High-precision non-contact blade vibration signal acquisition and feature extraction method and system

Abstract

This invention relates to the field of turbomachinery data processing technology, and particularly to a high-precision non-contact blade vibration signal acquisition and feature extraction method and system. The method involves: step S1 to acquire the sensing distance; step S2 to acquire the blade arrival time and theoretical arrival time; step S3 to fit the blade vibration signal; step S4 to extract features from the blade vibration signal; step S5 to perform real-time correction during feature extraction; step S6 to analyze the processed blade vibration features; step S7 to increase the rationality of the blade vibration parameter model; step S8 to reduce the impact of excessively long blade usage time on the accuracy of result confidence determination; and step S9 to optimize the preset blade usage time. The system includes a signal acquisition module, a signal acquisition module, a feature extraction module, a real-time monitoring module, and a feedback optimization module.

Description

Technical Field

[0001] This invention relates to the field of turbomachinery data processing technology, and in particular to a high-precision non-contact blade vibration signal acquisition and feature extraction method and system. Background Technology

[0002] In high-speed turbomachinery, non-contact vibration monitoring is crucial for ensuring blade safety. However, this field still has significant shortcomings. First, sensors typically operate in harsh environments with high temperatures and electromagnetic interference, making signals susceptible to noise contamination. Furthermore, sparsely distributed sensors result in insufficient spatial sampling of vibration signals, introducing uncertainty into subsequent analysis. Second, existing feature extraction methods largely rely on manual design based on expert knowledge, making them insensitive to complex, coupled early-stage faults and lacking adaptive capabilities. Traditional methods face challenges in terms of accuracy and reliability. These shortcomings limit the accuracy and timeliness of early fault warnings.

[0003] Chinese patent application CN120883037A discloses a method for predicting the life of rolling bearings under arbitrary load conditions. This method calculates the stress amplification factor range ΔKII accompanying the rolling element's passage under load conditions based on the size of inclusions near the rolling surface of the rolling bearing, and predicts the inclusion initiation peeling life of the rolling bearing based on the stress amplification factor range ΔKII. However, this method still suffers from insufficient accuracy of blade vibration signals and low feature extraction efficiency due to a lack of analysis of the core scenario and monitoring of blade condition. Summary of the Invention

[0004] To address this, the present invention provides a high-precision non-contact blade vibration signal acquisition and feature extraction method and system, which overcomes the problems of insufficient accuracy of blade vibration signals and low feature extraction efficiency caused by the lack of analysis of core scenarios and monitoring of blade status in the prior art.

[0005] To achieve the above objectives, on the one hand, the present invention provides a high-precision non-contact blade vibration signal acquisition and feature extraction method, the method comprising: Step S1: Acquire sensing distance and target environment data; Step S2: Obtain the blade arrival time and the theoretical blade arrival time based on the sensing distance; Step S3: Obtain the arrival time difference based on the blade arrival time and the theoretical arrival time of the blade, and obtain the blade vibration signal based on the arrival time difference; Step S4: Extract features from the blade vibration signal to obtain the processed blade vibration features; Step S5: Acquire the core scene and correct the feature extraction process based on the core scene; Step S6: Output the blade feature extraction results and the confidence level of the results based on the processed blade vibration characteristics; Step S7: Make basic adjustments to the process of acquiring blade vibration signals based on the confidence level of the results; Step S8: Obtain the blade usage time and update the basic adjustment process based on the blade usage time. Step S9: Obtain the environmental impact level based on the target environmental data, and perform basic optimization on the basic update process based on the environmental impact level.

[0006] Furthermore, in step S1, when acquiring the sensing distance, a microwave signal is emitted to the blade through a sensor probe and a reflected signal is received. The microwave signal and the reflected signal are input to a mixer to obtain the phase information intermediate frequency signal output by the mixer. The phase information intermediate frequency signal is then input to an ADC analog-to-digital converter to acquire the phase information digital signal, and the sensing distance is acquired based on the phase information digital signal. In step S1, when collecting target environmental data, the target environmental data includes ambient temperature, salt spray concentration and dust particle size distribution. In this embodiment, ambient temperature is collected by a temperature sensor, salt spray concentration is collected by an isokinetic sampling probe, and dust particle size distribution is collected by a laser diffraction particle size analyzer. In step S2, when the blade arrival time and theoretical arrival time are obtained based on the sensing distance, the sensing distance acquisition process is subjected to multiple distance measurement processes to obtain a distance-time series. The distance-time series is then subjected to local minimum value search processing to obtain the minimum distance point. The time point corresponding to the minimum distance point is output as the blade arrival time, and the theoretical arrival time of the blade is obtained through bond phase measurement.

[0007] Further, in step S3, the arrival time difference tp is calculated based on the blade arrival time t1 and the theoretical arrival time t2, and tp is set to |t1-t2| to obtain the arrival time difference tp; In step S3, when acquiring the blade vibration signal based on the arrival time difference, the historical time signal library is used as the training set to train the recurrent neural network model to obtain the blade vibration parameter model, and the arrival time difference tp is input into the blade vibration parameter model to obtain the blade vibration signal output by the blade vibration parameter model.

