Method and system for realizing quality analysis of wind power equipment based on deep learning
Through deep learning technology, combined with image and vibration data analysis, the problem of infrared thermal imaging technology being difficult to detect deep faults in wind power equipment is solved, achieving more efficient and accurate quality detection, and extending the service life of the equipment.
Patent Information
- Application Number
- CN202411031631.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-30
- Publication Date
- 2024-11-12
AI Technical Summary
Existing infrared thermal imaging technology is difficult to accurately detect deep faults or hidden problems in wind power equipment, resulting in limited accuracy of quality detection.
A deep learning-based method is adopted to construct a quality analysis report of wind power equipment through image preprocessing, anchor frame construction, image shear, peak signal-to-noise ratio and structural similarity calculation, combined with vibration data analysis.
It improves the accuracy and efficiency of quality inspection of wind power equipment, can more accurately analyze the external and internal components of the equipment, find potential problems, and extend the service life of the equipment.
Smart Images

Figure CN118941794A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of low-carbon optimization of bridges, and in particular to a quality analysis method and system for wind power equipment based on deep learning. Background Art
[0002] Wind power equipment refers to equipment that uses wind energy to generate electricity, usually wind generators, wind turbines and related control systems. These devices capture the kinetic energy of wind and convert it into electrical energy to supply electricity demand. During the use of wind power equipment, analyzing the quality of the equipment helps to discover potential defects or problems in the equipment, take timely measures to repair it, optimize the operating efficiency and performance of the equipment to avoid greater losses.
[0003] At present, the quality analysis method for wind power equipment is generally analyzed through infrared thermal imaging technology. The key components of the wind power equipment are comprehensively scanned by infrared thermal imaging instruments, and the thermal map is saved. Then, by analyzing the thermal map, abnormal hot spots or areas with uneven temperature distribution are identified and compared with the baseline data under normal operating conditions to distinguish between normal and abnormal conditions. However, infrared thermal imaging technology mainly reflects the temperature of the surface of the object, and may not be able to accurately detect deep faults or hidden problems, resulting in certain quality problems of wind power equipment that cannot be detected, which will affect the accuracy of quality inspection of wind power equipment. Summary of the invention
[0004] In order to solve the above problems, the present invention provides a quality analysis method for wind power equipment based on deep learning, which can improve the quality detection accuracy of wind power equipment.
[0005] In a first aspect, the present invention provides a method for implementing quality analysis of wind power equipment based on deep learning, comprising:
[0006] Collecting equipment images of wind power equipment, and performing image preprocessing on the equipment images to obtain preprocessed images;
[0007] Constructing an image anchor frame of the preprocessed image, based on the image anchor frame, performing an adaptive anchor frame on the preprocessed image to obtain an anchor frame image, and based on the anchor frame image, performing image shearing on the preprocessed image to obtain a sheared image;
[0008] Calculating a peak signal-to-noise ratio and a structural similarity of the cut image, and calculating an image hierarchy of the cut image based on the peak signal-to-noise ratio and the structural similarity, and obtaining a target image when the image hierarchy is greater than a preset hierarchy;
[0009] Performing a first quality analysis on the wind power equipment using the target image to obtain a first analysis result;
[0010] Collecting vibration data of the wind power equipment, converting the vibration data into a frequency spectrum chart, and using the frequency spectrum chart to analyze the vibration frequency, amplitude and extreme value of the wind power equipment;
[0011] A second quality analysis is performed on the wind power equipment using the vibration frequency, the amplitude, and the extreme value to obtain a second analysis result, and a quality analysis report of the wind power equipment is constructed based on the first analysis result and the second analysis result.
[0012] In a possible implementation manner of the first aspect, performing image preprocessing on the device image to obtain a preprocessed image includes:
[0013] Performing image scaling on the device image to obtain a scaled image;
[0014] Performing filtering on the scaled image to obtain a filtered image;
[0015] Perform image enhancement on the filtered image to obtain a preprocessed image.
