A hydraulic plunger pump intelligent fault diagnosis method and system based on pressure signals

Abstract

The application relates to the technical field of image recognition, and discloses a hydraulic plunger pump intelligent fault diagnosis method and system based on pressure signals, which comprises the following steps: arranging pressure signals of a hydraulic plunger pump according to a spiral structure, and performing interpolation processing on the arranged signals to obtain an optimized pressure image; performing visual enhancement processing on the optimized pressure image to obtain an enhanced pressure image; analyzing multi-scale features in the enhanced pressure image to obtain global feature parameters of image brightness statistical features; performing local texture feature analysis on the enhanced pressure image to obtain local feature parameters of image gradient changes; taking the global feature parameters as an upper layer framework and taking the local feature parameters as a bottom layer benchmark to construct overall feature parameters of a running state; comparing the overall feature parameters with preset standard feature threshold values in multiple levels, and determining a fault state according to a comparison result; and the application can improve the efficiency of the hydraulic plunger pump intelligent fault diagnosis based on the pressure signals.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to an intelligent fault diagnosis method and system for hydraulic piston pumps based on pressure signals. Background Technology

[0002] In the field of hydraulic piston pump fault diagnosis, pressure signal-based diagnostic methods rely heavily on the effective conversion between signals and images. However, existing technologies lack a reasonable structural design for the arrangement of pressure signals and fail to fully correlate the temporal characteristics and spatial distribution logic of the signals. This results in initial images after conversion often exhibiting problems such as blurred feature representations and missing information. From the perspective of related image processing technologies, existing interpolation methods are insufficient to accurately compensate for pixel deficiencies in blank areas of the image, making it impossible to construct optimized images that truly reflect the essential laws of pressure signals. This deprives subsequent visual analysis of reliable data support, directly impacting the overall efficiency of the diagnostic process.

[0003] Existing technologies have significant shortcomings in feature extraction and fault diagnosis. Feature analysis of pressure images is often limited to a single dimension, failing to achieve deep fusion of multi-scale global features and local texture features. This results in insufficient completeness and weak correlation of the acquired feature parameters, making it difficult to comprehensively characterize the operating status of hydraulic piston pumps. Furthermore, the fault diagnosis process lacks a scientific multi-level threshold comparison and cross-validation mechanism. Simple feature matching alone is insufficient to distinguish fault differences under complex operating conditions, leading to low accuracy and reliability of diagnostic results, and failing to meet the practical application requirements for efficient and accurate fault diagnosis of hydraulic piston pumps. Summary of the Invention

[0004] This invention provides a method and system for intelligent fault diagnosis of hydraulic piston pumps based on pressure signals, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an intelligent fault diagnosis method for hydraulic piston pumps based on pressure signals, comprising:

[0006] S1. Arrange the pressure signals of the hydraulic piston pump according to the spiral structure, and perform interpolation processing on the arranged signals to obtain the optimized pressure image of the hydraulic piston pump.

[0007] S2. Perform visual enhancement processing on the optimized pressure image to obtain the enhanced pressure image of the hydraulic piston pump;

[0008] S3. Analyze the multi-scale features in the enhanced pressure image to obtain the global feature parameters of the image brightness statistical features in the hydraulic piston pump;

[0009] S4. Perform local texture feature analysis on the enhanced pressure image to obtain local feature parameters of the image gradient change in the hydraulic plunger pump;

[0010] S5. Using the global feature parameters as the upper-level framework and the local feature parameters as the lower-level reference, construct the overall feature parameters of the operating state of the hydraulic piston pump.

[0011] S6. Perform multi-level comparison between the overall feature parameters and the preset standard feature threshold, and determine the fault status of the hydraulic piston pump based on the comparison results.

[0012] In a preferred embodiment, arranging the pressure signals of the hydraulic plunger pump according to a helical structure and interpolating the arranged signals to obtain an optimized pressure image of the hydraulic plunger pump includes:

[0013] The pressure signal of the hydraulic piston pump is collected during operation.

[0014] The starting point of the spiral structure is taken as the center of the image plane, and the unfolding direction of the spiral path in the spiral structure is determined by the equal angle increment of the spiral structure.

[0015] The pressure signals are distributed to equal-angle positions along the spiral path in chronological order to obtain a spiral distribution of the pressure signals.

[0016] In the spiral distribution, the pressure signal corresponding to the equiangular position is mapped to the image grayscale value to obtain the initial image of the spiral grayscale distribution in the hydraulic plunger pump;

[0017] Identify blank areas in the initial image caused by path gaps in the spiral path;

[0018] Fill in the missing pixel values ​​in the blank areas of the image to obtain an optimized pressure image of the hydraulic piston pump.

[0019] In a preferred embodiment, the step of mapping the pressure signals corresponding to the equiangular positions in the spiral distribution to image grayscale values ​​to obtain an initial image of the spiral grayscale distribution in the hydraulic piston pump includes:

[0020] Using the time-domain characteristics of the pressure signal and the combined characterization of the position and amplitude characteristics in the spiral distribution, a pressure signal characterization space for the spiral distribution is constructed.

[0021] In the pressure signal characterization space, the pressure signals at the isoangular positions are mapped onto a preset grayscale base to obtain grayscale characterization values ​​with visual discrimination at the isoangular positions.

[0022] According to the topological structure of the spiral path, the grayscale representation value is projected onto the image coordinate position of the image plane;

[0023] Along the spiral path, the grayscale representation value is reconstructed using the visual saliency distribution of the pressure signal to obtain an initial image of the spiral grayscale distribution.

[0024] In a preferred embodiment, the step of performing visual enhancement processing on the optimized pressure image to obtain the enhanced pressure image of the hydraulic piston pump includes:

[0025] By analyzing the visual characteristics of the optimized stress image, structural feature information of the optimized stress image at different scales is obtained.

[0026] The structural feature information is reconstructed to obtain the structure-aware enhanced map of the optimized pressure image;

[0027] Based on the structure-aware enhanced map, adaptive contrast enhancement is performed on the optimized pressure image to obtain the visual mode of the operating status related to the hydraulic piston pump in the optimized pressure image.

[0028] The visual mode of the running state is subjected to detail restoration processing to obtain the essential feature texture information of the pressure signal in the optimized pressure image;

[0029] The essential feature texture information is input into the optimized pressure image to obtain the enhanced pressure image of the hydraulic plunger pump.

[0030] In a preferred embodiment, the step of analyzing the multi-scale features in the enhanced pressure image to obtain global feature parameters of the image brightness statistical features in the hydraulic plunger pump includes:

[0031] Using visual information at different scales in the enhanced stress image, an image pyramid structure of the enhanced stress image is constructed;

[0032] In the image pyramid structure, the brightness statistics of the brightness distribution characteristics in the enhanced pressure image are extracted;

[0033] Cross-scale correlation analysis is performed on the brightness statistics to obtain the correlation matrix of the enhanced stress image;

[0034] Based on the correlation matrix, tensor synthesis is performed on the overall operating state of the hydraulic piston pump to obtain the brightness feature vector of the hydraulic piston pump.

