrect judgments and cannot provide sufficiently accurate leakage information.

[0006] In view of this, the present invention proposes an intelligent inspection method for gas turbine leakage with multi-sensor data fusion to solve the above problems. Summary of the Invention

[0007] In order to overcome the above defects of the prior art and to achieve the above object, the present invention provides the following technical solution, an intelligent inspection method for gas turbine leakage with multi-sensor data fusion, including:

[0008] Step SS1: Deploy M sensors in each of the T ranges, and based on the data collected from the M sensors, use a multi-level regression model to perform mutual verification of cross-sensor data and obtain a verification result. If the data of a sensor deviates from the verification result, the automatic calibration mechanism of the sensor is automatically triggered for calibration;

[0009] Step SS2: In each range, use the calibrated data collected by the automatically calibrated sensors, use the SVR model to predict the leakage risk data of each range, and set a leakage threshold. If the leakage risk data is greater than or equal to the leakage threshold, there is a leakage risk in this range; if the leakage risk data is less than the leakage threshold, this range is safe;

[0010] Step SS3: For the range with leakage risk, use an infrared thermal imaging camera for inspection and use the dynamic error connection marking method to obtain the abnormal area;

[0011] Step SS4: For the abnormal area, trigger the multi-spectral camera to inspect the abnormal area, use the dynamic fusion method to obtain the band comprehensive data, and set the leakage determination threshold. If the band comprehensive data of the abnormal area is greater than the leakage determination threshold, it means that the abnormality in this area is caused by the leakage of the gas turbine; if the band comprehensive data of the abnormal area is less than the leakage determination threshold, it means that the abnormality in this area is caused by the equipment itself;

[0012] Step SS5: When the area abnormality is caused by the leakage of the gas turbine, immediately trigger an alarm; when the area abnormality is caused by the equipment itself, trigger the infrared thermal imaging camera and the multi-spectral camera to perform high-frequency inspections on the abnormal area.

[0013] Further, the specific method of deploying M sensors in each of the T ranges includes:

[0014] Install fiber-optic methane sensors, temperature sensors, and pressure sensors in the ranges of the gas turbine valves, fuel pipe expansion joints, gas turbine fuel ring pipes, combustion chambers, bottoms of combustion chambers, tops of combustion chambers, and air outlet ranges inside the gas turbine casing.

[0015] Further, based on the data collected from the M sensors, a multi-level regression model is used to mutually verify the cross-sensor data and obtain the verification result. If the data of the sensor deviates from the verification result, the specific method of automatically triggering the automatic calibration mechanism of the sensor for calibration includes:

[0016] Based on the data collected from the M sensors, use a multi-level regression model to model the methane concentration data, temperature data, and pressure data. Input the temperature data and pressure data into the multi-level regression model to obtain the verified methane concentration data; input the methane concentration data and pressure data into the multi-level regression model to obtain the verified temperature data, and input the methane concentration data and temperature data into the multi-level regression model to obtain the verified pressure data;

[0017] Compare the real-time methane concentration data, temperature data, and pressure data with the verified methane concentration data, temperature data, and pressure data of the multi-level regression model. If the data measured by the sensor deviates from the verification result, and the verification result is the verified methane concentration data, temperature data, and pressure data, it means that the sensor has a deviation, and automatically trigger the automatic calibration mechanism of the sensor for calibration.

[0018] Further, the specific method of using the calibration data collected by the automatically calibrated sensor in each range and using the SVR model to predict the leakage risk data of each range includes:

[0019] Establish T databases for T ranges, store the data collected by the automatically calibrated sensor in each range in the database. The database includes M groups of samples, and each group of samples consists of methane concentration data, temperature data, pressure data, and leakage risk data. Input the M groups of samples as input data into the SVR model to obtain the predicted leakage risk data for each range.

[0020] Further, the specific method of using the infrared thermal imaging camera to conduct patrol inspections on the ranges with leakage risks and using the dynamic error connected component labeling method to obtain the abnormal areas includes:

[0021] Use the infrared thermal imaging camera to conduct real-time patrol inspections on the ranges with leakage risks to obtain thermal imaging images, where the patrol inspection frequency of the infrared thermal imaging camera is u;

[0022] Use the median filtering method to denoise the thermal imaging map to obtain the denoised thermal imaging map;

[0023] Use the sliding window method to calculate the average value of each pixel point in the thermal imaging map within the historical n time periods to obtain the average thermal imaging map;

[0024] Calculate the error of each pixel point between the average thermal imaging map and the denoised thermal imaging map to obtain the error thermal imaging map;

[0025] For the error thermal imaging map, calculate the mean and standard deviation of each pixel point in the error thermal imaging map, and set the dynamic error threshold based on the mean and standard deviation;

[0026] In the error thermal imaging map, mark the pixel points with errors greater than the dynamic error threshold in blue, assign a unique label 0 to each blue pixel point in the error thermal imaging map, use the traversal method to detect whether the labels of the upper, lower, left, right, and four diagonal directions of the blue pixel points with label 0 are the same, and adopt the connected component labeling method to identify the blue areas formed in the error thermal imaging map. The blue areas are the abnormal areas.