[0016] In a possible implementation manner of the first aspect, performing an adaptive anchor frame on the preprocessed image based on the image anchor frame to obtain an anchor frame image includes:
[0017] Constructing a feature matrix of the preprocessed image;
[0018] Using the feature matrix to perform feature point detection on the preprocessed image to obtain image feature points;
[0019] The feature points of the image are located using the following formula to obtain the feature point positions:
[0020]
[0021] Among them, FS represents the feature point, T represents the image gradient value of the preprocessed image, represents the direction of the feature point of the preprocessed image, P represents the feature image, P(a+1, b), P(a-1, b), P(a, b+1), and P(a, b-1) represent the neighboring points of the origin P(a, b) of the feature image in the multi-scale space;
[0022] Calculating the gradient value of the preprocessed image using the feature points;
[0023] identifying image features of the preprocessed image based on the gradient value;
[0024] identifying image information of the preprocessed image using the image features;
[0025] An adaptive anchor frame is performed on the preprocessed image based on the image information to obtain an anchor frame image.
[0026] In a possible implementation manner of the first aspect, performing image shearing on the preprocessed image based on the anchor frame image to obtain a sheared image includes:
[0027] Based on the anchor frame image, determining a target area of the preprocessed image;
[0028] Based on the target area, construct an image cutting tool for the preprocessed image;
[0029] The image cutting tool is used to cut the pre-processed image to obtain a cut image.
[0030] In a possible implementation manner of the first aspect, calculating the peak signal-to-noise ratio and the structural similarity of the sheared image includes:
[0031] The image mean square error of the sheared image is calculated using the following formula:
[0032]
[0033] Where MSE represents the mean square error of the image, d1 represents the pixel value of the original image at position (i, j) corresponding to the sheared image, d2 represents the pixel value of the sheared image at position (i, j), H represents the height of the sheared image, and L represents the width of the sheared image;
[0034] The peak signal-to-noise ratio of the sheared image is calculated based on the mean square error and the following formula:
[0035]
[0036] Where pnsr represents the peak signal-to-noise ratio, MSE represents the mean square error of the sheared image, and X represents the maximum pixel value of the sheared image;
[0037] The structural similarity of the cut image is calculated using the following formula:
[0038]
[0039] Among them, ssim represents the structural similarity, l represents the brightness ratio of the sheared image, c represents the contrast of the sheared image, s represents the structural degree of the sheared image, ρ, σ, τ are constants, and A i represents the i-th image in the sheared image, B i represents the comparison image, and m represents the number of cropped images.
[0040] In a possible implementation manner of the first aspect, calculating the image hierarchy of the cut image based on the peak signal-to-noise ratio and the structural similarity includes:
[0041] The image hierarchy of the cut image is calculated based on the peak signal-to-noise ratio and the structural similarity using the following formula:
[0042] w=w1q pnsr +w2q ssim
[0043] Among them, w represents the image level, w1 represents the weight coefficient of the peak signal-to-noise ratio, and q pnsr represents the weight of the peak signal-to-noise ratio, q ssim Represents the weight of structural similarity.
[0044] In a possible implementation manner of the first aspect, performing a first quality analysis on the wind power equipment using the target image to obtain a first analysis result includes:
[0045] Inputting the target image into a trained deep learning model to perform random forest classification on the target image using the deep learning model to obtain a random classification result;
[0046] Calculating the contrast loss of the random classification result;
[0047] Extracting the classification results whose contrast loss satisfies the preset contrast loss to obtain the target classification results;
[0048] A first quality analysis is performed on the wind power equipment using the target classification result to obtain a first analysis result.
[0049] In a possible implementation manner of the first aspect, converting the vibration data into a frequency spectrum chart includes:
[0050] Performing signal filtering processing on the vibration data to obtain a filtered signal;
[0051] The filtered signal is transformed in the frequency domain using the following formula to obtain a transformed signal:
[0052]
[0053] Wherein, K represents the transformed signal, N represents the length of the filtered signal, x(n) represents the discrete time of the filtered signal, k represents the signal frequency of the filtered signal, t represents the discrete time of the filtered signal, and j represents the complex amplitude spectrum of the filtered signal in the frequency domain;
[0054] The transformed signal is converted into a line graph, a visualization chart of the vibration data is constructed, and the line graph is input into the visualization chart to obtain a spectrum chart.
[0055] In a possible implementation manner of the first aspect, analyzing the vibration frequency, amplitude, and extreme value of the wind power equipment by using the frequency spectrum chart includes:
[0056] Identifying the horizontal axis coordinate of the spectrum chart;
[0057] Determine the vibration frequency of the wind power equipment based on the horizontal axis coordinate, and identify the vertical axis coordinate of the frequency spectrum chart;
[0058] The amplitude of the wind power device is determined based on the vertical axis coordinate, the peaks and troughs of the line graph in the frequency spectrum diagram are identified, and the extreme value of the wind power device is determined based on the peaks and the troughs.