[0035] By normalizing the different dimensions of the brightness feature vector, the global feature parameters of the hydraulic plunger pump are obtained.

[0036] In a preferred embodiment, the step of performing local texture feature analysis on the enhanced pressure image to obtain local feature parameters of the image gradient changes in the hydraulic plunger pump includes:

[0037] The local texture features in the enhanced stress image are scanned region by region to obtain the spatial variation pattern of the enhanced stress image;

[0038] Based on the changing trend of the spatial variation pattern, the directional gradient distribution of texture orientation characteristics in the enhanced stress image is determined;

[0039] Regional aggregation of gradient change descriptors in the directional gradient distribution yields a texture primitive feature map of the enhanced stress image.

[0040] The gradient statistics in the texture primitive feature map are used as local feature parameters of the enhanced stress image.

[0041] In a preferred embodiment, constructing the overall characteristic parameters of the operating state of the hydraulic piston pump, using the global characteristic parameters as the upper-level framework and the local characteristic parameters as the lower-level reference, includes:

[0042] The global feature parameters are configured as the top-level representation of the hydraulic piston pump, and the local feature parameters are configured as the bottom-level representation of the hydraulic piston pump.

[0043] The spatial distribution of the bottom-level representation and the feature dimensions of the top-level representation are synchronously aligned to obtain the preliminary feature relationship of the hydraulic piston pump.

[0044] Verify the feature integrity in the preliminary feature relationship to obtain the feature association relationship of the hydraulic piston pump;

[0045] Based on the aforementioned feature association, the local feature parameters are embedded as nodes into the overall framework of the global feature parameters to obtain the overall feature parameters of the operating state of the hydraulic piston pump.

[0046] In a preferred embodiment, the calculation formula for the preliminary feature relationship is as follows: ;

[0047] In the formula, For the first The first underlying representation and the first The initial relationship strength between the top-level representations The local feature parameter is the first one. The spatial location of the first 1 eigenvalue, The first of the global feature parameters 1 eigenvalue, It is a preset small positive number.

[0048] In a preferred embodiment, the step of performing a multi-level comparison of the overall feature parameters with a preset standard feature threshold, and determining the fault state of the hydraulic piston pump based on the comparison results, includes:

[0049] The preset standard feature thresholds are divided into macro-consistency thresholds and micro-difference thresholds according to feature level;

[0050] The global feature components in the overall feature parameters are compared with the macro-consistency threshold at the first level to obtain the consistency evaluation result of the global feature components.

[0051] The local feature components in the overall feature parameters are compared with the micro-difference threshold at a second level to obtain the difference evaluation result of the local feature components;

[0052] Cross-validate the consistency assessment results and the difference assessment results to obtain the comprehensive diagnostic decision vector of the overall feature parameters;

[0053] Based on the pointing characteristics of the comprehensive diagnostic decision vector, the fault state type of the hydraulic piston pump is determined.

[0054] To address the aforementioned problems, the present invention also provides an intelligent fault diagnosis system for hydraulic piston pumps based on pressure signals, the system comprising:

[0055] An image construction module is used to arrange the pressure signals of the hydraulic piston pump in a spiral structure and to perform interpolation processing on the arranged signals to obtain an optimized pressure image of the hydraulic piston pump.

[0056] An image enhancement module is used to perform visual enhancement processing on the optimized pressure image to obtain an enhanced pressure image of the hydraulic plunger pump;

[0057] The global feature module is used to analyze the multi-scale features in the enhanced pressure image to obtain the global feature parameters of the image brightness statistical features in the hydraulic piston pump.

[0058] The local feature module is used to perform local texture feature analysis on the enhanced pressure image to obtain local feature parameters of the image gradient change in the hydraulic plunger pump;

[0059] The overall feature module is used to construct the overall feature parameters of the operating state of the hydraulic piston pump, with the global feature parameters as the upper-level framework and the local feature parameters as the lower-level reference.

[0060] The fault diagnosis module is used to perform multi-level comparisons between the overall feature parameters and preset standard feature thresholds, and determine the fault status of the hydraulic piston pump based on the comparison results.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] 1. This invention provides an intelligent fault diagnosis method and system for hydraulic piston pumps based on pressure signals, which significantly improves the efficiency and accuracy of fault diagnosis. The technology arranges the hydraulic piston pump pressure signals in a spiral structure and performs interpolation processing to generate an optimized pressure image. This image is then enhanced with visual enhancement processing to strengthen essential texture features. By analyzing multi-scale global brightness statistical features through an image pyramid structure and scanning region by region to extract local texture gradient change features, the technology achieves comprehensive capture of fault-related visual features, providing complete and accurate image feature support for diagnosis.

[0063] 2. This technical solution boasts significant advantages in feature fusion and fault diagnosis. Using global features as the upper-level framework and local features as the lower-level benchmark, it constructs overall feature parameters covering both macroscopic and microscopic aspects, quantifying feature correlations using formulas. The overall feature parameters are then compared at multiple levels with hierarchical preset thresholds, and a comprehensive diagnostic decision vector is formed through cross-validation. Based on the vector's direction, the fault state type is accurately determined. The entire process is based on image visual analysis technology, achieving end-to-end intelligent diagnosis from signal imaging to feature extraction, fusion, and judgment, significantly improving the efficiency and reliability of hydraulic piston pump fault identification and providing a scientific basis for equipment maintenance. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating an intelligent fault diagnosis method for a hydraulic piston pump based on pressure signals, provided in an embodiment of the present invention.

[0065] Figure 2 A functional block diagram of a hydraulic piston pump intelligent fault diagnosis system based on pressure signals, provided in an embodiment of the present invention;

[0066] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0067] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0068] This application provides an intelligent fault diagnosis method for hydraulic piston pumps based on pressure signals. The executing entity of this intelligent fault diagnosis method for hydraulic piston pumps based on pressure signals includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the intelligent fault diagnosis method for hydraulic piston pumps based on pressure signals can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0069] Reference Figure 1 The diagram shown is a flowchart illustrating an intelligent fault diagnosis method for a hydraulic piston pump based on pressure signals, according to an embodiment of the present invention. In this embodiment, the intelligent fault diagnosis method for a hydraulic piston pump based on pressure signals includes:

[0070] S1. Arrange the pressure signals of the hydraulic piston pump according to the spiral structure, and perform interpolation processing on the arranged signals to obtain the optimized pressure image of the hydraulic piston pump.