[0008] Furthermore, in step S4, when extracting features from the blade vibration signal, the blade vibration signal is extracted using a feature extraction method to obtain the processed blade vibration features. The feature extraction method includes: Step A01: Perform data cleaning processing on the blade vibration signal using Kalman filtering to obtain a clean blade vibration signal; Step A02: Perform preliminary feature extraction on the vibration signal of the clean blade to obtain preliminary blade vibration features; Step A03: Normalize the preliminary characteristics of the blade vibration to obtain the processed blade vibration characteristics.

[0009] Furthermore, in step S5, when acquiring the core scene and correcting the feature extraction process based on the core scene, the core scene includes the speed stability scene and the unit start-up and shutdown scene. When the core scenario is a stable rotation speed scenario, the feature extraction process is corrected, the vibration signal of the clean blade in the feature extraction is subjected to synchronous and asynchronous vibration analysis to obtain the first vibration monitoring result, the first vibration monitoring result is marked to obtain the marked clean blade vibration signal, and the marked clean blade vibration signal is subjected to preliminary feature extraction again. When the core scenario is the unit start-up and shutdown scenario, the feature extraction process is corrected, the preliminary feature extraction is updated, and the updated preliminary feature extraction is obtained. The preliminary feature extraction is then replaced with the updated preliminary feature extraction.

[0010] Furthermore, in step S6, when outputting the blade feature extraction result based on the processed blade vibration characteristics, a feature recognition model is constructed using a feature recognition model construction method, and the processed blade vibration characteristics are input into the feature recognition model to obtain the blade feature extraction result and result confidence level output by the feature recognition model.

[0011] Further, in step S7, when making basic adjustments to the blade vibration signal acquisition process based on the result confidence level, the result confidence level Yp is compared with the preset result confidence level Yp0. The compliance of the result confidence level is judged based on the comparison result, and the blade vibration signal acquisition process is adjusted based on the judgment result, wherein: When Yp≥Yp0, the confidence level of the judgment result is considered to be up to standard, and no basic adjustments are made to the process of acquiring the blade vibration signal. When Yp < Yp0, the confidence level of the judgment result is deemed unqualified. The acquisition process of the blade vibration signal is adjusted, and the physical constraints of the blade vibration are added as a penalty term to the blade vibration parameter model to obtain the adjusted blade vibration parameter model. The arrival time difference tp is then re-inputted into the adjusted blade vibration parameter model.

[0012] Further, in step S8, the blade usage time is acquired, and the basic adjustment process is updated based on the blade usage time. The total operating time from the engine's initial operation to the current moment is taken as the blade usage time tc. The blade usage time tc is compared with the preset blade usage time tc0. The state of the blade usage time is judged based on the comparison result, and the basic adjustment process is updated based on the judgment result, wherein: When tc≤tc0, the blade usage time is determined to be short, and no basic update is performed during the basic adjustment process; When tc > tc0, the blade usage time is determined to be long, the basic adjustment process is updated, and the result confidence level is directly judged as unqualified.

[0013] Further, in step S9, when the environmental influence intensity is obtained based on the target environmental data and the basic update process is optimized based on the environmental influence intensity, the target environmental data is normalized to obtain a set of normalized environmental data values. The set of normalized environmental data values ​​includes environmental temperature value, salt spray concentration value, and dust particle size distribution value. The environmental influence intensity Zc is calculated based on the environmental temperature value ws, salt spray concentration value cs, dust particle size distribution value bs, environmental temperature value weight v1, salt spray concentration value weight v2, and dust particle size distribution value weight v3. Zc is set as Zc = ws × v1 + cs × v2 + bs × v3, and v1 = 0.3, v2 = 0.4, v3 = 0.3. The environmental influence magnitude Zc is compared with the preset environmental influence magnitude Zc0. Based on the comparison result, the state of the environmental influence magnitude is determined, and the basic update process is optimized based on the determination result. When Zc≤Zc0, the environmental influence is determined to be a weak influence, and no basic optimization is performed on the basic update process; When Zc > Zc0, the environmental influence is determined to be strong. Basic optimization is performed on the basic update process. The preset blade usage time tc0 is optimized according to the basic optimization coefficient α. The optimized preset blade usage time is set as tc01, and tc01 = tc0 × α, 0.5 < α < 1. The optimized preset blade usage time tc0 is used as the preset blade usage time tc, and the blade usage time tc is compared with the preset blade usage time tc0 again.