[0059] In a second aspect, the present invention provides a quality analysis system for wind power equipment based on deep learning, the system comprising:
[0060] An image processing module is used to collect equipment images of wind power equipment and perform image preprocessing on the equipment images to obtain preprocessed images;
[0061] An image shearing module is used to construct an image anchor frame of the preprocessed image, perform an adaptive anchor frame on the preprocessed image based on the image anchor frame to obtain an anchor frame image, and perform image shearing on the preprocessed image based on the anchor frame image to obtain a sheared image;
[0062] An image analysis module, used to calculate a peak signal-to-noise ratio and a structural similarity of the cut image, and based on the peak signal-to-noise ratio and the structural similarity, calculate an image hierarchy of the cut image, and obtain a target image when the image hierarchy is greater than a preset hierarchy;
[0063] A first analysis module, configured to perform a first quality analysis on the wind power equipment using the target image to obtain a first analysis result;
[0064] A data acquisition module, used for acquiring vibration data of the wind power equipment, converting the vibration data into a frequency spectrum chart, and using the frequency spectrum chart to analyze the vibration frequency, amplitude and extreme value of the wind power equipment;
[0065] The second analysis module is used to perform a second quality analysis on the wind power equipment using the vibration frequency, the amplitude and the extreme value to obtain a second analysis result, and to construct a quality analysis report of the wind power equipment based on the first analysis result and the second analysis result.
[0066] Compared with the prior art, the technical principle and beneficial effects of this solution are:
[0067] The embodiment of the present invention can obtain the external image of the wind power equipment by collecting the equipment image of the wind power equipment, so as to analyze the external situation of the wind power equipment by using the external image, and construct the image anchor frame of the pre-processed image to frame the target object to be analyzed in an image. Furthermore, the embodiment of the present invention can cut out a large amount of irrelevant information in an image by performing image shearing on the pre-processed image based on the anchor frame image to obtain the sheared image, so as to reduce interference factors when the model is used to analyze the image later, thereby improving the efficiency and accuracy of image analysis, and calculating the peak signal-to-noise ratio and structural similarity of the sheared image can understand the results of the analysis. After processing each step, the processing level of the sheared image is obtained, such as image loss, clarity, etc., to improve the accuracy of the analysis of the sheared image. Furthermore, the embodiment of the present invention uses the target image to perform a first quality analysis on the wind power equipment, and obtains a first analysis result to understand the external situation of the wind power equipment, and then analyzes the external quality of the wind power equipment such as the blades, towers and bearings of the wind turbine, and uses the vibration frequency, the amplitude and the extreme value to perform a second quality analysis on the wind power equipment. The second analysis result can understand the internal components of the wind power equipment when it is running, such as motors, gears and other components. A method and system for quality analysis of wind power equipment based on deep learning proposed in an embodiment of the present invention can improve the accuracy of quality detection of wind power equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0068] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0069] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0070] Figure 1 A schematic flow chart of a method for implementing quality analysis of wind power equipment based on deep learning provided by an embodiment of the present invention;
[0071] Figure 2 A module schematic diagram of a quality analysis system for wind power equipment based on deep learning provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0072] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0073] An embodiment of the present invention provides a method for quality analysis of wind power equipment based on deep learning, and the execution subject of the method for quality analysis of wind power equipment based on deep learning includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present invention. In other words, the method for quality analysis of wind power equipment based on deep learning can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.
[0074] See also Figure 1 FIG. 1 is a flow chart of a method for implementing a quality analysis method of wind power equipment based on deep learning provided by an embodiment of the present invention. Figure 1 A method for quality analysis of wind power equipment based on deep learning described in the literature includes:
[0075] S1. Collecting equipment images of wind power equipment, and performing image preprocessing on the equipment images to obtain preprocessed images.
[0076] In the embodiment of the present invention, the external image of the wind power equipment can be obtained by collecting the equipment image of the wind power equipment, so as to analyze the external condition of the wind power equipment by using the external image. The equipment image can be collected by a pan-tilt camera.