[0071] In this embodiment of the invention, the step of arranging the pressure signals of the hydraulic plunger pump according to a spiral structure and interpolating the arranged signals to obtain an optimized pressure image of the hydraulic plunger pump includes:

[0072] The pressure signal of the hydraulic piston pump is collected during operation.

[0073] The starting point of the spiral structure is taken as the center of the image plane, and the unfolding direction of the spiral path in the spiral structure is determined by the equal angle increment of the spiral structure.

[0074] The pressure signals are distributed to equal-angle positions along the spiral path in chronological order to obtain a spiral distribution of the pressure signals.

[0075] In the spiral distribution, the pressure signal corresponding to the equiangular position is mapped to the image grayscale value to obtain the initial image of the spiral grayscale distribution in the hydraulic plunger pump;

[0076] Identify blank areas in the initial image caused by path gaps in the spiral path;

[0077] Fill in the missing pixel values ​​in the blank areas of the image to obtain an optimized pressure image of the hydraulic piston pump.

[0078] In the spiral distribution, mapping the pressure signals corresponding to the equiangular positions into image grayscale values ​​to obtain an initial image of the spiral grayscale distribution in the hydraulic piston pump includes:

[0079] Using the time-domain characteristics of the pressure signal and the combined characterization of the position and amplitude characteristics in the spiral distribution, a pressure signal characterization space for the spiral distribution is constructed.

[0080] In the pressure signal characterization space, the pressure signals at the isoangular positions are mapped onto a preset grayscale base to obtain grayscale characterization values ​​with visual discrimination at the isoangular positions.

[0081] According to the topological structure of the spiral path, the grayscale representation value is projected onto the image coordinate position of the image plane;

[0082] Along the spiral path, the grayscale representation value is reconstructed using the visual saliency distribution of the pressure signal to obtain an initial image of the spiral grayscale distribution.

[0083] After the hydraulic plunger pump starts and enters a stable operating state, a high-precision pressure sensor fixedly connected to the oil outlet or oil inlet of the hydraulic plunger pump continuously captures the pressure value at every moment during the pump's operation. These pressure values ​​are continuously recorded in the order of time, forming a pressure signal sequence that can completely reflect the pressure fluctuations under the working state of the hydraulic plunger pump.

[0084] A fixed-size two-dimensional image plane is preset, and the starting point of the spiral structure is precisely located at the geometric center of the image plane. Then, a fixed equal angle increment is set, and starting from the starting point, the spiral path gradually extends to the four sides of the image plane according to the equal angle increment, clarifying the specific direction of the spiral path from the center outward, and ensuring that the spiral path extends in a uniform and equal angle with the center as the origin.

[0085] The collected pressure signals are sorted sequentially according to the time they were recorded. Starting from the earliest collected pressure signal, each signal is mapped to a predetermined equiangular position on the spiral path. Each pressure signal at any given time point is uniquely matched with an equiangular position, so that the pressure signals are arranged in an orderly manner along the spiral path in chronological order, ultimately forming a spiral distribution of pressure signals.

[0086] Using the fluctuation pattern of pressure signal over time as the time domain feature, the pressure signal's corresponding equiangular position information in the spiral distribution as the position feature, and the numerical value of each pressure signal as the amplitude feature, these three features are fused together to construct a pressure signal representation space that can comprehensively describe the integrated characteristics of pressure signals in the spiral distribution. This space can simultaneously reflect the signal's time attributes, spatial position attributes, and numerical attributes.

[0087] A standard grayscale base containing multiple continuous grayscale levels from light to dark is pre-defined. In the constructed pressure signal representation space, for each pressure signal at an equal angle position, its temporal, positional, and amplitude characteristics are comprehensively compared and matched with the grayscale levels in the pre-defined grayscale base. A unique grayscale level is determined for each pressure signal. This grayscale level is the grayscale representation value with obvious visual distinguishability, ensuring that the grayscale representation values ​​corresponding to pressure signals with different characteristics can be clearly distinguished visually.

[0088] The connection method and spatial layout of the spiral path on the image plane are clearly defined, that is, the topological structure of the spiral path. Based on this structure, the specific coordinates of each isoangular position on the image plane are determined. Then, the grayscale representation value corresponding to each isoangular position is accurately projected onto its corresponding image coordinate position, so that the grayscale representation value forms a preliminary distribution on the image plane according to the layout of the spiral path.

[0089] The pressure signal changes along the spiral path are analyzed to identify those signal segments with large fluctuations that clearly reflect the operating status of the hydraulic piston pump. The distribution of these signal segments along the spiral path is determined, i.e., the visual saliency distribution of the pressure signal. Based on this distribution, the grayscale values ​​at each image coordinate position are adjusted and optimized. For positions with significant signal changes, the visual contrast of the grayscale values ​​is enhanced, while for positions with gradual signal changes, the stability of the grayscale values ​​is maintained. Through this reconstruction process, an initial image of the spiral grayscale distribution is formed.

[0090] The initial image is thoroughly examined by scanning row by row and column by column. Each image coordinate position is carefully observed to see if there are any areas without grayscale values. Since there are fixed intervals between the spiral paths, the image coordinate positions corresponding to these intervals will form blank areas without pixel information. The boundary range and image coordinates of each blank area are accurately determined by scanning.

[0091] For each identified blank area, the pixel information of the surrounding effective grayscale values ​​is extracted, the grayscale values ​​of these adjacent effective pixels are counted, and the average value of these grayscale values ​​is calculated. This average value is used as the grayscale value of the corresponding missing pixel position in the blank area. All missing pixels in the blank area are filled one by one to ensure that the grayscale value of the filled pixel transitions naturally with the grayscale value of the surrounding effective pixels. When all missing pixels in the blank areas are filled, an optimized pressure image of the hydraulic piston pump that is complete, without blanks, and with uniform pixel distribution is obtained.

[0092] The beneficial effects are that the pressure signal of the hydraulic piston pump is transformed into a complete optimized pressure image through a spiral structure arrangement and a series of image processing. This not only fully preserves the key features of the pressure signal, such as time domain, position, and amplitude, but also eliminates image blanks through precise mapping, projection, and interpolation filling, thereby improving the integrity and visual clarity of the image. This provides high-quality and reliable image data support for subsequent image-based visual enhancement, feature analysis, and other fault diagnosis processes, effectively ensuring the accuracy of subsequent feature extraction and laying a solid image foundation for the accurate judgment of the fault state of the hydraulic piston pump.

[0093] S2. Perform visual enhancement processing on the optimized pressure image to obtain the enhanced pressure image of the hydraulic piston pump;

[0094] In this embodiment of the invention, the step of performing visual enhancement processing on the optimized pressure image to obtain the enhanced pressure image of the hydraulic plunger pump includes:

[0095] By analyzing the visual characteristics of the optimized stress image, structural feature information of the optimized stress image at different scales is obtained.