[0014] On the other hand, the present invention also provides a system for high-precision non-contact blade vibration signal acquisition and feature extraction, the system comprising: The signal acquisition module is used to acquire the sensing distance and to acquire the blade arrival time and the theoretical blade arrival time based on the sensing distance. The signal acquisition module is used to acquire the time difference of arrival based on the blade's arrival time and theoretical arrival time, and to acquire the blade vibration signal based on the time difference of arrival. The feature extraction module is used to extract features from the blade vibration signal, obtain the processed blade vibration features, acquire the core scene, and correct the feature extraction process based on the core scene. The real-time monitoring module is used to output the blade feature extraction results and the confidence level of the results based on the processed blade vibration characteristics, and to make basic adjustments to the blade vibration signal acquisition process based on the confidence level of the results. The feedback optimization module is used to acquire the blade usage time and perform basic updates to the basic adjustment process based on the blade usage time. It is also used to acquire the environmental impact intensity and perform basic optimization to the basic update process based on the environmental impact intensity.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: the method acquires the sensing distance in step S1, and acquires the blade arrival time and theoretical arrival time in step S2; the method also fits the blade vibration signal in step S3 to intuitively measure the health status of the blade structure, thereby enabling real-time health monitoring of high-speed rotating turbomachinery and reducing the risks and losses caused by blade failures; furthermore, the method extracts features from the blade vibration signal in step S4, removes noise from the blade vibration signal, and converts it into high-quality, processed blade vibration features that can be quickly recognized by computers and models, thereby improving the efficiency of signal feature extraction and subsequent processing, saving computing resources, and further... 5. Based on the monitoring requirements of high-speed rotating turbomachinery under different core scenarios, the feature extraction process is corrected in real time to increase the scenario rationality of the processed blade vibration features. In step S6, the processed blade vibration features are analyzed to obtain the blade feature extraction results, so as to monitor the specific situation of the impeller system in high-speed rotating turbomachinery in real time, so as to facilitate prevention and maintenance. In step S7, the rationality of the blade vibration parameter model is increased, thereby improving the accuracy of the blade vibration signal. The method also reduces the impact of excessive blade usage time on the accuracy of result confidence judgment in step S8, and optimizes the preset blade usage time in step S9 to reduce the impact of environmental influence on basic updates. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the high-precision non-contact blade vibration signal acquisition and feature extraction method of this embodiment; Figure 2 This is a flowchart illustrating the feature extraction method of this embodiment; Figure 3 This is a schematic diagram of the system structure of the high-precision non-contact blade vibration signal acquisition and feature extraction method in this embodiment. Detailed Implementation

[0017] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0018] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0019] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0020] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0021] Please see Figure 1 As shown, this is a flowchart illustrating the high-precision non-contact blade vibration signal acquisition and feature extraction method of this embodiment. The method includes: Step S1: Acquire sensing distance and target environment data; Step S2: Obtain the blade arrival time and the theoretical blade arrival time based on the sensing distance; Step S3: Obtain the arrival time difference based on the blade arrival time and the theoretical arrival time of the blade, and obtain the blade vibration signal based on the arrival time difference; Step S4: Extract features from the blade vibration signal to obtain the processed blade vibration features; Step S5: Acquire the core scene and correct the feature extraction process based on the core scene; Step S6: Output the blade feature extraction results and the confidence level of the results based on the processed blade vibration characteristics; Step S7: Make basic adjustments to the process of acquiring blade vibration signals based on the confidence level of the results; Step S8: Obtain the blade usage time and update the basic adjustment process based on the blade usage time. Step S9: Obtain the environmental impact level based on the target environmental data, and perform basic optimization on the basic update process based on the environmental impact level.

[0022] Specifically, the high-precision non-contact blade vibration signal acquisition and feature extraction method is applied to high-speed rotating turbomachinery, such as aero engines and steam turbines. The method acquires blade vibration signals, extracts features from these signals, and optimizes the feature extraction process in real time to improve the accuracy of signal acquisition and the efficiency of feature extraction. Specifically, step S1 acquires the sensing distance, and step S2 acquires the blade arrival time and theoretical arrival time. Step S3 fits the blade vibration signal to provide a visual assessment of the blade's structural health, enabling real-time health monitoring of the high-speed rotating turbomachinery and reducing the risks and losses caused by blade failures. Step S4 extracts features from the blade vibration signal, removing noise and converting it to high-quality signals. The computer and model can quickly identify the processed blade vibration characteristics to improve the efficiency of signal feature extraction and subsequent processing, saving computing resources. Step S5 further corrects the feature extraction process in real time according to the monitoring needs of high-speed rotating turbomachinery in different core scenarios, increasing the scenario rationality of the processed blade vibration characteristics. Step S6 analyzes the processed blade vibration characteristics to obtain blade feature extraction results, enabling real-time monitoring of the specific condition of the impeller system in high-speed rotating turbomachinery for prevention and maintenance. Step S7 increases the rationality of the blade vibration parameter model, thereby improving the accuracy of the blade vibration signal. Step S8 reduces the impact of excessively long blade usage time on the accuracy of result confidence judgment. Step S9 optimizes the preset blade usage time to reduce the impact of environmental influence on basic updates.