[0077] Furthermore, the embodiment of the present invention can improve the clarity of the image by performing image preprocessing on the device image to obtain a preprocessed image, thereby enabling better analysis of the information in the image. The preprocessing refers to a processing method before analyzing the image, such as filtering, graying, and standardization.
[0078] As an embodiment of the present invention, the image preprocessing of the device image to obtain the preprocessed image includes: performing image scaling on the device image to obtain a scaled image, filtering the scaled image to obtain a filtered image, and image enhancing the filtered image to obtain a preprocessed image.
[0079] Optionally, performing image scaling on the device image to obtain a scaled image refers to adjusting the size of the image, which can be achieved through a nearest neighbor interpolation method; performing filtering on the scaled image to obtain a filtered image refers to a technique for removing image noise, which can be achieved through a low-pass filtering technique; and performing image enhancement on the filtered image refers to enhancing image details to improve recognition capabilities, which can be achieved through a deep learning model.
[0080] S2. Construct an image anchor frame of the preprocessed image, perform an adaptive anchor frame on the preprocessed image based on the image anchor frame to obtain an anchor frame image, and perform image shearing on the preprocessed image based on the anchor frame image to obtain a sheared image.
[0081] In the embodiment of the present invention, the target object to be analyzed can be framed in an image by constructing the image anchor frame of the preprocessed image, wherein the image anchor frame refers to a technology in the field of target detection, which is used to accurately locate the target in the image.
[0082] Optionally, the image anchor frame can be constructed by a deep learning model.
[0083] Furthermore, the embodiment of the present invention performs an adaptive anchor frame on the preprocessed image based on the image anchor frame, and obtains an anchor frame image, which can extract the target object to be analyzed in the image, avoid interference from excessive information, and improve the efficiency of image analysis.
[0084] As an embodiment of the present invention, based on the image anchor frame, the preprocessed image is adaptively anchored to obtain the anchor frame image, including: constructing a feature matrix of the preprocessed image, using the feature matrix to detect feature points of the preprocessed image to obtain image feature points, and locating the image feature points using the following formula to obtain feature point positions:
[0085]
[0086] Among them, FS represents the feature point, T represents the image gradient value of the preprocessed image, represents the direction of the feature points of the preprocessed image, P represents the feature image, P(a+1, b), P(a-1, b), P(a, b+1), and P(a, b-1) represent the neighboring points of the origin P(a, b) of the feature image in the multi-scale space.
[0087] The feature points are used to calculate the gradient value of the preprocessed image, the image features of the preprocessed image are identified based on the gradient values, the image information of the preprocessed image is identified using the image features, and an adaptive anchor frame is performed on the preprocessed image based on the image information to obtain an anchor frame image.
[0088] Among them, the feature matrix refers to a concept commonly used in machine learning and data analysis. It is a matrix composed of samples in the data set and their corresponding features, which can be constructed through feature engineering technology. The feature point refers to the location in the image with unique, significant or important information, and the gradient value refers to the rate of change of grayscale intensity at each pixel in the image.
[0089] Optionally, the feature matrix is used to perform feature point detection on the preprocessed image to obtain image feature points by constructing a multi-scale space of the preprocessed image, and the multi-scale space is used to perform feature point detection. The feature point position is used to calculate the gradient value of the preprocessed image by calculating the gradient function. The image features of the preprocessed image based on the gradient values are identified by constructing a gradient histogram of the preprocessed image using the gradient values, and the gradient histogram is used to perform feature extraction. The image information of the preprocessed image is identified by using the image features through a preconfigured neural network to identify the image features, and the image features are translated. The preprocessed image is adaptively anchored based on the image information to obtain the anchor frame image by determining the target area in the preprocessed image using the image information, and adaptively anchoring the target area.
[0090] Furthermore, the embodiment of the present invention performs image shearing on the preprocessed image based on the anchor frame image to obtain a sheared image, which can cut out a large amount of irrelevant information in an image, thereby reducing interference factors when subsequently using the model to analyze the image, thereby improving the efficiency and accuracy of image analysis.
[0091] As an embodiment of the present invention, the method of performing image shearing on the preprocessed image based on the anchor frame image to obtain a sheared image includes: determining a target area of the preprocessed image based on the anchor frame image, constructing an image shearing tool for the preprocessed image based on the target area, and performing image shearing on the preprocessed image using the image shearing tool to obtain a sheared image.
[0092] Optionally, the image cutting tool can be constructed by a script generated by Java.
