Vision-based rotating blade vibration measurement method

Abstract

The invention discloses a rotating blade vibration measurement method based on vision, and relates to the technical field of measurement. In order to solve the technical defect that circumferential and radial vibration displacements of a blade cannot be simultaneously and synchronously obtained in a high-speed rotating state in the prior art, the technical scheme provided by the invention is as follows: calculating the time when the blade reaches a preset measurement position based on a non-contact sensor, and respectively triggering a high-speed camera to collect circumferential and radial original images; the method comprises the following steps: performing micro motion amplification processing on an image, tracking pixel movement by using a sparse optical flow algorithm and converting the pixel movement into vibration displacement, and finally performing time domain and frequency domain feature extraction on circumferential and radial vibration data to obtain vibration feature parameters. The method is suitable for online vibration monitoring and analysis work of rotating blades of wind driven generators or aero-engines and the like under the condition of high-speed rotation.

Description

Technical Field

[0001] The invention relates to the field of measurement technology, and in particular to vision-based vibration measurement of rotating blades. Background Art

[0002] Rotating blades (such as those in wind turbines and aircraft engines) generate vibrations during operation, and this vibration directly impacts the safety and reliability of the equipment. To monitor and analyze the vibration characteristics of rotating blades, researchers at home and abroad have conducted research on various measurement techniques.

[0003] On the one hand, traditional vibration measurement methods primarily rely on contact sensors, such as strain gauges and accelerometers. These sensors require direct installation on the blade or blade root, indirectly inferring vibration characteristics by measuring strain or acceleration on the blade surface. However, due to factors such as centrifugal forces and aerodynamic forces involved in high-speed blade rotation, contact sensors often struggle to withstand the high-frequency impacts of high-speed environments. Furthermore, due to limited mounting locations, they cannot simultaneously capture vibration information in both the radial and circumferential directions. Furthermore, the sensor wiring is complex and susceptible to electromagnetic interference, which can also affect the aerodynamic performance of the blade itself.

[0004] Meanwhile, non-contact vibration measurement technologies have garnered widespread attention in recent years, primarily including laser Doppler vibrometry (LDV), digital image correlation (DIC), and high-speed imaging. Laser Doppler vibrometry can measure radial vibration with micron-level accuracy, but it is limited in the number of sampling points, and the instrument is expensive, making it inconvenient for simultaneous measurement of large blade areas. Digital image correlation methods can obtain displacement information by tracking markers or natural textures in the image. However, when the blade rotates at high speeds, image blur and the difficulty of synchronous acquisition increase, making measurement accuracy and real-time performance challenging. High-speed cameras combined with structured light or fringe projection techniques can improve dynamic measurement accuracy, but they require strict lighting conditions, and the subsequent image processing algorithms are complex and computationally intensive, making them difficult to implement in real-time on-site. Furthermore, most of these methods only capture vibration information in a single direction of the blade and lack the ability to simultaneously and synchronously measure circumferential and radial vibrations.

[0005] In recent years, some vision-based vibration measurement studies have begun to attempt to use video magnification technology to enhance subtle vibration signals and combine it with optical flow methods to extract pixel-level displacements. However, existing research often only focuses on planar structures or fixed targets. For high-speed rotating blades, especially multi-blade synchronous measurement, high-speed triggering and alignment, and circumferential and radial vibration separation and extraction, a mature overall solution has not yet been formed. For example, some studies use a monocular high-speed camera to capture blade vibration images at a fixed viewing angle. However, due to the lack of clock synchronization and position reference, it is difficult to accurately map the image displacement to the actual vibration displacement of the blade, and it is also impossible to eliminate the interference of the overall blade motion on the vibration measurement results. Other studies have attempted to use binocular or multi-cameras for three-dimensional displacement measurement, but this increases the complexity of the system and the difficulty of calibration. In addition, the experimental environment lighting requirements are stringent, and field applications are greatly limited.

[0006] In summary, the existing technology has the following defects: it is impossible to simultaneously and synchronously obtain the circumferential and radial vibration displacements of the blade under high-speed rotation; the existing visual measurement method has difficulty in accurately separating the overall motion of the blade from the vibration signal, and has high requirements for system synchronization triggering and image alignment; the measurement accuracy and real-time performance are insufficient, making it difficult to meet the online monitoring needs under complex working conditions. Summary of the Invention

[0007] In order to solve the technical defect in the prior art that the circumferential and radial vibration displacements of the blade cannot be obtained simultaneously and synchronously under high-speed rotation, the technical solution provided by the present invention is as follows: A vision-based rotating blade vibration measurement method, comprising: The step of calculating the time it takes for the blade to reach a preset measurement position according to the rotation speed value output by the non-contact rotation speed sensor; The step of collecting a circumferential original image of the blade when the blade reaches a preset measurement position, performing micro-motion amplification, and generating a circumferential enhanced video; Taking the current video as input, tracking the pixel movement of the blade leading edge or visual marker based on sparse optical flow, and converting the pixel movement into circumferential vibration displacement; The step of collecting a radial original image of the blade when the blade is triggered by a photoelectric switch, performing micro-motion amplification, and generating a radial enhanced video; Taking radially enhanced video as input, tracking the pixel movement of the leaf tip visual marker based on sparse optical flow, and converting the pixel movement into radial vibration displacement; The steps of performing time domain and frequency domain feature extraction on the circumferential vibration displacement and the radial vibration displacement respectively to obtain vibration characteristic parameters.