[0023] Specifically, in step S1, when acquiring the sensing distance, a microwave signal is emitted to the blade through a sensor probe and a reflected signal is received. The microwave signal and the reflected signal are input into a mixer to obtain the phase information intermediate frequency signal output by the mixer. The phase information intermediate frequency signal is then input into an ADC analog-to-digital converter to acquire the phase information digital signal, and the sensing distance is acquired based on the phase information digital signal. In step S1, when collecting target environmental data, the target environmental data includes ambient temperature, salt spray concentration, and dust particle size distribution. In this embodiment, ambient temperature is collected by a temperature sensor, salt spray concentration is collected by an isokinetic sampling probe, and dust particle size distribution is collected by a laser diffraction particle size analyzer.

[0024] Specifically, the sensor probe refers to hardware integrating a microwave antenna and front-end circuitry to transmit microwaves to the blades; the blades refer to the fan blades in a high-speed rotating impeller; the reflected signal refers to the microwave signal reflected back after the emitted microwave signal reaches the blades; the mixer refers to an electronic device that converts the microwave signal and the reflected signal into a phase information intermediate frequency signal; the phase information intermediate frequency signal refers to an intermediate frequency signal containing the phase change between the microwave signal and the reflected signal; the ADC analog-to-digital converter refers to a device that converts a continuous analog signal into a discrete digital signal; and the phase information digital signal refers to a discrete digital signal representing the phase information intermediate frequency signal. This embodiment does not limit the specific method of obtaining the sensing distance based on the phase information digital signal. For example, the phase information digital signal can be input into a digital signal processor to obtain the phase angle, and the sensing distance can be obtained based on the phase angle using a phase expansion algorithm. The ambient temperature refers to the temperature value of the environment around the blades; the salt spray concentration refers to the mass of soluble salt contained in a unit volume of air in the environment around the blades; and the dust particle size distribution refers to the number of dust particles of different diameters in the dust particle group in the environment around the blades.

[0025] Specifically, in step S1, the sensing distance is monitored in real time to perform predictive maintenance on the high-speed rotating impeller machinery in order to prevent catastrophic accidents.

[0026] Specifically, in step S2, when acquiring the blade arrival time and the theoretical blade arrival time based on the sensing distance, the sensing distance acquisition process is subjected to multiple distance measurement processes to obtain a distance-time series. The distance-time series is then subjected to local minimum value search processing to obtain the minimum distance point. The time point corresponding to the minimum distance point is output as the blade arrival time, and the theoretical blade arrival time is acquired through bond phase measurement.

[0027] Specifically, the multiple ranging process refers to acquiring the sensing distance multiple times according to a preset sampling rate and marking the time point of each acquired sensing distance. This embodiment does not limit the preset sampling rate, and those skilled in the art can freely choose it according to actual needs. For example, this embodiment sets the preset sampling rate to 100kHz based on engineering experience. The distance-time series refers to the sequence composed of multiple sensing distances and their corresponding time points obtained after multiple ranging processes. The local minimum value finding process refers to the process of finding the minimum distance point in the distance-time series. The minimum distance point refers to the smallest sensing distance in the distance-time series.

[0028] Specifically, in step S2, the arrival time and theoretical arrival time of the blade are obtained so that the blade vibration signal can be obtained subsequently based on the arrival time and theoretical arrival time of the blade, thereby improving the accuracy and efficiency of blade vibration signal acquisition.

[0029] Specifically, in step S3, the arrival time difference tp is calculated based on the blade arrival time t1 and the theoretical arrival time t2, and tp is set to |t1-t2| to obtain the arrival time difference tp; In step S3, when acquiring the blade vibration signal based on the arrival time difference, the historical time signal library is used as the training set to train the recurrent neural network model to obtain the blade vibration parameter model, and the arrival time difference tp is input into the blade vibration parameter model to obtain the blade vibration signal output by the blade vibration parameter model.

[0030] Specifically, the historical time signal library refers to a database that stores the arrival time differences of historical events and their corresponding blade vibration signals; the recurrent neural network model refers to the basic model architecture for constructing the blade vibration parameter model; and the blade vibration signal refers to the physical quantity that comprehensively describes the dynamic motion characteristics of the blade obtained from the blade vibration parameter model.

[0031] Specifically, in step S3, the blade vibration signal is fitted to make the blade structure health status more intuitive, thereby enabling real-time health monitoring of the high-speed rotating impeller and reducing the risks and losses caused by blade failure.

[0032] Specifically, in step S4, when extracting features from the blade vibration signal, the blade vibration signal is extracted using a feature extraction method to obtain the processed blade vibration features.

[0033] Specifically, in step S4, by extracting features from the blade vibration signal, the noise in the blade vibration signal is removed and converted into high-quality processed blade vibration features that can be quickly identified by computers and models, so as to improve the efficiency of signal feature extraction and subsequent processing and save computing resources.

[0034] Specifically, in step S5, when acquiring the core scene and correcting the feature extraction process based on the core scene, the core scene includes the speed stability scene and the unit start-up and shutdown scene. When the core scenario is a stable rotation speed scenario, the feature extraction process is corrected, the vibration signal of the clean blade in the feature extraction is subjected to synchronous and asynchronous vibration analysis to obtain the first vibration monitoring result, the first vibration monitoring result is marked to obtain the marked clean blade vibration signal, and the marked clean blade vibration signal is subjected to preliminary feature extraction again. When the core scenario is the unit start-up and shutdown scenario, the feature extraction process is corrected, the preliminary feature extraction is updated, and the updated preliminary feature extraction is obtained. The preliminary feature extraction is then replaced with the updated preliminary feature extraction.