[0093] S3, calculating the peak signal-to-noise ratio and the structural similarity of the cut image, and calculating the image hierarchy of the cut image based on the peak signal-to-noise ratio and the structural similarity, and obtaining the target image when the image hierarchy is greater than a preset hierarchy.
[0094] The embodiment of the present invention can understand the processing level of the sheared image obtained after each step of processing, such as image loss, clarity, etc., by calculating the peak signal-to-noise ratio and structural similarity of the sheared image, so as to improve the accuracy of the sheared image analysis. It should be understood that the higher the processing level, the more accurate the result of the quality analysis of the wind power equipment, so it is important to ensure the image processing level. Among them, the peak signal-to-noise ratio refers to the distortion of the image, and the structural similarity refers to the degree of image details retained by the processed image compared with the original image.
[0095] As an embodiment of the present invention, the calculating the peak signal-to-noise ratio and the structural similarity of the sheared image includes: calculating the image mean square error of the sheared image using the following formula:
[0096]
[0097] Where MSE represents the mean square error of the image, d1 represents the pixel value of the original image at position (i, j) corresponding to the sheared image, d2 represents the pixel value of the sheared image at position (i, j), H represents the height of the sheared image, and L represents the width of the sheared image;
[0098] The peak signal-to-noise ratio of the sheared image is calculated based on the mean square error and the following formula:
[0099]
[0100] Where pnsr represents the peak signal-to-noise ratio, MSE represents the mean square error of the sheared image, and X represents the maximum pixel value of the sheared image;
[0101] The structural similarity of the cut image is calculated using the following formula:
[0102]
[0103] Among them, ssim represents the structural similarity, l represents the brightness ratio of the sheared image, c represents the contrast of the sheared image, s represents the structural degree of the sheared image, ρ, σ, τ are constants, and A i represents the i-th image in the sheared image, B i represents the comparison image, and m represents the number of cropped images.
[0104] In the embodiment of the present invention, the overall processing level of the cut-out image can be understood by calculating the image hierarchy of the cut-out image based on the peak signal-to-noise ratio and the structural similarity.
[0105] As an embodiment of the present invention, the calculating the image hierarchy of the cut image based on the peak signal-to-noise ratio and the structural similarity includes: calculating the image hierarchy of the cut image based on the peak signal-to-noise ratio and the structural similarity using the following formula:
[0106] w=w1q pnsr +w2q ssim
[0107] Among them, w represents the image level, w1 represents the weight coefficient of the peak signal-to-noise ratio, and q pnsr represents the weight of the peak signal-to-noise ratio, q ssim Represents the weight of structural similarity.
[0108] It should be understood that when the image hierarchy is greater than the preset hierarchy, obtaining the target image means that the calculated value w of the image hierarchy is greater than the preset value, indicating that the processing level of the cut image meets the requirements. The preset value can be 0.8, or it can be set according to the actual application scenario.
[0109] S4. Perform a first quality analysis on the wind power equipment using the target image to obtain a first analysis result.
[0110] The embodiment of the present invention uses the target image to perform a first quality analysis on the wind power equipment, and obtains a first analysis result to understand the external condition of the wind power equipment, and then analyze the external quality of the wind power equipment such as the blades, towers and bearings of the wind turbine.
[0111] As an embodiment of the present invention, the method of performing a first quality analysis on the wind power equipment using the target image to obtain a first analysis result includes: inputting the target image into a trained deep learning model to perform random forest classification on the target image using the deep learning model to obtain a random classification result, and calculating the contrast loss of the random classification result, extracting the classification result whose contrast loss satisfies a preset contrast loss to obtain a target classification result, and performing a first quality analysis on the wind power equipment using the target classification result to obtain a first analysis result.
[0112] Optionally, the using the deep learning model to perform random forest classification on the target image to obtain a random classification result means performing feature classification on the target image by utilizing historical image features of the target image, and the obtained classification result can be obtained by a random forest function, and the calculating of the contrast loss of the random classification result means performing image detail comparison between the target image and an image that matches the target image classification, and determining the final classification result based on the contrast ratio of the image details, which can be achieved by a contrast loss function.
[0113] S5. Collect vibration data of the wind power equipment, convert the vibration data into a frequency spectrum chart, and use the frequency spectrum chart to analyze the vibration frequency, amplitude, and extreme value of the wind power equipment.