[0027] Further, for the abnormal areas, trigger the multi-spectral camera to conduct patrol inspections on the abnormal areas, and the specific method of using the dynamic fusion method to obtain the band comprehensive data includes:

[0028] When the abnormal area is determined, trigger the multi-spectral camera. The patrol inspection frequency of the multi-spectral camera is f, and use the multi-spectral camera to collect N bands of each pixel point in the abnormal area;

[0029] Each band includes radiation intensity, absorption bandwidth, spectral reflectance, and normalized absorption index data;

[0030] For the N bands of each pixel, D bands are obtained using a dynamic adaptive method;

[0031] For the D bands, the weighted average method is used to obtain the integrated band data of each pixel, and the integrated band data of each pixel is weighted and fused to obtain the integrated band data of the abnormal area.

[0032] Further, the specific method of obtaining D bands using the dynamic adaptive method for the N bands of each pixel includes:

[0033] When the multispectral camera is triggered, the multispectral camera acquires the N-band data of each pixel in the abnormal area, and uses the spectral angle matching analysis correction method to obtain the response intensity data of the N bands of each pixel to the abnormal area;

[0034] Sort the response intensities of the N bands to the abnormal area in descending order, set the response intensity threshold using the expert experience method, retain the bands greater than or equal to the response threshold, and remove the bands less than the response threshold, so that each pixel retains D bands.

[0035] Further, the specific method of obtaining the response intensity data of the N bands of each pixel to the abnormal area using the spectral angle matching analysis correction method includes:

[0036] Construct N spectral vectors from the radiation intensity, absorption bandwidth, spectral reflectance, and normalized absorption index data of the N bands of the pixel;

[0037] Calculate the spectral angle between the spectral vector of each band and the spectral feature vector of the abnormal area. The formula is: where S λb represents the spectral vector of the b-th band, S a represents the spectral data vector of the abnormal area, and θ b represents the spectral angle between the spectral vector of the b-th band and the spectral data vector of the abnormal area. The smaller the spectral angle, the more the band conforms to the abnormal area;

[0038] For each band, use the Pearson correlation method to calculate the linear relationship between the spectral vector of each band and the spectral feature vector of the abnormal area to obtain the correlation value;

[0039] Use the correlation value to correct the spectral angle between the spectral vector of each band and the spectral data vector of the abnormal area to obtain the correlation intensity data of each band; through the correlation intensity data, calculate the response intensity data. The formula is: where rb denotes the correlation intensity data of the b-th band, max(r) denotes the maximum correlation intensity data among N bands, and R b denotes the response intensity data of the b-th band.

[0040] Furthermore, the method for obtaining the spectral feature vector of the abnormal area includes:

[0041] It is known that the abnormal area includes H pixel points. For each pixel point, spectral data of N bands are extracted, and the spectral data are constructed into a spectral vector. Then, the weighted average method is used to perform weighted calculation on the spectral vectors of H pixel points, and thus the spectral feature vector of the abnormal area is obtained.

[0042] Furthermore, when the area anomaly is caused by the device itself, the specific method for triggering the infrared thermal imaging camera and the multispectral camera to perform high-frequency inspection on the abnormal area includes:

[0043] For the abnormal area caused by the device itself, high-frequency inspection is triggered. Among them, the inspection frequency of the infrared thermal imaging camera is set to 2u, and the inspection frequency of the multispectral camera is set to 2f.

[0044] The technical effects and advantages of a multi-sensor data fusion intelligent inspection method for gas turbine leakage of the present invention:

[0045] Through the multi-sensor fusion technology, combining methane concentration data, temperature data, and pressure data, the present invention can accurately identify the gas turbine leakage risk and equipment failures, predict in advance and trigger early warnings; through the multi-level regression model, mutual verification of cross-sensor data is carried out, and according to the verification results, the automatic calibration mechanism of the sensor is automatically triggered for calibration, ensuring the high accuracy and reliability of the data, and effectively reducing the possibility of missed reports and false alarms;

[0046] According to different ranges, the SVR model is combined to predict the leakage risk data of the range, and targeted inspection measures are intelligently triggered to perform global inspection, local inspection, and fine inspection, improving the accuracy and efficiency of the inspection, and ensuring that the ranges with leakage risks can be discovered in time;

[0047] For the ranges with leakage risks, the infrared thermal imaging camera is used to inspect the ranges with leakage risks, and the areas with abnormal temperatures can be accurately identified, helping to quickly discover potential leakage sources; the infrared thermal imaging camera provides a non-contact detection method, avoiding the risks of manual operation. Especially when inspecting in high-temperature, high-pressure or dangerous areas, it can ensure the safety of the inspection personnel;

[0048] Collecting the band data of the abnormal area through a multispectral camera can obtain more detailed physical characteristic information, such as radiation intensity, absorption bandwidth, spectral reflectivity, etc. These multi-dimensional data can more accurately distinguish the abnormalities caused by gas turbine leakage from those caused by equipment failures or other factors, providing a more scientific and detailed means of leakage detection;

[0049] As a non-contact monitoring tool, the multispectral camera can accurately monitor the abnormal area without disturbing the normal operation of the equipment. This method can effectively avoid human interference and potential safety risks, especially when dealing with equipment in high-temperature, high-pressure or dangerous environments, which can greatly enhance safety;

[0050] Through precise inspection of the abnormal area, potential problems caused by the equipment itself can be identified in a timely manner. This can not only reduce the equipment downtime caused by equipment failures, but also help technicians implement more precise maintenance or replacement strategies, further improving the overall operational reliability and stability of the equipment;