[0035] Specifically, the stable speed scenario refers to the scenario where a high-speed rotating impeller operates continuously at a nearly fixed speed and load. The unit start-up and shutdown scenario refers to the transitional scenario where a high-speed rotating impeller starts from a standstill and rises to its rated speed, and then decelerates from its rated speed back to a standstill. The synchronous and asynchronous vibration analysis includes synchronous vibration analysis and asynchronous vibration analysis. The synchronous vibration analysis refers to the process of monitoring the vibration response components in the feature-extracted clean blade vibration signal that are precisely synchronized with the blade rotor speed, such as monitoring unbalanced responses. The asynchronous vibration analysis refers to the process of monitoring the vibration response components in the feature-extracted clean blade vibration signal that are unrelated to the blade rotor speed, such as monitoring flutter caused by aeroelastic instability. The updated preliminary feature extraction refers to the process of performing a Fast Fourier Transform on segments of the clean blade vibration signal to extract the frequency domain features in the clean blade vibration signal. This embodiment does not limit the specific segmentation method of performing Fast Fourier Transform on segments of the clean blade vibration signal. Those skilled in the art can freely choose according to actual needs, such as segmenting the clean blade vibration signal according to a preset time interval of 0.1s.

[0036] Specifically, in step S5, the feature extraction process is corrected in real time according to the monitoring needs of high-speed rotating impeller machinery in different scenarios, so as to increase the scenario rationality of the processed blade vibration features and thus improve the accuracy of blade vibration signal feature extraction.

[0037] Specifically, in step S6, when outputting the blade feature extraction result based on the processed blade vibration characteristics, a feature recognition model is constructed using a feature recognition model construction method, and the processed blade vibration characteristics are input into the feature recognition model to obtain the blade feature extraction result and result confidence level output by the feature recognition model.

[0038] Specifically, the feature recognition model refers to a gradient boosting decision tree model that takes the processed blade vibration characteristics as input data and the blade feature extraction results as output data. This embodiment does not limit the construction method of the feature recognition model. Those skilled in the art can freely choose according to actual needs. For example, the gradient boosting decision tree model can be trained by using historical processed blade vibration characteristics and their corresponding blade feature extraction results as training sets to obtain the feature recognition model. The blade feature extraction results refer to the specific situation of the impeller system in the high-speed rotating turbomachinery obtained by the feature recognition model, such as normal, unbalanced, misaligned, and loose. The result confidence level refers to a numerical value that measures the consistency of the feature recognition model in making decisions. For example, if the voting consistency of the gradient boosting decision tree model for the blade feature extraction results is 95%, then the final result confidence level is 95%.

[0039] Specifically, in step S6, the vibration characteristics of the processed blades are analyzed by a feature recognition model to obtain blade feature extraction results, so as to monitor the specific situation of the impeller system in high-speed rotating turbomachinery in real time, so as to facilitate prevention and maintenance.

[0040] Specifically, in step S7, when making basic adjustments to the blade vibration signal acquisition process based on the result confidence level, the result confidence level Yp is compared with the preset result confidence level Yp0. Based on the comparison result, the attainment of the result confidence level is judged, and based on the judgment result, the blade vibration signal acquisition process is made basic adjustments, wherein: When Yp≥Yp0, the confidence level of the judgment result is considered to be up to standard, and no basic adjustments are made to the process of acquiring the blade vibration signal. When Yp < Yp0, the confidence level of the judgment result is deemed unqualified. The acquisition process of the blade vibration signal is adjusted, and the physical constraints of the blade vibration are added as a penalty term to the blade vibration parameter model to obtain the adjusted blade vibration parameter model. The arrival time difference tp is then re-inputted into the adjusted blade vibration parameter model.

[0041] Specifically, the preset result confidence level refers to a preset value used to judge the compliance status of the result confidence level. This embodiment does not limit the specific value setting of the preset result confidence level Yp0. Those skilled in the art can freely choose according to actual needs. For example, this embodiment sets Yp0=95% based on historical experience. The compliance status of the result confidence level refers to the degree of compliance of the result confidence level judged based on the result confidence level and the preset result confidence level. The physical constraints of blade vibration refer to the constraints on blade vibration added to the loss function of the blade vibration parameter model, such as the motion differential equation and boundary conditions. In this embodiment, when the compliance status of the result confidence level is judged to be unacceptable for three consecutive times, the basic adjustment is stopped and the compliance status of the confidence level is pushed to the system control interaction interface.

[0042] Specifically, in step S7, a penalty term is added to the blade vibration parameter model to prevent overfitting when the confidence level of the result does not meet the standard, thereby increasing the rationality of the blade vibration parameter model and improving the accuracy of the blade vibration signal.