[0114] The embodiment of the present invention can understand the operation status and health status of the equipment in real time by collecting the vibration data of the wind power equipment. The vibration data refers to the motion signal generated by the equipment during operation, which can be collected by a displacement sensor.
[0115] Furthermore, the embodiment of the present invention can display complex spectrum data in an intuitive manner by converting the vibration data into a spectrum chart, thereby enabling better understanding and analysis of the data. The spectrum chart refers to a chart used to represent the energy distribution of a signal at different frequencies.
[0116] As an embodiment of the present invention, the converting the vibration data into a frequency spectrum chart includes: performing signal filtering processing on the vibration data to obtain a filtered signal, and performing frequency domain transformation on the filtered signal using the following formula to obtain a transformed signal:
[0117]
[0118] Wherein, K represents the transformed signal, N represents the length of the filtered signal, x(n) represents the discrete time of the filtered signal, k represents the signal frequency of the filtered signal, t represents the discrete time of the filtered signal, and j represents the complex amplitude spectrum of the filtered signal in the frequency domain;
[0119] The transformed signal is converted into a line graph, a visualization chart of the vibration data is constructed, and the line graph is input into the visualization chart to obtain a spectrum chart.
[0120] The visual chart refers to a graphical representation method used to display data, information or relationships, including types such as line, column and scatter points.
[0121] Optionally, the visualization chart can be constructed by a script generated by Java.
[0122] The embodiment of the present invention can evaluate the performance status of the wind power equipment by analyzing the vibration frequency, amplitude and extreme value of the wind power equipment using the frequency spectrum chart. Abnormal vibration frequency or amplitude may indicate that the equipment has problems or needs maintenance. The vibration frequency refers to the number of vibration cycles completed by an object in a unit time, the amplitude refers to the maximum deviation or displacement of a wave or vibration, that is, the distance from the peak (positive peak) or trough (negative peak) to the equilibrium position of the wave, and the extreme value refers to the maximum and minimum amplitude values reached during the vibration process.
[0123] As an embodiment of the present invention, the use of the frequency spectrum chart to analyze the vibration frequency, amplitude and extreme values of the wind power equipment includes: identifying the horizontal axis coordinates of the frequency spectrum chart, determining the vibration frequency of the wind power equipment based on the horizontal axis coordinates, identifying the vertical axis coordinates of the frequency spectrum chart, determining the amplitude of the wind power equipment based on the vertical axis coordinates, identifying the peaks and troughs of the line graph in the frequency spectrum chart, and determining the extreme values of the wind power equipment based on the peaks and the troughs.
[0124] S6. Perform a second quality analysis on the wind power equipment using the vibration frequency, the amplitude, and the extreme value to obtain a second analysis result, and construct a quality analysis report on the wind power equipment based on the first analysis result and the second analysis result.
[0125] The embodiment of the present invention performs a second quality analysis on the wind power equipment by utilizing the vibration frequency, the amplitude and the extreme value, and obtains a second analysis result to understand the internal components of the wind power equipment such as motors, gears and other parts when the wind power equipment is running.
[0126] Furthermore, the embodiment of the present invention can intuitively understand the detailed data about the quality analysis of the wind power equipment by constructing the quality analysis report of the wind power equipment based on the first analysis result and the second analysis result, so as to better analyze whether there are quality problems or faults in the wind power equipment, thereby facilitating fault repair or equipment maintenance of the wind power equipment.
[0127] Optionally, the quality analysis report can be obtained by integrating the first analysis result and the second analysis result using an Excel tool.
[0128] like Figure 2 As shown, it is a functional module diagram of a quality analysis system for wind power equipment based on deep learning according to the present invention.
[0129] The quality analysis system 200 for wind power equipment based on deep learning described in the present invention can be installed in an electronic device. According to the functions implemented, the quality analysis system for wind power equipment based on deep learning can include an image processing module 201, an image cutting module 202, an image analysis module 203, a first analysis module 204, a data acquisition module 205, and a second analysis module 206. The module described in the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.