[0008] Furthermore, a preferred embodiment is provided in which the micro-motion amplification process automatically selects a phase amplification algorithm or a grayscale amplification algorithm according to the contrast between the leaf and the background.

[0009] Furthermore, a preferred embodiment is provided in which the sparse optical flow tracking algorithm uses the Lucas–Kanade method to track the pixel movement of the visual marker point at the leading edge or tip of the blade.

[0010] Furthermore, a preferred embodiment is provided in which the inter-frame alignment eliminates the circumferential motion of the entire blade by automatically detecting and aligning the positions of the blade tip visual markers.

[0011] Furthermore, a preferred embodiment is provided, wherein the time domain feature extraction includes calculation of peak displacement and root mean square displacement, and the frequency domain feature extraction includes analysis of the main vibration frequency and its harmonic amplitude.

[0012] Based on the same inventive concept, the present invention also provides a vision-based rotating blade vibration measurement device, comprising: A module for calculating the time it takes for the blade to reach a preset measurement position based on the speed value output by the non-contact speed sensor; A module that captures the original circumferential image of the blade when it reaches the preset measurement position, amplifies the slight movement, and generates a circumferential enhanced video; A module that takes the current video as input, tracks the pixel movement of the blade leading edge or visual markers based on sparse optical flow, and converts the pixel movement into circumferential vibration displacement; A module that captures the original radial image of the blade when it is triggered by a photoelectric switch, amplifies the tiny movement, and generates a radial enhanced video; A module that takes radially enhanced video as input, tracks the pixel movement of the leaf tip visual marker based on sparse optical flow, and converts the pixel movement into radial vibration displacement; A module that extracts time-domain and frequency-domain features of circumferential and radial vibration displacements to obtain vibration characteristic parameters.

[0013] Based on the same inventive concept, the present invention also provides a vision-based rotating blade vibration measurement system for implementing the method described above, comprising: A speed module for collecting the speed of the rotating shaft; A trigger module for calculating the blade arrival measurement moment; a circumferential processing module for driving the first high-speed camera to capture circumferential images and extract circumferential vibration displacement at a triggering moment; a photoelectric trigger module for detecting the passage of a blade tip and triggering a second high-speed camera to capture radial images; A radial processing module for performing inter-frame alignment of radial images and extracting radial vibration displacement; A feature analysis module is used to perform time domain and frequency domain analysis on circumferential and radial vibration displacements and output vibration characteristics.

[0014] Based on the same inventive concept, the present invention also provides a computer storage medium for storing a computer program. When the computer program is read by a computer, the computer executes the method described above.

[0015] Based on the same inventive concept, the present invention also provides a computer, comprising a processor and a storage medium. When the processor reads the computer program stored in the storage medium, the computer executes the method described above.

[0016] Based on the same inventive concept, the present invention also provides a computer program product, which is a computer program. When the computer program is executed, the method described above is implemented.

[0017] Compared with the prior art, the technical solution provided by the present invention is beneficial in that: By adopting a synchronously triggered dual high-speed camera system in the scheme, the effect of simultaneously collecting the circumferential and radial vibration information of the blade is achieved. This can fully record the vibration characteristics of the blade in different directions under high-speed rotation conditions. Compared with the existing research methods that only use monocular high-speed cameras or laser Doppler measurements to obtain vibration in only a single direction, this scheme significantly improves the comprehensiveness and reliability of vibration measurement.

[0018] By calculating the shaft speed in real time based on the eddy current sensor and deriving the theoretical arrival time of the blade, the camera is triggered to capture images at the precise moment, which significantly reduces the image blur and timing errors caused by the overall movement of the blade. Compared with the existing technology where there is no synchronous triggering or simple timed acquisition, which often leads to the problem of acquisition timing deviation, this solution improves the accuracy of image acquisition, thereby improving the accuracy of vibration extraction.

[0019] By applying a micro-motion amplification algorithm to the collected video, the blade vibration signal is enhanced at the sub-pixel level, amplifying the weak vibration signal to the visible range. Clear vibration information can be obtained even when the vibration amplitude is small. Compared with the limited signal sensitivity of traditional laser Doppler or digital image correlation methods, this scheme can still obtain vibration data with a high signal-to-noise ratio in a low-amplitude environment.