[0096] The structural feature information is reconstructed to obtain the structure-aware enhanced map of the optimized pressure image;

[0097] Based on the structure-aware enhanced map, adaptive contrast enhancement is performed on the optimized pressure image to obtain the visual mode of the operating status related to the hydraulic piston pump in the optimized pressure image.

[0098] The visual mode of the running state is subjected to detail restoration processing to obtain the essential feature texture information of the pressure signal in the optimized pressure image;

[0099] The essential feature texture information is input into the optimized pressure image to obtain the enhanced pressure image of the hydraulic plunger pump.

[0100] When performing visual characteristic analysis on optimized stress images, multiple observation windows of different sizes are first set up. Small windows are used to capture the details of grayscale changes in local pixels, while large windows are used to capture the overall regional structural distribution of the image. The optimized stress image is scanned pixel by pixel through the small windows, and the magnitude and direction of the differences in pixel grayscale values ​​within each window are recorded to determine the edge positions and subtle texture directions in the image. The image is then divided into regions using the large windows, and the average level and distribution range of pixel grayscale values ​​within each region are statistically analyzed to determine the division of bright and dark regions and the overall structural contour of the image. The edge information, texture direction, division of bright and dark regions, and overall structural contour obtained through the analysis of windows of different sizes are classified and organized according to the size of the observation windows to form structural feature information that can completely reflect the structural characteristics of the optimized stress image at different scales.

[0101] When reconstructing the structural feature information atlas, a blank atlas framework with the same size as the optimized stress image is first constructed, with each pixel coordinate in the framework corresponding one-to-one with the pixel coordinates in the optimized stress image. Structural feature information at different scales is then filled into the blank atlas framework according to its corresponding image position. For positions containing edge information, the direction of the edge is highlighted in the atlas using specific visual markers. For regions containing texture direction, continuous visual trajectories are drawn in the atlas according to the extension direction of the texture. For parts containing the division of light and dark regions, regions with different degrees of lightness and darkness are distinguished in the atlas using different visual depth markers. In this way, the scattered structural feature information is integrated into a complete atlas that can intuitively reflect the structural perception characteristics of the optimized stress image, thus obtaining the structure perception enhancement atlas of the optimized stress image.

[0102] When performing adaptive contrast enhancement on optimized pressure images based on structure-aware enhanced maps, the structure types of different regions in the structure-aware enhanced map are first analyzed to determine whether each region is an edge region, a texture region, or a region with uneven brightness. For edge regions marked in the map, the gray-level difference between adjacent pixels in the region is adjusted to further widen the gray-level difference between adjacent pixels that originally had a small gray-level difference, making the edge lines clearer and more distinguishable in the image. For texture regions marked in the map, the overall gray-level span of pixels in the region is calculated, and the gray-level range with a small span is appropriately expanded so that each detail unit of the texture can be clearly distinguished in the image. For regions with uneven brightness marked in the map, the gray-level values ​​of pixels in overly bright areas are appropriately reduced, and the gray-level values ​​of pixels in overly dark areas are appropriately increased to make the gray-level distribution in the region more uniform, while preserving the original structural features in the region. After the above targeted contrast adjustments, the visual features that can reflect the operating status of the hydraulic piston pump in the optimized pressure image are highlighted, forming a visual pattern of the operating status related to the hydraulic piston pump.

[0103] When performing detail restoration processing on the visual pattern of the operating status, the process first checks each region for areas with blurred textures or missing details. Blurred textures are characterized by breaks in what should be continuous textures, while missing details are characterized by smooth pixel grayscale changes and a lack of obvious structural features in localized areas. For blurred textures, the direction and grayscale change patterns of the surrounding clear textures are referenced, extending the clear textures into the blurred areas. Simultaneously, the missing grayscale values ​​in the blurred areas are supplemented according to the grayscale change intervals of the clear textures, thus reconnecting the broken textures. For missing details, the structural features of the missing area are compared with those of adjacent areas. Based on the structural features of the adjacent areas, the expected texture shape of the missing part is inferred, and the corresponding grayscale value is assigned to the area according to the inferred texture shape, thus supplementing the missing details. Through the above processing, all key texture details related to the pressure signal in the visual pattern of the operating status are restored, resulting in essential feature texture information that accurately reflects the essential characteristics of the hydraulic piston pump pressure signal.

[0104] When inputting essential feature texture information into the optimized pressure image, the coordinate correspondence between the essential feature texture information and the optimized pressure image is first established to ensure that the position of each pixel in the essential feature texture information can be accurately matched to the same pixel position in the optimized pressure image. Then, the gray value corresponding to each pixel in the essential feature texture information replaces the original pixel gray value at the same coordinate position in the optimized pressure image one by one. During the replacement process, it is necessary to ensure that the gray value of each pixel is accurately transmitted without omission or misalignment. After all the gray values ​​corresponding to the essential feature texture information have completed the pixel update of the optimized pressure image, the original optimized pressure image is integrated with clear essential feature texture, forming an enhanced pressure image of the hydraulic piston pump that can more clearly present the pressure signal characteristics.

[0105] The beneficial effects are that by performing visual characteristic analysis, map reconstruction, adaptive contrast enhancement, detail restoration, and texture information fusion on the optimized pressure image step by step, key visual features related to the hydraulic piston pump pressure signal are gradually highlighted and restored. This effectively solves the problems of unclear structure and blurred details in the optimized pressure image. The resulting enhanced pressure image can more accurately present the visual information corresponding to the pressure signal, providing a high-quality image foundation for subsequent analysis of global and local feature parameters from the image. This ensures the accuracy and completeness of subsequent feature extraction, and thus provides image processing support for the reliability of hydraulic piston pump fault diagnosis.

[0106] S3. Analyze the multi-scale features in the enhanced pressure image to obtain the global feature parameters of the image brightness statistical features in the hydraulic piston pump;

[0107] In this embodiment of the invention, the step of analyzing the multi-scale features in the enhanced pressure image to obtain global feature parameters of the image brightness statistical features in the hydraulic plunger pump includes:

[0108] Using visual information at different scales in the enhanced stress image, an image pyramid structure of the enhanced stress image is constructed;

[0109] In the image pyramid structure, the brightness statistics of the brightness distribution characteristics in the enhanced pressure image are extracted;

[0110] Cross-scale correlation analysis is performed on the brightness statistics to obtain the correlation matrix of the enhanced stress image;

[0111] Based on the correlation matrix, tensor synthesis is performed on the overall operating state of the hydraulic piston pump to obtain the brightness feature vector of the hydraulic piston pump.

[0112] By normalizing the different dimensions of the brightness feature vector, the global feature parameters of the hydraulic plunger pump are obtained.