[0051] By distinguishing the abnormalities caused by leakage from those caused by equipment itself failures and taking corresponding early warning and high-frequency inspection measures, the problem source can be more accurately located, reducing false alarms and missed alarms, ensuring that effective measures are taken in a timely manner for intervention, thereby improving the response speed and accuracy of the system, optimizing the operation and maintenance of the equipment, and enhancing the safety guarantee level and operation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 Schematic diagram of an intelligent inspection method for gas turbine leakage with multi-sensor data fusion according to the present invention;

[0053] Figure 2 Schematic diagram of the fine inspection method according to the present invention;

[0054] Figure 3 Schematic diagram of an intelligent inspection system for gas turbine leakage with multi-sensor data fusion according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] Embodiment 1

[0057] Please refer to Figure 1 and Figure 2As shown in the figure, an intelligent inspection method for gas turbine leakage with multi-sensor data fusion in this embodiment includes:

[0058] Step SS1: Deploy M sensors in each of the T ranges. Based on the data collected from the M sensors, a multi-level regression model is used to mutually verify the cross-sensor data and obtain the verification result. If the data of a sensor deviates from the verification result, the automatic calibration mechanism of the sensor is automatically triggered for calibration.

[0059] Step SS2: In each range, using the calibrated data collected by the automatically calibrated sensors, an SVR model is used to predict the leakage risk data of each range, and a leakage threshold is set. If the leakage risk data is greater than or equal to the leakage threshold, there is a leakage risk in this range; if the leakage risk data is less than the leakage threshold, this range is safe.

[0060] Step SS3: For the ranges with leakage risks, an infrared thermal imaging camera is used for inspection, and the dynamic error connected component labeling method is used to obtain the abnormal areas.

[0061] Step SS4: For the abnormal areas, a multi-spectral camera is triggered to inspect the abnormal areas, and a dynamic fusion method is used to obtain the band comprehensive data, and a leakage determination threshold is set. If the band comprehensive data of the abnormal area is greater than the leakage determination threshold, it indicates that the abnormality in this area is caused by gas turbine leakage; if the band comprehensive data of the abnormal area is less than the leakage determination threshold, it indicates that the abnormality in this area is caused by the equipment itself.

[0062] Step SS5: When the area abnormality is caused by gas turbine leakage, an alarm is immediately triggered; when the area abnormality is caused by the equipment itself, the infrared thermal imaging camera and the multi-spectral camera are triggered to perform high-frequency inspections on the abnormal areas.

[0063] The specific method of deploying M sensors in each of the T ranges includes:

[0064] Install fiber optic methane sensors, temperature sensors, and pressure sensors in the gas turbine valve range, fuel pipe expansion joint range, gas turbine fuel ring pipe range, combustion chamber range, bottom of the combustion chamber range, top of the combustion chamber range, and the air outlet range inside the gas turbine housing.

[0065] Existing sensors will be calibrated regularly at set time intervals. For example, every certain number of days, months, or hours of use, the sensor will start the calibration equipment for standardization adjustment. The advantage of this method is simple operation, but the disadvantage is that the calibration time point may not be consistent with the actual moment when the performance deviation of the sensor occurs. Therefore, the real-time changes in the performance deviation of the sensor may be missed.

[0066] Secondly, calibration is performed based on a standard reference object. For a gas sensor (such as an optical fiber methane sensor), a common practice is to calibrate it using a standard gas with a known concentration, compare the standard gas data with the data collected by the sensor, and adjust the sensor output to be consistent with the standard gas concentration. Although this method is relatively accurate, it requires manual operation and has high requirements for the experimental environment and gas concentration.

[0067] Based on the data collected from M sensors, a multi-level regression model is used to cross-validate the cross-sensor data and obtain a validation result. If the data of a sensor deviates from the validation result, the specific methods for automatically triggering the automatic calibration mechanism of this sensor for calibration include:

[0068] Based on the data collected from M sensors, a multi-level regression model is used to model the methane concentration data, temperature data, and pressure data. The temperature data and pressure data are input into the multi-level regression model to obtain the validated methane concentration data; the methane concentration data and pressure data are input into the multi-level regression model to obtain the validated temperature data, and the methane concentration data and temperature data are input into the multi-level regression model to obtain the validated pressure data.

[0069] Compare the real-time methane concentration data, temperature data, and pressure data with the validated methane concentration data, temperature data, and pressure data obtained by the multi-level regression model. If the data measured by the sensor deviates from the validation result, and the validation result is the validated methane concentration data, temperature data, and pressure data, it indicates that there is a deviation in this sensor, and the automatic calibration mechanism of this sensor is automatically triggered for calibration.

[0070] Among them, the calculation formula of the multi-level regression model is: Among them, T pre and P pre represent the validated methane concentration data, temperature data, and pressure data, and represent the intercepts of the methane concentration data, temperature data, and pressure data, X T 、 and X P represent the actual temperature data value, methane gas concentration data value, and pressure data value measured by the sensor, and represent the regression coefficients of the methane concentration data, temperature data, and pressure data. The intercept and regression coefficients are obtained by the least squares method.