[0020] By using a sparse optical flow algorithm to extract pixel-level displacement in the enhanced video, and combining it with camera calibration and magnification to convert it into physical displacement, a contactless method for obtaining high-precision vibration displacement is achieved. This method overcomes the shortcomings of contact sensors in high-speed environments, such as insufficient impact resistance and limited installation space. At the same time, compared with the shortcomings of digital image correlation methods, which have high image quality requirements and large computational complexity, this scheme reduces computational complexity while ensuring accuracy.

[0021] By aligning the marker points of the radial vibration video and then performing micro-motion amplification and optical flow extraction, the interference of the overall rotational motion of the blade on the radial vibration measurement is effectively filtered out, making the extraction of the radial vibration signal purer and more accurate. Compared with the problem of existing research that often mixes the overall motion and vibration signals, this scheme has higher reliability and repeatability in signal separation.

[0022] It is suitable for online vibration monitoring and analysis of rotating blades of wind turbines or aircraft engines under high-speed rotation conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart of the method; Figure 2 This is a schematic diagram of the camera and photoelectric switch installation. DETAILED DESCRIPTION

[0024] In order to make the advantages and benefits of the technical solution provided by the present invention more clearly reflected, the technical solution provided by the present invention is now further described in detail with reference to the accompanying drawings, specifically: Embodiment 1: This embodiment provides a vision-based rotating blade vibration measurement method, including: The step of calculating the time it takes for the blade to reach a preset measurement position according to the rotation speed value output by the non-contact rotation speed sensor; The step of collecting a circumferential original image of the blade when the blade reaches a preset measurement position, performing micro-motion amplification, and generating a circumferential enhanced video; Taking the current video as input, tracking the pixel movement of the blade leading edge or visual marker based on sparse optical flow, and converting the pixel movement into circumferential vibration displacement; The step of collecting a radial original image of the blade when the blade is triggered by a photoelectric switch, performing micro-motion amplification, and generating a radial enhanced video; Taking radially enhanced video as input, tracking the pixel movement of the leaf tip visual marker based on sparse optical flow, and converting the pixel movement into radial vibration displacement; The steps of performing time domain and frequency domain feature extraction on the circumferential vibration displacement and the radial vibration displacement respectively to obtain vibration characteristic parameters.

[0025] The micro-motion amplification processing automatically selects the phase amplification algorithm or the grayscale amplification algorithm according to the contrast between the leaves and the background.

[0026] The sparse optical flow tracking algorithm uses the Lucas–Kanade method to track the pixel movement of the visual marker point on the leading edge or tip of the blade.

[0027] Inter-frame alignment eliminates the circumferential motion of the entire blade by automatically detecting and aligning the positions of the blade tip visual markers.

[0028] The time domain feature extraction includes the calculation of peak displacement and root mean square displacement, and the frequency domain feature extraction includes the analysis of the main vibration frequency and its harmonic amplitude.

[0029] A vision-based rotating blade vibration measurement system is also provided, for implementing the method described, comprising: A speed module for collecting the speed of the rotating shaft; A trigger module for calculating the blade arrival measurement moment; a circumferential processing module for driving the first high-speed camera to capture circumferential images and extract circumferential vibration displacement at a triggering moment; a photoelectric trigger module for detecting the passage of a blade tip and triggering a second high-speed camera to capture radial images; A radial processing module for performing inter-frame alignment of radial images and extracting radial vibration displacement; A feature analysis module is used to perform time domain and frequency domain analysis on circumferential and radial vibration displacements and output vibration characteristics.

[0030] Implementation Method 2: This implementation method further describes the technical solution provided in Implementation Method 1 in detail. Specifically: Step 1 The rotating blade assembly to be tested is fixed on the test bench, and a non-contact sensor for measuring the rotational speed, a photoelectric switch for triggering, a high-speed camera, and visual markers are attached to the blade tip to prepare for subsequent vibration signal collection and processing.

[0031] 1.1 Insert the shaft end of the rotating blade assembly to be tested into the main shaft flange of the test bench and fix it with locking bolts and positioning pins to ensure that the blade has no eccentricity or shaking when rotating at high speed.

[0032] 1.2 Set grooves at equal intervals on the main shaft, corresponding to the number of blades; install a non-contact speed sensor near the main shaft, with the sensor tip aligned with the side of the groove so that it can output a pulse signal when the groove rotates past, which is used for subsequent speed calculation.

[0033] 1.3 A photoelectric switch is set in front of the position where the blade is expected to pass along the axial direction of the blade's rotation plane. The position of the photoelectric switch is adjusted so that when the visual mark on the blade tip passes this position, it triggers the photoelectric switch to output a synchronization signal, which serves as the clock reference for the high-speed camera to collect data.

[0034] 1.4 Affix small, high-contrast visual markers to the blade tips to ensure that the markers do not affect the aerodynamic performance of the blades and can be clearly identified in images taken by high-speed cameras.

[0035] 1.5 Set up the first high-speed camera with its lens aligned with the center of the blade along the axis of rotation and at a certain distance from the blade tip. Adjust the focus to cover the entire blade profile. This camera is used to capture circumferential motion images when the blade approaches a specific position.