[0113] When processing enhanced stress images, the number of scale levels required to construct the image pyramid structure is first determined, starting from the original size of the enhanced stress image as the bottom layer of the pyramid. Each time, the image at the current level is scaled down. During the scaling down process, the average grayscale value of each 2×2 pixel region of the current image is calculated, and this average value is used as the grayscale value of the corresponding single pixel in the scaled-down image. This ensures that the scaled-down image retains the brightness information of the original image region. Image levels with decreasing sizes are generated sequentially in this way until the generated image size reaches the preset minimum size. These images of different sizes are stacked sequentially in descending order to form the image pyramid structure of the enhanced stress image.

[0114] For each layer of the image pyramid structure, the grayscale values ​​of all pixels in that layer are read pixel by pixel, and the brightness statistics of that layer are calculated. The sum of the grayscale values ​​of all pixels in that layer is calculated, and the sum is divided by the total number of pixels in that layer to obtain the average brightness of that layer. The grayscale values ​​of all pixels in that layer are traversed, and the grayscale value with the largest value is recorded as the maximum brightness value of that layer, and the grayscale value with the smallest value is recorded as the minimum brightness value of that layer. The difference between the grayscale value of each pixel and the average brightness is calculated, and the squares of each difference are summed. The sum is then divided by the total number of pixels to obtain the brightness variance of that layer. The average brightness, maximum brightness value, minimum brightness value, and brightness variance of each layer are used together as the brightness statistics of that layer. After completing the statistics for all layers, the brightness statistics of the brightness distribution characteristics in the enhanced stress image are obtained.

[0115] First, number all levels of the image pyramid in order from top to bottom, labeling them as level 1, level 2, ..., level n. Then, calculate the correlation between the brightness statistics of adjacent levels. Taking level 1 and level 2 as an example, subtract the average brightness of level 1 from the average brightness of level 2, and take the absolute value of the difference as the correlation value of their average brightness. Similarly, subtract the maximum brightness of level 1 from the maximum brightness of level 2, and take the absolute value as the correlation value of their maximum brightness. The minimum brightness and maximum brightness are calculated using the same method. Correlation values ​​of variance; following the above method, calculate the correlation values ​​of various brightness statistics for layer 2 and layer 3, layer 3 and layer 4, ... layer n-1 and layer n in sequence; construct a two-dimensional matrix, where the rows of the matrix correspond to the four statistical types of average brightness, maximum brightness, minimum brightness, and brightness variance, and the columns of the matrix correspond to adjacent layer pairs such as layer 1-layer 2, layer 2-layer 3, ... layer n-1-layer n. Fill the calculated correlation values ​​into the corresponding rows and columns of the matrix to obtain the correlation matrix of the enhanced stress image.

[0116] First, clarify the meaning of each element in the correlation matrix. Each element represents the degree of correlation between specific brightness statistics at specific adjacent levels. These correlations directly reflect the relationship between the brightness changes of the enhanced pressure image at different scales and the operating state of the hydraulic piston pump. Extract all elements in the correlation matrix in row-major order to form a one-dimensional numerical sequence. Each value in the sequence corresponds to a brightness correlation information related to the pump's operating state. Bind this one-dimensional numerical sequence to the overall operating state of the hydraulic piston pump. For example, a certain value in the sequence corresponds to the brightness feature related to the pressure stability of the pump at a certain operating stage. Through this binding, the one-dimensional numerical sequence is transformed into a vector form that can comprehensively characterize the overall operating state of the hydraulic piston pump. This vector is the brightness feature vector of the hydraulic piston pump.

[0117] First, iterate through all dimensions of the brightness feature vector, recording the maximum and minimum values ​​of each dimension to determine the overall distribution range of the values ​​in the current vector. Then, set a unified target value range and normalize the values ​​of each dimension by subtracting the minimum value of the dimension from the current value, dividing the difference by the difference between the maximum and minimum values ​​of the dimension, and multiplying the result by 100 to obtain the normalized value of the dimension within the target range. Repeat this process for all dimensions of the brightness feature vector, and then rearrange all the normalized values ​​according to the original dimension order to form a new vector. This new vector is the global feature parameter of the hydraulic piston pump.

[0118] The beneficial effects are that by constructing an image pyramid structure, a systematic sorting of multi-scale visual information of enhanced pressure images is achieved. Then, through steps such as brightness statistics extraction, cross-scale correlation analysis, tensor synthesis, and vector normalization, global feature parameters of image brightness statistics that can reflect the operating status of hydraulic piston pumps are accurately obtained. The entire process ensures the integrity and accuracy of global feature parameters through explicit image processing operations, providing a reliable feature basis for global-level status judgment in subsequent hydraulic piston pump fault diagnosis, and effectively improving the scientificity and effectiveness of fault diagnosis in the global feature analysis stage.

[0119] S4. Perform local texture feature analysis on the enhanced pressure image to obtain local feature parameters of the image gradient change in the hydraulic plunger pump;

[0120] In this embodiment of the invention, the step of performing local texture feature analysis on the enhanced pressure image to obtain local feature parameters of the image gradient change in the hydraulic plunger pump includes:

[0121] The local texture features in the enhanced stress image are scanned region by region to obtain the spatial variation pattern of the enhanced stress image;

[0122] Based on the changing trend of the spatial variation pattern, the directional gradient distribution of texture orientation characteristics in the enhanced stress image is determined;

[0123] Regional aggregation of gradient change descriptors in the directional gradient distribution yields a texture primitive feature map of the enhanced stress image.

[0124] The gradient statistics in the texture primitive feature map are used as local feature parameters of the enhanced stress image.

[0125] When processing enhanced stress images, a square sliding window with a fixed size of 8×8 pixels is used. Starting from the top left pixel position of the enhanced stress image, the window is scanned column by column by moving one pixel horizontally each time. After scanning a row, the window moves down one pixel to start scanning the next row, until the sliding window covers all pixel areas of the enhanced stress image. Within the area covered by each sliding window, the gray value difference between adjacent pixels within the window is calculated one by one, and the sign and absolute value of each difference are recorded. The gray value change direction and amplitude information recorded in all sliding windows are integrated according to the position of the windows in the image to form a spatial change pattern that can reflect the gray value spatial change law of pixels in different areas of the enhanced stress image.

[0126] After obtaining the spatial change pattern, the gray-level change direction and amplitude information recorded in each region are analyzed to determine the dominant trend of gray-level change in each region. For example, if the sum of the absolute values ​​of gray-level differences in the horizontal direction is the largest in a certain region, then the gray-level change trend in that region is mainly in the horizontal direction; if the sum of the absolute values ​​of gray-level differences in the vertical direction is the largest in a certain region, then the gray-level change trend in that region is mainly in the vertical direction. At the same time, the sum of the gray-level change amplitude under the dominant change trend in each region is recorded. The dominant change direction and the sum of the amplitude under that direction for each region of the enhanced stress image are recorded one by one according to the region's position in the image, forming a directional gradient distribution that can reflect the texture direction and corresponding change intensity of each region of the enhanced stress image.