[0043] Specifically, in step S8, the blade usage time is acquired, and the basic adjustment process is updated based on the blade usage time. The total operating time from the engine's initial operation to the current moment is taken as the blade usage time tc. The blade usage time tc is compared with the preset blade usage time tc0. Based on the comparison result, the state of the blade usage time is determined, and the basic adjustment process is updated based on the determination result. When tc≤tc0, the blade usage time is determined to be short, and no basic update is performed during the basic adjustment process; When tc > tc0, the blade usage time is determined to be long, the basic adjustment process is updated, and the result confidence level is directly judged as unqualified.

[0044] Specifically, the engine refers to the engine in a high-speed rotating turbomachinery. The preset blade usage time refers to a preset value for judging the state of blade usage time. This embodiment does not limit the specific value of the preset blade usage time tc0. Those skilled in the art can freely choose according to actual needs. For example, this embodiment sets tc0 = 10 years based on engineering experience. The state of blade usage time refers to the length of time the blade is used. The state of blade usage time includes short time and long time.

[0045] Specifically, in step S8, the confidence level of the result is updated based on the status of the leaf usage time, so as to reduce the impact of excessive leaf usage time on the accuracy of the result confidence level determination, thereby improving the accuracy of the leaf feature extraction result.

[0046] Specifically, in step S9, when the environmental influence intensity is obtained based on the target environmental data and the basic update process is optimized based on the environmental influence intensity, the target environmental data is normalized to obtain a set of normalized environmental data values. The set of normalized environmental data values ​​includes environmental temperature value, salt spray concentration value, and dust particle size distribution value. The environmental influence intensity Zc is calculated based on the environmental temperature value ws, salt spray concentration value cs, dust particle size distribution value bs, environmental temperature value weight v1, salt spray concentration value weight v2, and dust particle size distribution value weight v3. Zc is set as Zc = ws × v1 + cs × v2 + bs × v3, and v1 = 0.3, v2 = 0.4, v3 = 0.3. The environmental influence magnitude Zc is compared with the preset environmental influence magnitude Zc0. Based on the comparison result, the state of the environmental influence magnitude is determined, and the basic update process is optimized based on the determination result. When Zc≤Zc0, the environmental influence is determined to be a weak influence, and no basic optimization is performed on the basic update process; When Zc > Zc0, the environmental influence is determined to be strong. Basic optimization is performed on the basic update process. The preset blade usage time tc0 is optimized according to the basic optimization coefficient α. The optimized preset blade usage time is set as tc01, and tc01 = tc0 × α, 0.5 < α < 1. The optimized preset blade usage time tc0 is used as the preset blade usage time tc, and the blade usage time tc is compared with the preset blade usage time tc0 again.

[0047] Specifically, the environmental influence degree refers to the degree of influence of the environment on the blade vibration. The preset environmental influence degree refers to the preset value for judging the state of the environmental influence degree. This embodiment does not limit the specific value setting of the preset environmental influence degree Zc0. Those skilled in the art can freely choose according to actual needs. For example, in this embodiment, Zc0 is set to 0.3. The state of the environmental influence degree includes weak influence and strong influence. The environmental temperature value weight refers to the value that measures the importance of the environmental temperature value in the environmental influence degree. The salt spray concentration value weight refers to the value that measures the importance of the salt spray concentration value in the environmental influence degree. The dust particle size distribution value weight refers to the value that measures the importance of the dust particle size distribution value in the environmental influence degree. In this embodiment, based on engineering experience, the salt spray concentration value is used as the main influencing factor of the blade environment, and v1=0.3, v2=0.4, and v3=0.3 are set.

[0048] Specifically, in step S9, by judging the state of the environmental impact intensity, when the state of the environmental impact intensity is strong, the preset blade usage time is reduced in a timely manner, so as to reduce the impact of the environmental impact intensity on the basic update.

[0049] Please see Figure 2 The diagram shown is a flowchart of the feature extraction method in this embodiment. The feature extraction method includes: Step A01: Perform data cleaning processing on the blade vibration signal using Kalman filtering to obtain a clean blade vibration signal; Step A02: Perform preliminary feature extraction on the vibration signal of the clean blade to obtain preliminary blade vibration features; Step A03: Normalize the preliminary characteristics of the blade vibration to obtain the processed blade vibration characteristics.

[0050] Specifically, the data cleaning process refers to the process of removing noise from the blade vibration signal. The preliminary feature extraction refers to the process of extracting quantitative indicators that can characterize the health status of the blade from the cleaned blade vibration signal. In this embodiment, the time-domain features of the cleaned blade vibration signal are extracted by performing autocorrelation analysis on the cleaned blade vibration signal. The frequency-domain features of the cleaned blade vibration signal are also extracted by performing fast Fourier transform. The time-domain features and frequency-domain features of the vibration signal are used as the preliminary features of the blade vibration. The normalization process refers to the process of mapping the preliminary features of the blade vibration to the [0,1] interval using the normalization function in the Paython library. The processed blade vibration features refer to the relevant feature values ​​that characterize the blade state after feature extraction, such as 1X amplitude, kurtosis, and spectral centroid.