[0036] 1.6 Set up a second high-speed camera with its lens aligned radially with the blade tip motion trajectory, at a certain distance from the blade tip and adjust the focus. This camera is used to capture the radial motion image of the blade when the photoelectric switch is triggered.

[0037] 1.7 Ensure that the signal cables of the speed sensor, photoelectric switch, and two cameras are connected to the signal processing module in the test bench control cabinet. Set up a trigger signal distribution unit in the control cabinet to transmit the speed sensor signal to the speed calculation unit and the photoelectric switch signal to the trigger channel of the second camera. Reserve the trigger interface for the first camera.

[0038] 1.8 Shield all sensor and camera signal cables to avoid electromagnetic interference, and place camera power and trigger cables in a safe area to prevent collisions during blade rotation.

[0039] 1.9 After completing the hardware installation, calibrate the two high-speed cameras using a standard checkerboard calibration plate. This will determine the camera's intrinsic and extrinsic parameters, as well as the spatial transformation relationship between the camera and leaf coordinate systems. This will provide a basis for converting pixel displacement in the image to physical displacement.

[0040] Step 2 The pulse signal output by the speed sensor is used to calculate the real-time speed of the rotating shaft, and the speed value is passed to the subsequent module used to calculate the time when the blade reaches the measurement position.

[0041] 2.1 The speed sensor generates a pulse signal when the spindle groove passes through. The signal processing module uses the pulse interval of the pulse signal as the basis for speed calculation.

[0042] 2.2 The speed calculation unit obtains the actual speed value of the current rotating shaft according to the time interval between two consecutive pulses and the total number of grooves, and updates the value in real time.

[0043] 2.3 The speed calculation results are displayed in real time on the control software interface, allowing the operator to confirm whether the speed is stable within the preset test range. If necessary, the operator can manually pause the test and adjust the blade installation status.

[0044] 2.4 If the system detects an abnormal pulse signal (such as signal loss or excessive pulse interval), the control software will issue an alarm prompting "The speed signal is abnormal, please check the sensor or blade installation status."

[0045] 2.5 The rotational speed value is stored in the buffer after each update and is used in the next step to calculate the moment when the blade reaches the measurement position.

[0046] Step 3 Based on the latest rotational speed data and the total number of blades, the time point when the blade reaches the preset measurement position in a vibration-free state is calculated and transmitted to the trigger module that triggers the first high-speed camera.

[0047] 3.1 The control software obtains the current speed from the speed calculation unit and, based on the total number of blades, calculates the time required for the blade to rotate one full circle around the axis. It also calculates the time interval required for two adjacent blades to reach the same measurement position based on the total number of blades.

[0048] 3.2 At the beginning of the test, the initial arrival time of the first blade at the measurement location is recorded manually or automatically by the system. Each subsequent time a blade arrives at the location, the system adds the time interval required for two adjacent blades to the arrival time recorded at the previous moment to obtain the precise moment when the blade arrives at the measurement location, and uses this moment as the trigger moment for this measurement.

[0049] 3.3 After calculating the new arrival time at the measurement position, the control software automatically updates the arrival time of the previous moment to the current moment for use in the next cycle.

[0050] 3.4 The system simultaneously displays the countdown to the next measurement position on the interface so that the operator can monitor and confirm that the entire system is working as expected.

[0051] 3.5 If a large fluctuation in the rotation speed is detected during the countdown, the system will recalculate the time required for one rotation and the time interval between two adjacent blades, and re-determine the time to reach the measurement position next time.

[0052] Step 4 When the blade reaches the preset measurement position, the control software sends a trigger signal to the first high-speed camera, which starts to capture multiple consecutive frames of images and synthesizes these images into a video sequence for subsequent processing.

[0053] 4.1 When the internal clock of the control software reaches the calculated moment when the next blade reaches the measurement position, it will send a pulse signal to the external trigger input of the first high-speed camera through the trigger signal distribution unit.

[0054] 4.2 Upon receiving the trigger signal, the first high-speed camera immediately begins shooting in the preset high-speed mode, with a frame rate of no less than one thousand frames per second, and the exposure time is kept within the range allowed by the blade tip movement without obvious blur, thus ensuring that each frame image is clear.

[0055] 4.3 The camera continuously captures hundreds of frames until the preset frame limit is reached or the system automatically stops capturing when it detects the next trigger signal.

[0056] 4.4 After the acquisition is completed, the camera saves all captured image frames to the storage medium in chronological order and feeds back the saving path of the batch of image data to the control software.

[0057] 4.5 After receiving the image save path, the control software records the path together with the corresponding acquisition speed and trigger time to provide input data for subsequent video processing.

[0058] Step 5 The video captured by the first high-speed camera is processed with micro-motion amplification to enhance the micro-vibration signal of the blade circumference, generate vibration-enhanced video data, and save parameters such as the enhancement multiple and vibration frequency band used.