[0127] First, the enhanced stress image is divided into multiple non-overlapping square regions of size 16×16 pixels, with each square region serving as a clustering unit. Simultaneously, the gradient change descriptor corresponding to each pixel location in the directional gradient distribution, composed of the gradient direction and magnitude at that location, is associated with the image pixel location. Within each clustering unit, gradient change descriptors for all pixel locations covered by that unit are collected. The sum of the gradient magnitudes for each gradient direction within that unit is calculated, and the proportion of the sum of the magnitudes for each gradient direction to the sum of all gradient magnitudes within that unit is used as the texture feature information of that clustering unit. The texture feature information of all clustering units is then presented in image form according to their position in the enhanced stress image. Each clustering unit corresponds to a pixel in the image, and the features of that pixel are determined by the texture feature information of the clustering unit, thus obtaining the texture primitive feature map of the enhanced stress image.

[0128] For each aggregated unit in the texture primitive feature map, its corresponding gradient statistics are extracted. Specifically, these include the average value of all gradient amplitudes within the aggregated unit (i.e., summing all gradient amplitudes within the unit and dividing by the total number of gradient amplitudes), the maximum value of the gradient amplitude (i.e., selecting the largest value after traversing all gradient amplitudes within the unit), the minimum value of the gradient amplitude (i.e., selecting the smallest value after traversing all gradient amplitudes within the unit), and the proportion value corresponding to the gradient direction with the largest proportion. These gradient statistics of all aggregated units are then arranged into a continuous set of data according to the order of the aggregated units in the texture primitive feature map. This set of data is the local feature parameters of the enhanced stress image.

[0129] The beneficial effects are that by scanning regions one by one, the spatial variation patterns of enhanced pressure images are accurately captured, and the gradient distribution of texture direction is clarified by combining the trend analysis. Then, texture primitive feature maps are formed through regional aggregation, and finally, gradient statistics are extracted to obtain local feature parameters. The whole process is progressive and the operation is clear, ensuring that the local feature parameters can accurately reflect the gradient change differences of local texture in the enhanced pressure image. These differences are directly related to the local state of the hydraulic piston pump during operation. For example, poor piston sealing may cause local pressure signal fluctuations, which are reflected in abnormal changes in local texture gradients of the image. Therefore, the local feature parameters obtained by this process can provide accurate feature support for the identification of local faults in hydraulic piston pumps, effectively improving the accuracy and reliability of fault diagnosis in the detection of local subtle anomalies.

[0130] S5. Using the global feature parameters as the upper-level framework and the local feature parameters as the lower-level reference, construct the overall feature parameters of the operating state of the hydraulic piston pump.

[0131] In this embodiment of the invention, the step of constructing the overall characteristic parameters of the operating state of the hydraulic piston pump, using the global characteristic parameters as the upper-level framework and the local characteristic parameters as the lower-level benchmark, includes:

[0132] The global feature parameters are configured as the top-level representation of the hydraulic piston pump, and the local feature parameters are configured as the bottom-level representation of the hydraulic piston pump.

[0133] The spatial distribution of the bottom-level representation and the feature dimensions of the top-level representation are synchronously aligned to obtain the preliminary feature relationship of the hydraulic piston pump.

[0134] Verify the feature integrity in the preliminary feature relationship to obtain the feature association relationship of the hydraulic piston pump;

[0135] Based on the aforementioned feature association, the local feature parameters are embedded as nodes into the overall framework of the global feature parameters to obtain the overall feature parameters of the operating state of the hydraulic piston pump.

[0136] The calculation formula for the preliminary characteristic relationship is as follows: ;

[0137] In the formula, For the first The first underlying representation and the first The initial relationship strength between the top-level representations The local feature parameter is the first one. The spatial location of the first 1 eigenvalue, The first of the global feature parameters 1 eigenvalue, It is a preset small positive number.

[0138] The global feature parameters are defined as the top-level representation of the hydraulic piston pump. This top-level representation directly corresponds to the macroscopic features of the overall operating state of the hydraulic piston pump, covering the overall information related to image brightness statistics, and can reflect the overall trend of pressure signal changes during pump operation. At the same time, the local feature parameters are defined as the bottom-level representation of the hydraulic piston pump. This bottom-level representation corresponds to the microscopic features of the local operating state of the hydraulic piston pump, covering the detailed information related to the changes in local image texture gradients, and can reflect the subtle fluctuations in pressure signals of individual pump components or local areas. The configuration of the top-level and bottom-level representations is completed through this definition.

[0139] First, the spatial distribution of the bottom-level representation is clarified, that is, the specific location of the enhanced pressure image corresponding to the local feature parameters. Each local feature parameter corresponds one-to-one with a specific spatial region in the image. Then, the feature dimensions of the top-level representation are clarified, that is, the specific brightness statistical features represented by each dimension in the global feature parameters. Each dimension corresponds to a certain macroscopic indicator of the overall operation of the pump. Subsequently, each spatial region of the bottom-level representation is checked one-to-one with each feature dimension of the top-level representation to determine the correspondence between the local features of each spatial region and each macroscopic indicator. For example, the local features of the image edge region are associated with the brightness variance dimension in the top-level representation, and the local features of the image center region are associated with the average brightness dimension in the top-level representation. Through such correspondence operations, the preliminary feature relationship of the hydraulic piston pump is obtained.

[0140] A comprehensive check of the preliminary feature relationships is conducted. First, it is checked whether all feature dimensions of the top-level representation have corresponding spatial regions associated with the bottom-level representation, ensuring that no macroscopic indicator is without underlying detail support. Next, it is checked whether all spatial regions of the bottom-level representation have corresponding feature dimensions associated with the top-level representation, ensuring that no microscopic feature of a local region is without a macroscopic framework. If unassociated feature dimensions or spatial regions are found, corresponding relationships are established based on the transmission law of pressure signals during the operation of the hydraulic piston pump. For example, the unassociated local region features are associated with the feature dimensions that best reflect the macroscopic impact of that region. After completing all relationship checks and additions, the feature relationship of the hydraulic piston pump is obtained.

[0141] Based on the overall framework formed by global feature parameters, the specific position of each local feature parameter in the global framework is determined according to the feature correlation. Each local feature parameter is treated as an independent node and embedded into the top-level feature dimension associated with it. For example, the local feature parameters of all local regions associated with the maximum brightness dimension in the top-level representation are uniformly embedded into the framework branch corresponding to the maximum brightness, so that each top-level feature dimension can contain all local detail information related to it. Finally, a complete feature structure that includes both macro-global features and micro-local features is formed. This structure is the overall feature parameter of the operating state in the hydraulic piston pump.