[0071] The present invention collects data from multiple sensors in real time (such as methane concentration data, temperature data, and pressure data), uses a multi-level regression model to mutually verify the sensor data, can timely detect the performance deviation of the sensors, and automatically trigger the calibration mechanism when the deviation occurs without manual intervention. This method eliminates the limitations of the existing regular calibration, and avoids the risks of missing the calibration opportunity or human operation errors.

[0072] Secondly, the adopted multi-level regression model realizes the cross-sensor data mutual verification by analyzing the relationships between different sensors. This not only improves the accuracy of the data, but also can minimize the errors caused by the failure or deviation of a single sensor. This model-based calibration method is more scientific, systematic, and has a high self-adaptability than the existing methods that rely on manual experience or regular calibration, and can operate stably in complex environments.

[0073] In each range, the specific ways of using the calibration data collected by the automatically calibrated sensors and using the SVR model to predict the leakage risk data of each range include:

[0074] Establish T databases for T ranges, store the data collected by the automatically calibrated sensors in each range in the databases. Each database includes M groups of samples, and each group of samples consists of methane concentration data, temperature data, pressure data, and leakage risk data. Input the M groups of samples as input data into the SVR model to obtain the predicted leakage risk data of each range.

[0075] Existing methods usually do not have a mechanism for real-time prediction of the leakage risk in each range, and usually directly judge whether a leakage occurs.

[0076] By establishing databases for multiple ranges and using the SVR model to model the methane concentration data, temperature data, and pressure data in each range, the leakage risk can be accurately predicted in each range.

[0077] Existing methods of multi-sensor fusion for leakage judgment usually rely on integrating the data collected by multiple sensors to identify whether a leakage occurs. However, this method has significant defects in judging minor leaks. First of all, although multi-sensor fusion can improve the accuracy of judgment by integrating the data of different sensors, when dealing with minor leaks, these methods often cannot sensitively capture the minor changes. Minor leaks may only cause extremely subtle fluctuations in some sensors, while the existing data fusion technologies are often designed to identify larger-scale leaks, which makes the signals of minor leaks easily submerged in the background noise, resulting in false negatives.

[0078] Secondly, differences in the response characteristics, accuracy, and sensitivity of sensors may lead to distortion of the fusion results. Especially when the leakage signal is weak, the deviation or failure of a single sensor can have a significant impact on the overall judgment, thus affecting the accurate detection of minor leaks.

[0079] In summary, although multi-sensor fusion is effective in detecting large leaks, it has defects such as insufficient sensitivity, poor adaptability, and susceptibility to sensor errors in judging minor leaks, which limits its application effect in fine monitoring and early leak detection.

[0080] If there is a leak at a certain point of the gas turbine, the leaked gas causes temperature changes in the surrounding area. Leakage points are likely to occur especially at pipe connections, valves, and seals. An infrared thermal imaging camera can detect these local temperature changes.

[0081] Install infrared thermal imaging cameras around the gas turbine to ensure that key areas of the gas turbine equipment (such as gas pipelines, seal joints, valves, etc.), especially areas prone to leakage, are covered for a non-blind-spot inspection of the gas turbine.

[0082] For the area with a leakage risk, the specific methods of using an infrared thermal imaging camera for inspection and obtaining the abnormal area using the dynamic error connected component labeling method include:

[0083] Use an infrared thermal imaging camera to conduct real-time inspection of the area with a leakage risk to obtain a thermal imaging image, where the inspection frequency of the infrared thermal imaging camera is u.

[0084] Use the median filtering method to denoise the thermal imaging map to obtain a denoised thermal imaging map.

[0085] Use the sliding window method to calculate the average value of each pixel point in the thermal imaging map over the past n time periods to obtain an average thermal imaging map.

[0086] Calculate the error of each pixel point between the average thermal imaging map and the denoised thermal imaging map to obtain an error thermal imaging map.

[0087] For the error thermal imaging map, calculate the mean and standard deviation of each pixel point in the error thermal imaging map, and set the dynamic error threshold based on the mean and standard deviation.

[0088] In the error thermal imaging map, identify the pixel points with errors greater than the dynamic error threshold in blue, assign a unique label 0 to each blue pixel point in the error thermal imaging map, use the traversal method to detect whether the labels of the upper, lower, left, right, and four diagonal directions of the blue pixel points with label 0 are the same, and adopt the connected component labeling method to identify the blue areas formed in the error thermal imaging map. The blue areas are the abnormal areas.

[0089] By using an infrared thermal imaging camera, temperature anomaly regions can be quickly identified; this method can accurately locate areas where leaks may exist, especially suitable for detecting temperature changes caused by leaks of methane, gas, etc.; the infrared thermal imaging camera has a high spatial resolution, can quickly scan a large area, and find regions with temperature anomalies or not meeting normal environmental conditions; this provides a clear target area for subsequent detailed inspections, thus improving the inspection efficiency and accuracy.

[0090] Furthermore, for the abnormal regions, a multispectral camera is triggered to conduct inspections on the abnormal regions. The specific ways to obtain band comprehensive data using the dynamic fusion method include:

[0091] When the abnormal region is determined, a multispectral camera is triggered. The inspection frequency of the multispectral camera is f, and N bands of each pixel point in the abnormal region are collected using the multispectral camera;

[0092] Each band includes radiation intensity, absorption bandwidth, spectral reflectivity, and normalized absorption index data;

[0093] For the N bands of each pixel point, D bands are obtained using the dynamic adaptive method;

[0094] For the D bands, the weighted average method is used to obtain the band comprehensive data of each pixel point, and the band comprehensive data of each pixel point are weighted and fused to obtain the band comprehensive data of the abnormal region.