[0059] 5.1 After receiving the save path of the first high-speed camera, the control software calls the video processing module to load the batch of image frames and obtain basic information such as their frame rate and resolution.

[0060] 5.2 The video processing module first analyzes the contrast between the leaf and the background, for example by calculating the grayscale difference in the leaf tip region across several frames. If the leaf outline clearly contrasts with the background, the system automatically selects a phase-based amplification method; otherwise, it uses a grayscale-based amplification method, ensuring effective amplification of subtle vibrations under various imaging conditions.

[0061] 5.3 If phase-based magnification is used, the system first converts each image frame to the phase domain, applies bandpass filtering to the blade contour area, extracts the phase changes associated with blade vibration, then multiplies the extracted phase changes by a preset magnification factor, and finally reconstructs the enhanced phase back into a visible image. If a grayscale-based amplification method is used, the system will apply bandpass filtering to each frame of the image in the grayscale domain, extract the grayscale changes corresponding to the vibration frequency, multiply the grayscale changes by the preset enhancement factor, and then superimpose them back on the original image to obtain a vibration-enhanced image.

[0062] 5.4 The system processes all image frames in sequence, generates a new image sequence, and saves the sequence as new vibration-enhanced video data.

[0063] 5.5 The enhancement factor and selected vibration frequency band information (i.e., the main vibration frequency range) used in video processing will be recorded in a side file or database for subsequent use in converting pixel-level displacement into physical displacement.

[0064] 5.6 The enhanced video data and the corresponding enhancement parameters are output and passed to the next module for extracting the blade circumferential vibration displacement.

[0065] Step 6 The pixel area corresponding to the blade leading edge or visual marker point is selected from the vibration-enhanced video. The pixel movement between consecutive frames is extracted using the sparse optical flow method. The pixel movement is then converted into the actual circumferential vibration displacement by combining camera calibration and enhancement factor.

[0066] 6.1 The video processing module reads the vibration-enhanced video frame and determines the mapping relationship between pixels and actual distance based on the camera calibration results, clarifying the actual length corresponding to each pixel.

[0067] 6.2 In the first frame, edge detection algorithm or template matching method is used to automatically locate the blade leading edge contour or the marked pixel area, and its coordinates are used as the reference position for tracking.

[0068] 6.3 For subsequent frames, calculate the horizontal and vertical pixel movement of the reference position in the current frame relative to the previous frame by applying the sparse optical flow algorithm within the leaf outline or marker area before alignment.

[0069] 6.4 The pixel movement at each moment is converted into actual distance by multiplying the number of pixel movements by the actual length corresponding to each pixel. This distance is then divided by the previously recorded enhancement factor to obtain the true circumferential vibration displacement value.

[0070] 6.5 The system saves the circumferential vibration displacement values obtained at all times in chronological order as time domain data, and records its value at the corresponding timestamp to ensure that the subsequent vibration feature extraction can accurately reflect the change of vibration over time.

[0071] 6.6 If tracking failure or pixel drift occurs during optical flow tracking, the system will interpolate the previous and next frames to fill the lost displacement value and mark the abnormal moment so that the abnormal data can be removed during subsequent feature extraction.

[0072] Step 7 When the blade reaches the position that triggers the photoelectric switch, the photoelectric switch sends a trigger signal, and the second high-speed camera starts to collect multiple consecutive frames of images and saves these images as data for subsequent radial vibration extraction.

[0073] 7.1 The photoelectric switch outputs a short pulse trigger signal when the visual mark point on the blade passes by. The signal is sent to the external trigger input terminal of the second high-speed camera through the trigger distribution unit.

[0074] 7.2 After receiving the trigger signal, the second high-speed camera begins shooting in the preset high-speed mode, with a frame rate of no less than 1,000 frames per second. The exposure time is kept within the range that can clearly capture the movement of the blade tip, thereby obtaining continuous multiple frames of clear images.

[0075] 7.3 The camera stops capturing after the preset number of frames is reached or when the next trigger signal arrives, and saves the captured images in chronological order as video data for subsequent processing.

[0076] 7.4 The system records the save path of the batch of videos and saves the path together with the corresponding acquisition speed, trigger time and other information for subsequent video alignment and vibration extraction steps.

[0077] 7.5 The output of this step includes: the video data collected and saved by the second camera, and the mapping relationship between the camera pixels used for the collection and the actual length, which will be used for the subsequent conversion of pixel-level movement to physical movement.

[0078] Step 8 The video captured by the second high-speed camera is aligned between frames so that the visual marker points of the blade tip in each frame remain at the same pixel position, thereby removing the overall circumferential motion of the blade and retaining only the radial vibration component.

[0079] 8.1 The system reads the video data saved by the second camera and performs visual marker detection on the first frame of the image. It automatically determines the pixel coordinates of the leaf tip marker in the first frame as the reference position for alignment.