[0142] The feature value of the corresponding dimension at a certain spatial location in the local feature parameters comes from the process of analyzing the local texture features of the enhanced stress image. The local texture features of the enhanced stress image are scanned region by region to obtain the spatial change pattern. The directional gradient distribution is determined according to the change trend of the spatial change pattern. The gradient change descriptors in the directional gradient distribution are regionally aggregated to obtain the texture primitive feature map. This value is the gradient statistic for a certain spatial location and corresponding dimension extracted from the texture primitive feature map. The feature value of the corresponding dimension in the global feature parameters comes from the process of analyzing the multi-scale features of the enhanced stress image. The image pyramid structure of the enhanced stress image is constructed. The brightness statistics are extracted and cross-scale correlation analysis is performed to obtain the correlation matrix. Tensor synthesis is performed based on the correlation matrix to obtain the brightness feature vector. After normalizing the different dimensions of the vector, the value of the corresponding dimension is extracted from the normalized vector, which is the feature value. The preset small positive number is a fixed constant set in advance before calculating the initial relationship strength. The purpose of setting it is to avoid the case that the denominator is zero during the calculation. Its value is determined before the calculation and does not change with the changes of local or global feature parameters.

[0143] The core of this calculation process is to quantify the correlation between the low-level representation corresponding to the local feature parameters and the high-level representation corresponding to the global feature parameters. The calculation first multiplies the corresponding dimensional feature value at a certain spatial location in the local feature parameters with the corresponding dimensional feature value in the global feature parameters. The product is then multiplied by 2 to obtain the numerator. Next, the square of the corresponding dimensional feature value at a certain spatial location in the local feature parameters is calculated, as is the square of the corresponding dimensional feature value in the global feature parameters. These two squared results are added to a preset small positive number to obtain the denominator. Dividing the numerator by the denominator yields the preliminary relationship strength between a low-level representation and a high-level representation. This result directly reflects the degree of fit between the local features at a specific spatial location and the corresponding global features in the same dimension.

[0144] When the value of the corresponding dimension feature at a certain spatial location in the local feature parameters becomes increasingly close to the value of the corresponding dimension feature in the global feature parameters, the calculated initial relationship strength increases accordingly. This means that the degree of fit between the local feature at that spatial location and the global feature of that dimension is continuously improving. Conversely, when the difference between the value of the corresponding dimension feature at a certain spatial location in the local feature parameters and the value of the corresponding dimension feature in the global feature parameters gradually widens, the calculated initial relationship strength decreases accordingly. This indicates that the degree of fit between the local feature at that spatial location and the global feature of that dimension is continuously decreasing. The preset small positive value is fixed and extremely small. Its existence will not have a significant impact on the overall trend of the initial relationship strength. It is only used when the value of the corresponding dimension feature at a certain spatial location in the local feature parameters and the value of the corresponding dimension feature in the global feature parameters are both zero, to avoid the denominator being zero and causing the calculation to fail. In this case, the initial relationship strength is zero, indicating that there is no effective correlation between the bottom-level representation and the top-level representation in this situation.

[0145] The beneficial effects are that, through clear top-level and bottom-level representation division, precise synchronous alignment of spatial distribution and feature dimensions, rigorous verification of feature integrity, and orderly embedding of local parameters into the global framework, a deep integration of macroscopic operating characteristics and microscopic local characteristics of hydraulic piston pumps is achieved. The constructed overall feature parameters not only contain global information reflecting the overall pressure signal change trend of the pump, but also microscopic information reflecting the details of pressure fluctuations in local components. This effectively solves the problem that a single feature parameter is difficult to fully characterize the pump's operating status, providing a hierarchical and comprehensive feature foundation for subsequent fault diagnosis through overall feature parameters, and significantly improving the comprehensiveness and accuracy of fault diagnosis in judging the pump's operating status.

[0146] S6. Perform multi-level comparison between the overall feature parameters and the preset standard feature threshold, and determine the fault status of the hydraulic piston pump based on the comparison results.

[0147] In this embodiment of the invention, the step of performing multi-level comparisons between the overall feature parameters and preset standard feature thresholds, and determining the fault state of the hydraulic piston pump based on the comparison results, includes:

[0148] The preset standard feature thresholds are divided into macro-consistency thresholds and micro-difference thresholds according to feature level;

[0149] The global feature components in the overall feature parameters are compared with the macro-consistency threshold at the first level to obtain the consistency evaluation result of the global feature components.

[0150] The local feature components in the overall feature parameters are compared with the micro-difference threshold at a second level to obtain the difference evaluation result of the local feature components;

[0151] Cross-validate the consistency assessment results and the difference assessment results to obtain the comprehensive diagnostic decision vector of the overall feature parameters;

[0152] Based on the pointing characteristics of the comprehensive diagnostic decision vector, the fault state type of the hydraulic piston pump is determined.

[0153] First, multiple sets of overall characteristic parameters of the hydraulic piston pump are collected under long-term, continuous, stable, and fault-free operation conditions using the aforementioned method of this invention. These overall characteristic parameters under normal conditions are classified and statistically analyzed. For the global characteristic components, the stable value range of each global characteristic component under normal operation is calculated, and these stable value ranges are set as macroscopic consistency thresholds. At the same time, for the local characteristic components in the overall characteristic parameters under normal conditions, the allowable fluctuation range of each local characteristic component under normal operation is calculated, and these allowable fluctuation ranges are set as microscopic difference thresholds. Through such classification, statistics, and interval setting, the operation of dividing the preset standard characteristic threshold into macroscopic consistency thresholds and microscopic difference thresholds according to the characteristic level is completed.

[0154] All global feature components are extracted from the overall feature parameters. The macro-consistency threshold corresponding to each global feature component is determined one by one. The specific value of each global feature component is directly compared with the corresponding macro-consistency threshold. If the value of a global feature component falls within its corresponding macro-consistency threshold range, the global feature component is determined to meet the macro-standard for normal operation. If the value of a global feature component exceeds its corresponding macro-consistency threshold range, the global feature component is determined to not meet the macro-standard for normal operation. The "compliant" or "non-compliant" judgment results of all global feature components are sorted in order of type of global feature component to form the consistency evaluation result of global feature components.