[0095] Existing methods usually directly use all bands or select several common fixed bands for analysis. This method has several significant defects:

[0096] First of all, using all bands will lead to a huge amount of data, increasing the burden of processing and storage. At the same time, it will also introduce a large amount of background information that is irrelevant to target detection and may even cause interference, reducing the accuracy and efficiency of anomaly detection;

[0097] Secondly, adopting several common fixed bands can simplify the processing process and reduce the demand for computing resources. However, this method ignores the spectral characteristic differences under different scenarios and different equipment states, and cannot flexibly cope with complex actual situations, resulting in inaccurate detection results or insufficient sensitivity;

[0098] In addition, the selection of fixed bands often based on general experience and historical data is difficult to adapt to special situations or new changes in a specific environment, and it is easy to miss important abnormal signals; Generally speaking, these existing methods lack pertinence and flexibility, cannot effectively distinguish truly useful bands, thus limiting their performance and reliability in application scenarios such as intelligent inspection of leaks in complex and dynamically changing gas turbines;

[0099] For the N bands of each pixel, the specific method of using the dynamic adaptive method to obtain D bands includes:

[0100] When triggering the multispectral camera, the multispectral camera acquires the N-band data of each pixel in the abnormal area, and uses the spectral angle matching analysis correction method to obtain the response intensity data of the N bands of each pixel to the abnormal area;

[0101] Sort the response intensities of the N bands to the abnormal area in descending order, set the response intensity threshold using the expert experience method, retain the bands greater than or equal to the response threshold, and remove the bands less than the response threshold. Then, each pixel retains D bands, where the D bands retained by each pixel are not the same. For example, if the y-th pixel retains 3 bands, then D = 3; if the j-th pixel retains 6 bands, then D = 6;

[0102] This dynamic adaptive method acquires the N-band data of each pixel in the abnormal area through a multispectral camera, calculates the response intensities of the N bands of each pixel to the abnormal area using the spectral angle matching analysis correction method, then sorts these response intensities in descending order and selects the D bands with the strongest response to the abnormal area according to the threshold set by expert experience, realizing that each pixel retains a different number of bands;

[0103] The advantage of this method is that it can specifically improve the detection accuracy because it dynamically selects the most relevant bands based on the actual response intensity, thus effectively reducing noise interference; at the same time, it has high flexibility and can adapt to complex and variable scenarios in gas turbine leakage detection without presetting fixed bands, reducing the need for manual intervention; in addition, this method can significantly reduce the amount of redundant data, optimize the data load and calculation efficiency of subsequent processing, and make the entire inspection process more efficient.

[0104] The specific method of using the spectral angle matching analysis correction method to obtain the response intensity data of the N bands of each pixel to the abnormal area includes:

[0105] Construct N spectral vectors from the radiation intensity, absorption bandwidth, spectral reflectivity, and normalized absorption index data of the N bands of the pixel;

[0106] Calculate the spectral angle between the spectral vector of each band and the spectral feature vector of the abnormal area. The formula is: where, S λb represents the spectral vector of the b-th band, S a represents the spectral data vector of the abnormal area, θ b represents the spectral angle between the spectral vector of the b-th band and the spectral data vector of the abnormal area. The smaller the spectral angle, the more the band conforms to the abnormal area;

[0107] For each band, the Pearson correlation method is used to calculate the linear relationship between the spectral vector of each band and the spectral feature vector of the abnormal region, and the correlation value is obtained;

[0108] The spectral angle between the spectral vector of each band and the spectral data vector of the abnormal region is corrected using the correlation value to obtain the correlation intensity data for each band; based on the correlation intensity data, the response intensity data is calculated using the formula: where r b represents the correlation intensity data for the b-th band, max(r) represents the maximum correlation intensity data among the N bands, and R b represents the response intensity data for the b-th band;

[0109] Regarding the spectral angle matching analysis and correction method: First, by constructing a spectral vector that includes radiation intensity, absorption bandwidth, spectral reflectance, and normalized absorption index data, the unique spectral characteristics of each band can be comprehensively captured, providing a rich information basis for subsequent analysis; then, spectral angle matching analysis is adopted, which can not only directly quantify the similarity between each band and the spectral characteristics of the abnormal region (the smaller the spectral angle, the higher the matching degree), thereby improving the accuracy and sensitivity of detection; in addition, combining the Pearson correlation method for linear relationship analysis and correcting the spectral angle can more precisely adjust the importance weights of each band, ensuring accurate identification of abnormal signals even in complex and variable environments; finally, the process of calculating the response intensity data based on the correlation intensity data can dynamically adjust the contribution degrees of each band according to actual observation situations, avoiding the biases and limitations that may be brought by traditional fixed-parameter methods, enhancing the robustness and reliability of the overall system, and achieving efficient and accurate monitoring of abnormal phenomena such as gas turbine leakage. This spectral angle matching analysis and correction method not only improves the accuracy of abnormal detection and can identify tiny leaks, but also optimizes the resource utilization efficiency, providing strong technical support for intelligent patrol inspection.