[0051] Please see Figure 3 The diagram shown is a structural schematic of the system for the high-precision non-contact blade vibration signal acquisition and feature extraction method of this embodiment. The system includes: The signal acquisition module is used to acquire the sensing distance and to acquire the blade arrival time and the theoretical blade arrival time based on the sensing distance. The signal acquisition module is used to acquire the time difference of arrival based on the arrival time of the blade and the theoretical arrival time of the blade, and to acquire the blade vibration signal based on the time difference of arrival. The signal acquisition module is connected to the signal acquisition module. The feature extraction module is used to extract features from the blade vibration signal to obtain the processed blade vibration features, acquire the core scene, and correct the feature extraction process based on the core scene. The feature extraction module is connected to the signal acquisition module. A real-time monitoring module is used to output the blade feature extraction results and result confidence scores based on the processed blade vibration characteristics, and to make basic adjustments to the blade vibration signal acquisition process based on the result confidence scores. The real-time monitoring module is connected to the feature extraction module. The feedback optimization module is used to acquire the blade usage time and perform basic updates to the basic adjustment process based on the blade usage time. It is also used to acquire the environmental impact intensity and perform basic optimization to the basic update process based on the environmental impact intensity. The feedback optimization module is connected to the real-time monitoring module.

[0052] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A high-precision non-contact blade vibration signal acquisition and feature extraction method, characterized in that, The method includes: Step S1: Acquire sensing distance and target environment data; Step S2: Obtain the blade arrival time and the theoretical blade arrival time based on the sensing distance; Step S3: Obtain the arrival time difference based on the blade arrival time and the theoretical arrival time of the blade, and obtain the blade vibration signal based on the arrival time difference; Step S4: Extract features from the blade vibration signal to obtain the processed blade vibration features; Step S5: Acquire the core scene and correct the feature extraction process based on the core scene; Step S6: Output the blade feature extraction results and the confidence level of the results based on the processed blade vibration characteristics; Step S7: Make basic adjustments to the process of acquiring blade vibration signals based on the confidence level of the results; Step S8: Obtain the blade usage time and update the basic adjustment process based on the blade usage time. Step S9: Obtain the environmental impact level based on the target environmental data, and perform basic optimization on the basic update process based on the environmental impact level.

2. The high-precision non-contact blade vibration signal acquisition and feature extraction method according to claim 1, characterized in that, In step S1, when acquiring the sensing distance, a microwave signal is emitted to the blade through a sensor probe and a reflected signal is received. The microwave signal and the reflected signal are input into a mixer to obtain the phase information intermediate frequency signal output by the mixer. The phase information intermediate frequency signal is then input into an ADC analog-to-digital converter to acquire the phase information digital signal, and the sensing distance is acquired based on the phase information digital signal. In step S1, when collecting target environmental data, the target environmental data includes ambient temperature, salt spray concentration and dust particle size distribution. Ambient temperature is collected by a temperature sensor, salt spray concentration is collected by an isokinetic sampling probe, and dust particle size distribution is collected by a laser diffraction particle size analyzer. In step S2, when the blade arrival time and theoretical arrival time are obtained based on the sensing distance, the sensing distance acquisition process is subjected to multiple distance measurement processes to obtain a distance-time series. The distance-time series is then subjected to local minimum value search processing to obtain the minimum distance point. The time point corresponding to the minimum distance point is output as the blade arrival time, and the theoretical arrival time of the blade is obtained through bond phase measurement.

3. The high-precision non-contact blade vibration signal acquisition and feature extraction method according to claim 2, characterized in that, In step S3, the arrival time difference tp is calculated based on the blade arrival time t1 and the theoretical arrival time t2, and tp is set to |t1-t2| to obtain the arrival time difference tp; In step S3, when acquiring the blade vibration signal based on the arrival time difference, the historical time signal library is used as the training set to train the recurrent neural network model to obtain the blade vibration parameter model, and the arrival time difference tp is input into the blade vibration parameter model to obtain the blade vibration signal output by the blade vibration parameter model.

4. The high-precision non-contact blade vibration signal acquisition and feature extraction method according to claim 3, characterized in that, In step S4, when extracting features from the blade vibration signal, the blade vibration signal is extracted using a feature extraction method to obtain the processed blade vibration features. The feature extraction method includes: Step A01: Perform data cleaning processing on the blade vibration signal using Kalman filtering to obtain a clean blade vibration signal; Step A02: Perform preliminary feature extraction on the vibration signal of the clean blade to obtain preliminary blade vibration features; Step A03: Normalize the preliminary characteristics of the blade vibration to obtain the processed blade vibration characteristics.