[0130] In the embodiment of the present invention, the functions of each module / unit are as follows:
[0131] The image processing module 201 is used to collect equipment images of wind power equipment and perform image preprocessing on the equipment images to obtain preprocessed images;
[0132] The image cutting module 202 is used as an image cutting module, and is used to construct an image anchor frame of the preprocessed image, and based on the image anchor frame, perform an adaptive anchor frame on the preprocessed image to obtain an anchor frame image, and based on the anchor frame image, perform image cutting on the preprocessed image to obtain a cut image; the image analysis module 203 is used to calculate the peak signal-to-noise ratio and structural similarity of the cut image, and calculate the image hierarchy of the cut image based on the peak signal-to-noise ratio and the structural similarity, and obtain the target image when the image hierarchy is greater than a preset hierarchy;
[0133] The first analysis module 204 is a first analysis module configured to perform a first quality analysis on the wind power equipment using the target image to obtain a first analysis result;
[0134] The data acquisition module 205 is used to collect vibration data of the wind power equipment, convert the vibration data into a frequency spectrum chart, and use the frequency spectrum chart to analyze the vibration frequency, amplitude and extreme value of the wind power equipment;
[0135] The second analysis module 206 is used to perform a second quality analysis on the wind power equipment using the vibration frequency, the amplitude and the extreme value to obtain a second analysis result, and to construct a quality analysis report of the wind power equipment based on the first analysis result and the second analysis result.
[0136] In detail, each module in the wind power equipment quality analysis system 200 based on deep learning described in the embodiment of the present invention is used in the same manner as described above. Figure 1 The technical means described in the article are similar to the method for quality analysis of wind power equipment based on deep learning and can produce the same technical effects, so they will not be repeated here.
[0137] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.
[0138] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0139] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.
[0140] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.
[0141] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.
[0142] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0143] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.
Claims
1. A method for quality analysis of wind power equipment based on deep learning, characterized in that: The method comprises: Collecting equipment images of wind power equipment, and performing image preprocessing on the equipment images to obtain preprocessed images; Constructing an image anchor frame of the preprocessed image, based on the image anchor frame, performing an adaptive anchor frame on the preprocessed image to obtain an anchor frame image, and based on the anchor frame image, performing image shearing on the preprocessed image to obtain a sheared image; Calculating a peak signal-to-noise ratio and a structural similarity of the cut image, and calculating an image hierarchy of the cut image based on the peak signal-to-noise ratio and the structural similarity, and obtaining a target image when the image hierarchy is greater than a preset hierarchy; Performing a first quality analysis on the wind power equipment using the target image to obtain a first analysis result; Collecting vibration data of the wind power equipment, converting the vibration data into a frequency spectrum chart, and using the frequency spectrum chart to analyze the vibration frequency, amplitude and extreme value of the wind power equipment; A second quality analysis is performed on the wind power equipment using the vibration frequency, the amplitude, and the extreme value to obtain a second analysis result, and a quality analysis report of the wind power equipment is constructed based on the first analysis result and the second analysis result.
2. The method according to claim 1, characterized in that The performing image preprocessing on the device image to obtain a preprocessed image includes: Performing image scaling on the device image to obtain a scaled image; Performing filtering on the scaled image to obtain a filtered image; Perform image enhancement on the filtered image to obtain a preprocessed image.
3. The method according to claim 1, characterized in that The step of performing an adaptive anchor frame operation on the preprocessed image based on the image anchor frame to obtain an anchor frame image includes: Constructing a feature matrix of the preprocessed image; Using the feature matrix to perform feature point detection on the preprocessed image to obtain image feature points; The feature points of the image are located using the following formula to obtain the feature point positions: Where FS represents the feature point, T represents the image gradient value of the preprocessed image, θ represents the direction of the feature point of the preprocessed image, P represents the feature image, P(a+1, b), P(a-1, b), P(a, b+1), and P(a, b-1) represent the neighboring points of the origin P(a, b) of the feature image in the multi-scale space; Calculating the gradient value of the preprocessed image using the feature points; identifying image features of the preprocessed image based on the gradient value; identifying image information of the preprocessed image using the image features; An adaptive anchor frame is performed on the preprocessed image based on the image information to obtain an anchor frame image.
4. The method according to claim 1, characterized in that: The step of performing image shearing on the preprocessed image based on the anchor frame image to obtain a sheared image comprises: Based on the anchor frame image, determining a target area of the preprocessed image; Based on the target area, construct an image cutting tool for the preprocessed image; The image cutting tool is used to cut the pre-processed image to obtain a cut image.