[0110] Furthermore, the method for obtaining the spectral feature vector of the abnormal region includes:

[0111] It is known that the abnormal region includes H pixel points. For each pixel point, the spectral data of N bands are extracted, and the spectral data are constructed into a spectral vector, and the spectral vectors of the H pixel points are weighted and calculated using the weighted average method, then the spectral feature vector of the abnormal region is obtained.

[0112] The advantages of this multi-spectral camera patrol inspection are as follows. By obtaining data in multiple bands through a multi-spectral camera and using a dynamic adaptive method to process this data, the accuracy and sensitivity of leak detection can be significantly improved. Different from the existing single-band detection method, the comprehensive data of multiple bands can provide more comprehensive information about the abnormal area, helping to more accurately identify tiny leaks. By fusing data of different bands through the weighted average method, environmental interference and data noise can be effectively reduced, and the response ability to weak leak signals is improved.

[0113] In addition, the dynamic adaptive method can automatically adjust the band selection according to the actual situation, making the detection process more flexible and accurate. This method can flexibly remove irrelevant or interfering bands according to the response intensity data of each pixel point, thereby improving the robustness and accuracy of detection. Especially when facing complex environments and anomalies caused by the equipment itself, it can accurately distinguish the leak source from equipment failures.

[0114] This meticulous patrol inspection mechanism makes the detection of tiny leaks more reliable, significantly reducing the probability of missed reports and false alarms, and improving the efficiency and security of the entire monitoring system.

[0115] Furthermore, the specific method of triggering the infrared thermal imaging camera and the multi-spectral camera to perform high-frequency patrol inspection on the abnormal area when the abnormal area is caused by the equipment itself includes:

[0116] For the abnormal area caused by the equipment itself, high-frequency patrol inspection is triggered. Among them, the patrol inspection frequency of the infrared thermal imaging camera is set to 2u, and the patrol inspection frequency of the multi-spectral camera is set to 2f.

[0117] The main purpose of using high-frequency patrol inspection is to increase the monitoring frequency of the abnormal area of the equipment, so as to more timely capture potential problems or changes caused by the equipment itself. In the case of equipment anomalies, by increasing the patrol inspection frequency (such as doubling the patrol inspection frequencies of the infrared thermal imaging camera and the multi-spectral camera), the equipment status can be monitored more accurately, avoiding missing subtle changes. High-frequency patrol inspection can provide more detection data in a short time, helping to discover equipment failures or anomalies faster, reducing the risk of equipment damage. In addition, the rapid response can improve the patrol inspection efficiency and take timely measures in the early stage of equipment failure, avoiding possible greater damage or safety hazards, and ensuring the normal operation of the equipment and the stability of the system.

[0118] In this embodiment, through the multi-sensor fusion technology, combined with methane concentration data, temperature data, and pressure data, it is possible to accurately identify the leakage risks and equipment failures of gas turbines, predict in advance and trigger early warnings; through a multi-level regression model for cross-sensor data mutual verification, according to the verification results, the automatic calibration mechanism of the sensor is automatically triggered for calibration, ensuring high data accuracy and reliability, and effectively reducing the possibility of false negatives and false positives;

[0119] According to different ranges, the SVR model is combined to predict the leakage risk data of the range, and targeted inspection measures are intelligently triggered for global inspection, improving the accuracy and efficiency of the inspection, and ensuring that the ranges with leakage risks can be discovered in time;

[0120] For the ranges with leakage risks, the ranges with leakage risks are inspected by an infrared thermal imaging camera, which can accurately identify the areas with abnormal temperatures and help quickly discover potential leakage sources; the infrared thermal imaging camera provides a non-contact detection method, avoiding the risks of manual operation. Especially when inspecting in high-temperature, high-pressure or dangerous areas, it can ensure the safety of the inspection personnel;

[0121] By collecting the band data of the abnormal area through a multi-spectral camera, more detailed physical characteristic information can be obtained, such as radiation intensity, absorption bandwidth, spectral reflectivity, etc. These multi-dimensional data can more accurately distinguish the abnormalities caused by gas turbine leakage from those caused by equipment failures or other factors, providing a more scientific and detailed leakage detection method;

[0122] As a non-contact monitoring tool, the multi-spectral camera can accurately monitor the abnormal area without disturbing the normal operation of the equipment. This method can effectively avoid human interference and potential safety risks, especially when dealing with equipment in high-temperature, high-pressure or dangerous environments, greatly enhancing safety;

[0123] Through the precise inspection of the abnormal area, potential problems caused by the equipment itself can be identified in time. This can not only reduce the equipment downtime caused by equipment failures, but also help technicians implement more precise maintenance or replacement strategies, further improving the overall operation reliability and stability of the equipment;

[0124] By distinguishing the abnormalities caused by leakage from those caused by equipment itself failures and taking corresponding early warning and high-frequency inspection measures, the problem source can be more accurately located, reducing the situations of false positives and false negatives, ensuring that effective measures are taken in time for intervention, thereby improving the response speed and accuracy of the system, optimizing the operation and maintenance of the equipment, and improving the safety guarantee level and operation efficiency.