5. The high-precision non-contact blade vibration signal acquisition and feature extraction method according to claim 4, characterized in that, In step S5, when the core scene is acquired and the feature extraction process is corrected based on the core scene, the core scene includes the speed stability scene and the unit start-up and shutdown scene. When the core scenario is a stable rotation speed scenario, the feature extraction process is corrected, the vibration signal of the clean blade in the feature extraction is subjected to synchronous and asynchronous vibration analysis to obtain the first vibration monitoring result, the first vibration monitoring result is marked to obtain the marked clean blade vibration signal, and the marked clean blade vibration signal is subjected to preliminary feature extraction again. When the core scenario is the unit start-up and shutdown scenario, the feature extraction process is corrected, the preliminary feature extraction is updated, and the updated preliminary feature extraction is obtained. The preliminary feature extraction is then replaced with the updated preliminary feature extraction.

6. The high-precision non-contact blade vibration signal acquisition and feature extraction method according to claim 5, characterized in that, In step S6, when outputting the blade feature extraction result based on the processed blade vibration characteristics, a feature recognition model is constructed using a feature recognition model construction method, and the processed blade vibration characteristics are input into the feature recognition model to obtain the blade feature extraction result and result confidence level output by the feature recognition model.

7. The high-precision non-contact blade vibration signal acquisition and feature extraction method according to claim 6, characterized in that, In step S7, when making basic adjustments to the blade vibration signal acquisition process based on the result confidence level, the result confidence level Yp is compared with the preset result confidence level Yp0. The compliance of the result confidence level is judged based on the comparison result, and the blade vibration signal acquisition process is adjusted based on the judgment result. Specifically: When Yp≥Yp0, the confidence level of the judgment result is considered to be up to standard, and no basic adjustments are made to the process of acquiring the blade vibration signal. When Yp < Yp0, the confidence level of the judgment result is deemed unqualified. The acquisition process of the blade vibration signal is adjusted, and the physical constraints of the blade vibration are added as a penalty term to the blade vibration parameter model to obtain the adjusted blade vibration parameter model. The arrival time difference tp is then re-inputted into the adjusted blade vibration parameter model.

8. The high-precision non-contact blade vibration signal acquisition and feature extraction method according to claim 7, characterized in that, In step S8, the blade usage time is acquired, and the basic adjustment process is updated based on the blade usage time. The total operating time from the engine's initial operation to the current moment is taken as the blade usage time tc. The blade usage time tc is compared with the preset blade usage time tc0. The state of the blade usage time is judged based on the comparison result, and the basic adjustment process is updated based on the judgment result. When tc≤tc0, the blade usage time is determined to be short, and no basic update is performed during the basic adjustment process; When tc > tc0, the blade usage time is determined to be long, the basic adjustment process is updated, and the result confidence level is directly judged as unqualified.

9. The high-precision non-contact blade vibration signal acquisition and feature extraction method according to claim 8, characterized in that, In step S9, when the environmental influence intensity is obtained based on the target environmental data and the basic update process is optimized based on the environmental influence intensity, the target environmental data is normalized to obtain a set of normalized environmental data values. The set of normalized environmental data values ​​includes environmental temperature value, salt spray concentration value, and dust particle size distribution value. The environmental influence intensity Zc is calculated based on the environmental temperature value ws, salt spray concentration value cs, dust particle size distribution value bs, environmental temperature value weight v1, salt spray concentration value weight v2, and dust particle size distribution value weight v3. Zc is set as Zc = ws × v1 + cs × v2 + bs × v3, and v1 = 0.3, v2 = 0.4, v3 = 0.

3. The environmental influence magnitude Zc is compared with the preset environmental influence magnitude Zc0. Based on the comparison result, the state of the environmental influence magnitude is determined, and the basic update process is optimized based on the determination result. When Zc≤Zc0, the environmental influence is determined to be a weak influence, and no basic optimization is performed on the basic update process; When Zc > Zc0, the environmental influence is determined to be strong. Basic optimization is performed on the basic update process. The preset blade usage time tc0 is optimized according to the basic optimization coefficient α. The optimized preset blade usage time is set as tc01, and tc01 = tc0 × α, 0.5 < α < 1. The optimized preset blade usage time tc0 is used as the preset blade usage time tc, and the blade usage time tc is compared with the preset blade usage time tc0 again.

10. A system for the high-precision non-contact blade vibration signal acquisition and feature extraction method as described in any one of claims 1-9, the system comprising: The signal acquisition module is used to acquire the sensing distance and to acquire the blade arrival time and the theoretical blade arrival time based on the sensing distance. The signal acquisition module is used to acquire the time difference of arrival based on the blade's arrival time and theoretical arrival time, and to acquire the blade vibration signal based on the time difference of arrival. The feature extraction module is used to extract features from the blade vibration signal, obtain the processed blade vibration features, acquire the core scene, and correct the feature extraction process based on the core scene. The real-time monitoring module is used to output the blade feature extraction results and the confidence level of the results based on the processed blade vibration characteristics, and to make basic adjustments to the blade vibration signal acquisition process based on the confidence level of the results. The feedback optimization module is used to acquire the blade usage time and perform basic updates to the basic adjustment process based on the blade usage time. It is also used to acquire the environmental impact intensity and perform basic optimization to the basic update process based on the environmental impact intensity.

Images