[0080] 8.2 For each subsequent frame, use the same detection method to automatically locate the pixel coordinates of the leaf tip marker point and calculate the horizontal and vertical pixel offsets between the pixel coordinates and the reference coordinates of the first frame.

[0081] 8.3 Shift the entire k-th frame image in the opposite direction by the corresponding pixel offset, so that after the shift, the leaf tip marker is aligned exactly with the reference coordinates of the first frame. This shift is performed in the software using image interpolation methods (such as bilinear interpolation) to ensure image quality.

[0082] 8.4 All aligned images are saved frame by frame, ultimately generating video data that has been stripped of the overall circumferential motion. At this point, the blade tip markers in the video remain in the same position in every frame, indicating that all inter-frame circumferential motion has been eliminated, leaving only radial vibration information.

[0083] 8.5 The aligned video data output from this step will be passed to the next step of the micro-motion amplification processing module, and the mapping relationship between the reference pixel coordinates of the leaf tip and the camera pixels and actual length will be output at the same time, so that the pixel displacement can be converted into physical displacement later.

[0084] Step 9 The micro-motion amplification processing is applied to the aligned video to enhance the radial micro-vibration signal of the blade, generate vibration-enhanced video data, and record parameters such as the enhancement multiple and vibration frequency band used.

[0085] 9.1 The control software reads the aligned video data and obtains information such as its frame rate and resolution for subsequent algorithm processing.

[0086] 9.2 Automatically select the appropriate amplification method based on the contrast between the blade tip and the background: if the contrast between the blade tip and the background is obvious, use the phase-based amplification method; otherwise, use the grayscale-based amplification method to ensure that the subtle vibration signal can be effectively enhanced under different lighting conditions.

[0087] 9.3 If a phase-based amplification method is used, the system will perform bandpass filtering on the blade tip region in the phase domain to extract the phase change associated with the vibration frequency. The extracted phase change is multiplied by the enhancement factor and then reconstructed back into a time domain image. If a grayscale-based amplification method is used, the system will perform grayscale domain bandpass filtering on the blade tip area image, extract the grayscale changes corresponding to the vibration frequency, multiply the grayscale changes by the enhancement factor, and superimpose them on the original image to obtain a vibration-enhanced image frame.

[0088] 9.4 The system processes all frames in sequence, generates a new vibration-enhanced video sequence, and saves the sequence to the storage medium, while recording parameters such as the enhancement multiple and vibration frequency band used.

[0089] 9.5 The vibration-enhanced video data and the corresponding enhancement parameters will be output and passed to the next module for extracting the radial vibration displacement of the blade.

[0090] Step 10 The pixel area corresponding to the blade tip area is selected from the vibration-enhanced video. The sparse optical flow method is used to calculate the pixel movement between consecutive frames. The pixel movement is then converted into the real radial vibration displacement by combining camera calibration and enhancement factor.

[0091] 10.1 The system reads the vibration-enhanced video frame and determines the mapping between pixels and actual distances based on the camera parameters obtained during calibration, clarifying the actual length corresponding to each pixel.

[0092] 10.2 In the first frame, use edge detection or template matching methods to automatically locate the pixel coordinates of the leaf tip visual marker in the image and use it as the initial reference position for optical flow tracking.

[0093] 10.3 For the kth frame, apply the sparse optical flow algorithm to track the horizontal and vertical pixel movement of the leaf tip marker in a small area around the marker in the current frame relative to the previous frame.

[0094] 10.4 The pixel movement at each moment is converted to actual length by multiplying the detected pixel movement by the actual length corresponding to each pixel. The result is then divided by the previously recorded enhancement factor to obtain the final true radial vibration displacement value.

[0095] 10.5 The system saves the radial vibration displacement values at all times in chronological order as time domain data, and records them together with the corresponding acquisition timestamps for subsequent analysis.

[0096] 10.6 If tracking fails or a marker cannot be correctly detected in a frame during optical flow tracking, the system will interpolate the data from previous and next frames to complete the missing displacement value and mark the moment as an anomaly to facilitate the removal of abnormal data during subsequent vibration feature extraction.

[0097] Step 11 The time domain and frequency domain features of the circumferential vibration displacement sequence and the radial vibration displacement sequence are extracted respectively to obtain the multi-dimensional characteristic parameters of the blade vibration. It is then judged whether the blade has abnormal vibration based on the characteristic parameters.

[0098] 11.1 The system pre-processes the extracted circumferential vibration displacement sequence and radial vibration displacement sequence separately, including removing the overall offset trend and applying band-pass filtering to ensure that only the signal components related to the blade vibration frequency band are retained.

[0099] 11.2 Time Domain Feature Extraction: Calculate peak displacement, root mean square displacement, and signal kurtosis from the preprocessed vibration displacement sequence to quantify the intensity and distribution of the vibration signal.

[0100] 11.3 Frequency Domain Feature Extraction: Apply fast Fourier transform to each vibration displacement sequence to obtain the vibration spectrum, identify the main vibration frequency and its harmonic components, and record the amplitude corresponding to each frequency to help determine whether the vibration frequency has abnormal drift.