[0125] Embodiment 2

[0126] Please refer to Figure 3 As shown, for the parts not described in detail in this embodiment, refer to the description in Embodiment 1. A multi-sensor data fusion intelligent inspection system for gas turbine leakage is provided, including:

[0127] Sensor dynamic calibration unit: Deploy M sensors in each of the T ranges. Based on the data collected from the M sensors, a multi-level regression model is used to cross-validate the data between sensors and obtain the validation result. If the data of a sensor deviates from the validation result, the automatic calibration mechanism of the sensor is automatically triggered for calibration;

[0128] Global inspection unit: In each range, using the calibrated data collected by the automatically calibrated sensors, the SVR model is used to predict the leakage risk data of each range, and a leakage threshold is set. If the leakage risk data is greater than or equal to the leakage threshold, there is a leakage risk in this range; if the leakage risk data is less than the leakage threshold, this range is safe;

[0129] Local inspection unit: For the ranges with leakage risks, an infrared thermal imaging camera is used for inspection, and the dynamic error connected component labeling method is used to obtain the abnormal area;

[0130] Fine inspection unit: For the abnormal area, a multi-spectral camera is triggered to inspect the abnormal area, and the dynamic fusion method is used to obtain the band comprehensive data, and a leakage determination threshold is set. If the band comprehensive data of the abnormal area is greater than the leakage determination threshold, it means that the abnormality in this area is caused by gas turbine leakage; if the band comprehensive data of the abnormal area is less than the leakage determination threshold, it means that the abnormality in this area is caused by the equipment itself;

[0131] Early warning unit: When the area abnormality is caused by gas turbine leakage, an early warning is immediately triggered; when the area abnormality is caused by the equipment itself, the infrared thermal imaging camera and the multi-spectral camera are triggered to perform high-frequency inspections on the abnormal area.

[0132] Embodiment 3

[0133] This embodiment publicly provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the operation mode of the above-provided multi-sensor data fusion intelligent inspection method for gas turbine leakage.

[0134] Since the electronic device introduced in this embodiment is the electronic device adopted in a multi-sensor data fusion-based intelligent inspection method for gas turbine leakage in the embodiments of the present application, based on the multi-sensor data fusion-based intelligent inspection method for gas turbine leakage introduced in the embodiments of the present application, those skilled in the art can understand the specific implementation manners and various variations of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device implements the method in the embodiments of the present application will not be introduced in detail here. As long as those skilled in the art implement the electronic device adopted in the multi-sensor data fusion-based intelligent inspection method for gas turbine leakage in the embodiments of the present application, it falls within the scope of protection of the present application.

[0135] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0136] The above is only the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for ordinary technical users in the technical field, several improvements and refinements made without departing from the principle of the present invention should also be regarded as within the protection scope of the present invention.

Claims

1. An intelligent inspection method for gas turbine leakage based on multi-sensor data fusion, characterized in that Including: Step SS1: Deploy M sensors in each of the T ranges. Based on the data collected from the M sensors, use a multi-level regression model to perform cross-sensor data mutual verification and obtain a verification result. If the data of a sensor deviates from the verification result, automatically trigger the automatic calibration mechanism of this sensor for calibration; Step SS2: In each range, use the calibrated data collected by the automatically calibrated sensors, and use the SVR model to predict the leakage risk data of each range, and set a leakage threshold. If the leakage risk data is greater than or equal to the leakage threshold, there is a leakage risk in this range; if the leakage risk data is less than the leakage threshold, this range is safe; Step SS3: For the ranges with leakage risks, use an infrared thermal imaging camera for inspection and use the dynamic error connectivity marking method to obtain the abnormal area; Step SS4: For the abnormal area, trigger a multi-spectral camera to inspect the abnormal area, use the dynamic fusion method to obtain the band comprehensive data, and set a leakage determination threshold. If the band comprehensive data of the abnormal area is greater than the leakage determination threshold, it means that the abnormality in this area is caused by the leakage of the gas turbine; if the band comprehensive data of the abnormal area is less than the leakage determination threshold, it means that the abnormality in this area is caused by the equipment itself; Step SS5: When the area abnormality is caused by the leakage of the gas turbine, immediately trigger an alarm; When the area abnormality is caused by the equipment itself, trigger the infrared thermal imaging camera and the multi-spectral camera to perform high-frequency inspections on the abnormal area.

2. The intelligent inspection method for gas turbine leakage with multi-sensor data fusion according to claim 1, characterized in that The specific method of deploying M sensors in each of the T ranges includes: Install fiber-optic methane sensors, temperature sensors, and pressure sensors in the gas turbine valve range, fuel pipe expansion joint range, gas turbine fuel ring pipe range, combustion chamber range, bottom of the combustion chamber range, top of the combustion chamber range, and the air outlet range inside the gas turbine housing.

3. The intelligent inspection method for leakage of gas turbine with multi-sensor data fusion according to claim 2, characterized in that, The specific method of using a multi-level regression model to perform cross-sensor data mutual verification based on the data collected from the M sensors and obtaining a verification result, and automatically triggering the automatic calibration mechanism of this sensor for calibration if the data of the sensor deviates from the verification result includes: Based on the data collected from the M sensors, use a multi-level regression model to model the methane concentration data, temperature data, and pressure data. Input the temperature data and pressure data into the multi-level regression model to obtain the verified methane concentration data; input the methane concentration data and pressure data into the multi-level regression model to obtain the verified temperature data, and input the methane concentration data and temperature data into the multi-level regression model to obtain the verified pressure data; Compare the real-time methane concentration data, temperature data, and pressure data with the verified methane concentration data, temperature data, and pressure data of the multi-level regression model. If the data measured by the sensor deviates from the verification result, and the verification result is the verified methane concentration data, temperature data, and pressure data, it means that this sensor has a deviation, and automatically trigger the automatic calibration mechanism of this sensor for calibration.