5. The method according to claim 1, characterized in that: The calculating the peak signal-to-noise ratio and the structural similarity of the sheared image includes: The image mean square error of the sheared image is calculated using the following formula: Where MSE represents the mean square error of the image, d1 represents the pixel value of the original image at position (i, j) corresponding to the sheared image, d2 represents the pixel value of the sheared image at position (i, j), H represents the height of the sheared image, and L represents the width of the sheared image; The peak signal-to-noise ratio of the sheared image is calculated based on the mean square error and the following formula: Where pnsr represents the peak signal-to-noise ratio, MSE represents the mean square error of the sheared image, and X represents the maximum pixel value of the sheared image; The structural similarity of the cut image is calculated using the following formula: Among them, ssim represents the structural similarity, l represents the brightness ratio of the sheared image, c represents the contrast of the sheared image, s represents the structural degree of the sheared image, ρ, σ, τ are constants, and A i represents the i-th image in the sheared image, B i represents the comparison image, and m represents the number of cropped images.
6. The method according to claim 1, characterized in that The calculating the image hierarchy of the cut image based on the peak signal-to-noise ratio and the structural similarity includes: The image hierarchy of the cut image is calculated based on the peak signal-to-noise ratio and the structural similarity using the following formula: w=w1q pnsr +w2q ssim Among them, w represents the image level, w1 represents the weight coefficient of the peak signal-to-noise ratio, and q pnsr represents the weight of the peak signal-to-noise ratio, q ssim Represents the weight of structural similarity.
7. The method according to claim 1, characterized in that The step of performing a first quality analysis on the wind power equipment using the target image to obtain a first analysis result includes: Inputting the target image into a trained deep learning model to perform random forest classification on the target image using the deep learning model to obtain a random classification result; Calculating the contrast loss of the random classification result; Extracting the classification results whose contrast loss satisfies the preset contrast loss to obtain the target classification results; A first quality analysis is performed on the wind power equipment using the target classification result to obtain a first analysis result.
8. The method according to claim 1, characterized in that The step of converting the vibration data into a frequency spectrum chart comprises: Performing signal filtering processing on the vibration data to obtain a filtered signal; The filtered signal is transformed in the frequency domain using the following formula to obtain a transformed signal: Wherein, K represents the transformed signal, N represents the length of the filtered signal, x(n) represents the discrete time of the filtered signal, k represents the signal frequency of the filtered signal, t represents the discrete time of the filtered signal, and j represents the complex amplitude spectrum of the filtered signal in the frequency domain; The transformed signal is converted into a line graph, a visualization chart of the vibration data is constructed, and the line graph is input into the visualization chart to obtain a spectrum chart.
9. The method according to claim 1, characterized in that: The using the frequency spectrum chart to analyze the vibration frequency, amplitude and extreme value of the wind power equipment includes: Identifying the horizontal axis coordinate of the spectrum chart; Determine the vibration frequency of the wind power equipment based on the horizontal axis coordinate, and identify the vertical axis coordinate of the frequency spectrum chart; The amplitude of the wind power device is determined based on the vertical axis coordinate, the peaks and troughs of the line graph in the frequency spectrum diagram are identified, and the extreme value of the wind power device is determined based on the peaks and the troughs.
10. A quality analysis system for wind power equipment based on deep learning, characterized in that: The system comprises: An image processing module is used to collect equipment images of wind power equipment and perform image preprocessing on the equipment images to obtain preprocessed images; An image shearing module is used to construct an image anchor frame of the preprocessed image, perform an adaptive anchor frame on the preprocessed image based on the image anchor frame to obtain an anchor frame image, and perform image shearing on the preprocessed image based on the anchor frame image to obtain a sheared image; An image analysis module, used to calculate a peak signal-to-noise ratio and a structural similarity of the cut image, and based on the peak signal-to-noise ratio and the structural similarity, calculate an image hierarchy of the cut image, and obtain a target image when the image hierarchy is greater than a preset hierarchy; A first analysis module, configured to perform a first quality analysis on the wind power equipment using the target image to obtain a first analysis result; A data acquisition module, used for acquiring vibration data of the wind power equipment, converting the vibration data into a frequency spectrum chart, and using the frequency spectrum chart to analyze the vibration frequency, amplitude and extreme value of the wind power equipment; The second analysis module is used to perform a second quality analysis on the wind power equipment using the vibration frequency, the amplitude and the extreme value to obtain a second analysis result, and to construct a quality analysis report of the wind power equipment based on the first analysis result and the second analysis result.