[0101] 11.4 Organize all time domain and frequency domain characteristics into tables or structured data to facilitate system generation of measurement reports and on-site review.

[0102] 11.5 The system automatically determines whether there is a vibration anomaly based on pre-set characteristic thresholds (such as peak displacement exceeding a set value or abnormal main frequency amplitude). If an anomaly is detected, the system will mark "Vibration Anomaly" in the report and generate an alarm message to prompt the operator to perform inspection and maintenance.

[0103] 11.6 All extracted feature data and judgment results will be saved in the database or report file and displayed graphically on the control interface, including time domain waveform graphs, spectrum comparison graphs, and historical trend comparison graphs of multiple measurements, for intuitive analysis by operators.

[0104] Step 12 The circumferential and radial vibration monitoring results are integrated and compared with historical measurement data to form a multi-dimensional vibration monitoring report, and maintenance recommendations or alarm decision support are given.

[0105] 12.1 The system compares the circumferential and radial vibration characteristics obtained from this measurement with the average characteristic values at the same speed in the historical database, calculates the difference and change trend between the two, and evaluates the blade condition.

[0106] 12.2 If the measured characteristic is significantly higher than the historical average (for example, the increase in a characteristic indicator exceeds a preset percentage), the system automatically generates a maintenance recommendation, prompting the technician to check the blade installation status, the balance weight condition, or whether the visual mark points are damaged.

[0107] 12.3 The system presents monitoring data in the form of multiple charts on the control interface, including: circumferential vibration time domain curve, radial vibration time domain curve, vibration spectrum comparison chart and historical trend curve, to help operators intuitively understand the changes in blade vibration.

[0108] 12.4 Vibration monitoring reports and maintenance recommendations can be exported to PDF or spreadsheet format and uploaded to a higher-level monitoring system or cloud platform for further analysis and long-term data management by experts.

[0109] 12.5 This step outputs a complete blade vibration monitoring report and maintenance / alarm recommendations, providing a basis for subsequent routine maintenance and fault diagnosis.

[0110] Implementation Method 3: Combination Figure 1-2 This embodiment further describes the above technical solution in detail through specific examples, specifically: A method for measuring vibration of a rotating blade based on vision, comprising the following steps: Step 1: Fix the object to be measured on the stand and set up two high-speed cameras, one of which C 1's optical axis coincides with the blade drive axis, and the other C 2. The optical axis coincides with the radius of the rotation axis along the radial direction; Step 2: Install the photoelectric switch in the axial direction of the blade S , ensure that each blade can trigger the photoelectric switch when it reaches the same position, and stick different marking points on the middle of the tip of each blade; Step 3: Use the speed channel to calculate the speed of the rotating blade drive shaft in real time; Step 4: Calculate the number of blades and the real-time speed of a blade Fan i The leading edge reaches the theoretical position without vibration pos std Moment t i ; Step 5: t i Trigger the camera at all times C 1. Capture images and synthesize multiple captured images into a video V 1; Step 6: Video V 1. Execute the micro-motion amplification algorithm to enhance the vibration signal and obtain the reconstructed video in a specific frequency band. V 1'; Step 7: Select the video V Specific pixel in 1' pixel , extracted using optical flow pixel The vibration signal sequence POS of the point act ; Step 8: Calculate POS act Each signal point in pos std The difference (POS act [ k ] - pos std ), obtain leaves Fan i The actual vibration displacement sequence of the rotating blade is extracted to extract the vibration characteristics. Fan i Circumferential vibration measurement; Step 9: When the leaves Fan i Trigger photoelectric switch S , using synchronization triggers to control the camera C 2. Take pictures and combine them into a video V 2; Step 10: Align the camera with the markers on each leaf. C 2 shots of leaves Fan i Fine-tune the image to ensure that V 2, leaves Fan i The marking points of the leaves are overlapped, thereby filtering out the leaves Fan i Circumferential vibration displacement; Step 11: VideoV 2. Execute the micro-motion amplification algorithm to enhance the vibration signal and obtain the reconstructed video in a specific frequency band. V 2'; Step 12: Select the video V Multiple pixels in 2' pixels , extracted using optical flow pixels Vibration signal sequences of multiple points in the rotating blade are extracted to extract vibration features. Fan i Tip vibration shape measurement.

[0111] For the system settings in steps 1 and 2, see Figure 2 shown.

[0112] The speed measurement scheme in step three can be implemented as follows: set equally spaced grooves on the main shaft, such as 3 at 120° intervals, or 4 at 90° intervals; use eddy current sensors to align with the grooves. According to the principle of eddy current sensors, when the shaft rotates, each time a groove rotates to the position of the eddy current sensor, a signal will be generated; calculate the time between two adjacent signals, and combine the number of grooves to calculate the real-time speed.