4. The intelligent inspection method for gas turbine leakage with multi-sensor data fusion according to claim 3, characterized in that The specific method of using the SVR model to predict the leakage risk data of each range by using the calibrated data collected by the automatically calibrated sensors in each range includes: Establish T databases for T ranges, store the data collected by the automatically calibrated sensors in each range in the databases. The databases include M groups of samples, and each group of samples consists of methane concentration data, temperature data, pressure data, and leakage risk data. Use the M groups of samples as input data and input them into the SVR model to obtain the predicted leakage risk data for each range.

5. A method for intelligent inspection of gas turbine leakage with multi-sensor data fusion according to claim 4, characterized in that, For the range with leakage risk, the specific method of using an infrared thermal imaging camera for inspection and using the dynamic error connected component labeling method to obtain the abnormal area includes: Use the infrared thermal imaging camera to conduct real-time inspection on the range with leakage risk to obtain a thermal imaging image, where the inspection frequency of the infrared thermal imaging camera is u; Use the median filtering method to denoise the thermal imaging image to obtain a denoised thermal imaging image; Use the sliding window method to calculate the average value of each pixel point in the thermal imaging images in the past n time periods to obtain an average thermal imaging image; Calculate the error of each pixel point between the average thermal imaging image and the denoised thermal imaging image to obtain an error thermal imaging image; For the error thermal imaging image, calculate the mean and standard deviation of each pixel point in the error thermal imaging image, and set the dynamic error threshold based on the mean and standard deviation; In the error thermal imaging image, mark the pixel points with errors greater than the dynamic error threshold in blue, assign a unique label 0 to each blue pixel point in the error thermal imaging image, use the traversal method to detect whether the labels of the upper, lower, left, right, and four diagonal directions of the blue pixel points with label 0 are the same, and use the connected component labeling method to identify the blue areas formed in the error thermal imaging image. The blue areas are the abnormal areas.

6. The intelligent inspection method for leakage of gas turbine with multi-sensor data fusion according to claim 5, characterized in that, For the abnormal area, trigger the multispectral camera to conduct inspection on the abnormal area. The specific method of using the dynamic fusion method to obtain the band comprehensive data includes: When the abnormal area is determined, trigger the multispectral camera. The inspection frequency of the multispectral camera is f, and use the multispectral camera to collect N bands of each pixel point in the abnormal area; Each band includes radiation intensity, absorption bandwidth, spectral reflectivity, and normalized absorption index data; For the N bands of each pixel point, use the dynamic adaptive method to obtain D bands; For the D bands, use the weighted average method to obtain the band comprehensive data of each pixel point, and perform weighted fusion on the band comprehensive data of each pixel point to obtain the band comprehensive data of the abnormal area.

7. A method for intelligent inspection of gas turbine leakage with multi-sensor data fusion according to claim 6, characterized in that, The specific method of using the dynamic adaptive method to obtain D bands for the N bands of each pixel point includes: When the multispectral camera is triggered, the multispectral camera obtains the N-band data of each pixel point in the abnormal area, and use the spectral angle matching analysis correction method to obtain the response intensity data of the N bands of each pixel point to the abnormal area; Sort the response intensities of the N bands to the abnormal area in descending order, use the expert experience method to set the response intensity threshold, retain the bands greater than or equal to the response threshold, and remove the bands less than the response threshold. Then each pixel point retains D bands.

8. The intelligent inspection method for gas turbine leakage with multi-sensor data fusion according to claim 7, characterized in that, The specific method of using the spectral angle matching analysis correction method to obtain the response intensity data of the N bands of each pixel point to the abnormal area includes: Construct N spectral vectors from the radiation intensity, absorption bandwidth, spectral reflectivity, and normalized absorption index data of N bands of pixel points; Calculate the spectral angle between the spectral vector of each band and the spectral feature vector of the abnormal area using the spectral angle formula; For each band, use the Pearson correlation method to calculate the linear relationship between the spectral vector of each band and the spectral feature vector of the abnormal area to obtain the correlation value; Use the correlation value to correct the spectral angle between the spectral vector of each band and the spectral data vector of the abnormal area to obtain the correlation intensity data for each band; through the correlation intensity data, calculate the response intensity data.

9. The intelligent inspection method for gas turbine leakage with multi-sensor data fusion according to claim 8, characterized in that, The method for obtaining the spectral feature vector of the abnormal area includes: It is known that the abnormal area includes H pixel points. For each pixel point, spectral data of N bands are extracted, and the spectral data are constructed into spectral vectors, and the spectral vectors of H pixel points are weighted and calculated using the weighted average method, then the spectral feature vector of the abnormal area is obtained.

10. A method for intelligent inspection of gas turbine leakage with multi-sensor data fusion according to claim 9, characterized in that, The specific method for triggering the infrared thermal imaging camera and the multispectral camera to perform high-frequency inspection on the abnormal area when the area abnormality is caused by the device itself includes: For the abnormal area caused by the device itself, trigger high-frequency inspection, where the inspection frequency of the infrared thermal imaging camera is set to 2u, and the inspection frequency of the multispectral camera is set to 2f.

Images