[0113] The time required for adjacent leaves to reach the same position in the calculation method in step 4 is 1 / ( speed × n ),in speed is the rotation speed, n Theoretical position pos std This can be selected during the experimental setup, such as but not limited to the horizontal position.

[0114] For the small motion amplification methods involved in steps 6 and 11, grayscale-based methods and phase-based methods can be used. The selection of these two methods can refer to but is not limited to the following criteria: (1) When the brightness difference between the leaf edge and the background brightness in the video is large, the phase-based method can be selected; (2) When the brightness difference between the leaf edge and the background brightness in the video is small, the grayscale-based method can be selected.

[0115] For the vibration amplitude calculation in steps 7 and 12, the sparse optical flow method is generally used to calculate the pixel displacement in the horizontal and vertical directions of the image. Based on the calibration results, the pixel displacement is converted into physical displacement and divided by the "micro motion amplification" factor to obtain the actual vibration displacement.

[0116] The above further describes the technical solution provided by the present invention in detail through several specific embodiments in order to highlight the advantages and benefits of the technical solution provided by the present invention. However, the several specific embodiments described above are not intended to limit the present invention. Any reasonable modification and improvement of the present invention, combination of embodiments and equivalent replacement based on the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for measuring vibration of rotating blades based on vision, characterized in that: include: The step of calculating the time it takes for the blade to reach a preset measurement position according to the rotation speed value output by the non-contact rotation speed sensor; The step of collecting a circumferential original image of the blade when the blade reaches a preset measurement position, performing micro-motion amplification, and generating a circumferential enhanced video; Taking the current video as input, tracking the pixel movement of the blade leading edge or visual marker based on sparse optical flow, and converting the pixel movement into circumferential vibration displacement; The step of collecting a radial original image of the blade when the blade is triggered by a photoelectric switch, performing micro-motion amplification, and generating a radial enhanced video; Taking radially enhanced video as input, tracking the pixel movement of the leaf tip visual marker based on sparse optical flow, and converting the pixel movement into radial vibration displacement; The steps of performing time domain and frequency domain feature extraction on the circumferential vibration displacement and the radial vibration displacement respectively to obtain vibration characteristic parameters.

2. The method for measuring vibration of a rotating blade based on vision according to claim 1, characterized in that: The micro-motion amplification processing automatically selects the phase amplification algorithm or the grayscale amplification algorithm according to the contrast between the leaves and the background.

3. The method for measuring vibration of a rotating blade based on vision according to claim 1, wherein: The sparse optical flow tracking algorithm uses the Lucas–Kanade method to track the pixel movement of the visual marker point on the leading edge or tip of the blade.

4. The method for measuring vibration of a rotating blade based on vision according to claim 1, wherein: Inter-frame alignment eliminates the circumferential motion of the entire blade by automatically detecting and aligning the positions of the blade tip visual markers.

5. The method for measuring vibration of a rotating blade based on vision according to claim 1, characterized in that: The time domain feature extraction includes the calculation of peak displacement and root mean square displacement, and the frequency domain feature extraction includes the analysis of the main vibration frequency and its harmonic amplitude.

6. A vision-based rotating blade vibration measurement device, characterized in that: include: A module for calculating the time it takes for the blade to reach a preset measurement position based on the speed value output by the non-contact speed sensor; A module that captures the original circumferential image of the blade when it reaches the preset measurement position, amplifies the slight movement, and generates a circumferential enhanced video; A module that takes the current video as input, tracks the pixel movement of the blade leading edge or visual markers based on sparse optical flow, and converts the pixel movement into circumferential vibration displacement; A module that captures the original radial image of the blade when it is triggered by a photoelectric switch, amplifies the tiny movement, and generates a radial enhanced video; A module that takes radially enhanced video as input, tracks the pixel movement of the leaf tip visual marker based on sparse optical flow, and converts the pixel movement into radial vibration displacement; A module that extracts time-domain and frequency-domain features of circumferential and radial vibration displacements to obtain vibration characteristic parameters.

7. A vision-based rotating blade vibration measurement system, characterized in that: The method for implementing claim 1 comprises: A speed module for collecting the speed of the rotating shaft; A trigger module for calculating the blade arrival measurement moment; a circumferential processing module for driving the first high-speed camera to capture circumferential images and extract circumferential vibration displacement at a triggering moment; a photoelectric trigger module for detecting the passage of a blade tip and triggering a second high-speed camera to capture radial images; A radial processing module for performing inter-frame alignment of radial images and extracting radial vibration displacement; A feature analysis module is used to perform time domain and frequency domain analysis on circumferential and radial vibration displacements and output vibration characteristics.

8. A computer storage medium for storing a computer program, characterized in that When the computer program is read by a computer, the computer executes the method according to claim 1 .

9. A computer comprising a processor and a storage medium, characterized in that When the processor reads the computer program stored in the storage medium, the computer executes the method according to claim 1 .

10. A computer program product, being a computer program, characterized in that When the computer program is executed, the method according to claim 1 is implemented.

Images