[0155] All local feature components are extracted from the overall feature parameters. Based on the image region location corresponding to each local feature component, its corresponding micro-difference threshold is found. The specific value of each local feature component is compared with its corresponding micro-difference threshold. If the value of a local feature component is within the allowable fluctuation range of its corresponding micro-difference threshold, it is determined that the local feature component meets the micro-difference requirements for normal operation. If the value of a local feature component exceeds the allowable fluctuation range of its corresponding micro-difference threshold, it is determined that the local feature component has micro-abnormal fluctuations. The "normal" or "abnormal" judgment results of all local feature components are arranged in order of their corresponding image region location to form the difference evaluation result of the local feature components.

[0156] First, examine all "non-compliant" global feature components in the consistency assessment results. Then, in the difference assessment results, find the local feature components corresponding to the image regions associated with these "non-compliant" global feature components. Determine whether there are "abnormalities" in the local feature components of these associated regions. For example, if the "brightness variance" in the consistency assessment results does not meet the macro threshold, check whether there are abnormalities in the local feature components of each local region within the overall region reflected by the brightness variance. If there are abnormalities, it indicates that there is a causal relationship between the two. At the same time, check whether the "abnormal" local feature components in the difference assessment results have corresponding global feature components that show "non-compliance" in the consistency assessment results. If they do, further confirm the correlation. If they do not, record them as independent micro anomalies. Organize all correlation judgment results and anomaly labels according to feature type to form a comprehensive diagnostic decision vector that includes the correlation between macro and micro anomalies.

[0157] By analyzing the anomaly types, correlations, and corresponding image regions recorded in the comprehensive diagnostic decision vector, if all global feature components in the decision vector are "compliant" and all local feature components are "normal," the fault state type of the hydraulic piston pump is determined to be fault-free. If a certain global feature component in the decision vector is "non-compliant" and there are corresponding local feature components that are "abnormal," such as "average brightness" not meeting the macroscopic threshold and the corresponding local feature components in the pump piston movement area being abnormal, the fault state type is determined to be piston sealing failure based on the function of the pump component corresponding to that area. If only local feature components are "abnormal" in the decision vector and there are no corresponding global feature components that are "non-compliant," the fault state type is determined to be an early minor fault of that component based on the pump component associated with the image region corresponding to the abnormal local feature component. The fault state type of the hydraulic piston pump is determined through this analysis.

[0158] The beneficial effects are as follows: by dividing the preset standard feature thresholds into layers, targeted comparisons of macroscopic and microscopic features in the overall feature parameters are achieved, avoiding the one-sidedness of single threshold comparisons; the separate comparisons of the first and second levels ensure that both macroscopic global feature anomalies and microscopic local feature anomalies can be accurately identified; the cross-validation step effectively establishes the causal relationship between macroscopic and microscopic anomalies, eliminating misjudgments caused by occasional interference; the comprehensive diagnostic decision vector fully integrates anomaly information, and based on its directional features, it can accurately locate the fault location and fault degree, solving the problems of missed judgments, misjudgments, and ambiguous fault location that are prone to occur in traditional diagnostic methods, significantly improving the accuracy, reliability, and precision of fault location in hydraulic piston pump fault diagnosis, and providing a scientific basis for timely maintenance and fault repair of hydraulic piston pumps.

[0159] like Figure 2 The diagram shown is a functional block diagram of an intelligent fault diagnosis system for a hydraulic piston pump based on pressure signals, provided by an embodiment of the present invention.

[0160] The intelligent fault diagnosis system 100 for hydraulic piston pumps based on pressure signals described in this invention can be installed in electronic devices. Depending on the functions implemented, the intelligent fault diagnosis system 100 for hydraulic piston pumps based on pressure signals may include an image construction module 101, an image enhancement module 102, a global feature module 103, a local feature module 104, an overall feature module 105, and a fault judgment module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0161] In this embodiment, the functions of each module / unit are as follows:

[0162] The image construction module 101 is used to arrange the pressure signals of the hydraulic plunger pump in a spiral structure and to perform interpolation processing on the arranged signals to obtain an optimized pressure image of the hydraulic plunger pump.

[0163] The image enhancement module 102 is used to perform visual enhancement processing on the optimized pressure image to obtain the enhanced pressure image of the hydraulic piston pump;

[0164] The global feature module 103 is used to analyze the multi-scale features in the enhanced pressure image to obtain the global feature parameters of the image brightness statistical features in the hydraulic piston pump.

[0165] The local feature module 104 is used to perform local texture feature analysis on the enhanced pressure image to obtain local feature parameters of the image gradient change in the hydraulic plunger pump.

[0166] The overall feature module 105 is used to construct the overall feature parameters of the operating state of the hydraulic piston pump, with the global feature parameters as the upper framework and the local feature parameters as the lower benchmark.

[0167] The fault judgment module 106 is used to perform multi-level comparison of the overall feature parameters with preset standard feature thresholds, and determine the fault status of the hydraulic piston pump based on the comparison results.

[0168] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0169] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0170] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0171] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0172] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0173] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent fault diagnosis of hydraulic piston pumps based on pressure signals, characterized in that, The method includes: S1. Arrange the pressure signals of the hydraulic piston pump according to the spiral structure, and perform interpolation processing on the arranged signals to obtain the optimized pressure image of the hydraulic piston pump. S2. Perform visual enhancement processing on the optimized pressure image to obtain the enhanced pressure image of the hydraulic plunger pump; S3. Analyze the multi-scale features in the enhanced pressure image to obtain the global feature parameters of the image brightness statistical features in the hydraulic piston pump, including: Using visual information at different scales in the enhanced stress image, an image pyramid structure of the enhanced stress image is constructed; In the image pyramid structure, the brightness statistics of the brightness distribution characteristics in the enhanced pressure image are extracted; Cross-scale correlation analysis is performed on the brightness statistics to obtain the correlation matrix of the enhanced stress image; Based on the correlation matrix, tensor synthesis is performed on the overall operating state of the hydraulic piston pump to obtain the brightness feature vector of the hydraulic piston pump. By normalizing the different dimensions of the brightness feature vector, the global feature parameters of the hydraulic plunger pump are obtained; S4. Perform local texture feature analysis on the enhanced pressure image to obtain local feature parameters of the image gradient change in the hydraulic plunger pump; S5. Using the global feature parameters as the upper framework and the local feature parameters as the lower benchmark, construct the overall feature parameters of the operating state of the hydraulic piston pump. S6. Perform multi-level comparisons between the overall feature parameters and preset standard feature thresholds, and determine the fault state of the hydraulic piston pump based on the comparison results, including: The preset standard feature thresholds are divided into macro-consistency thresholds and micro-difference thresholds according to feature level; The global feature components in the overall feature parameters are compared with the macro-consistency threshold at the first level to obtain the consistency evaluation result of the global feature components. The local feature components in the overall feature parameters are compared with the micro-difference threshold at a second level to obtain the difference evaluation result of the local feature components; Cross-validate the consistency assessment results and the difference assessment results to obtain the comprehensive diagnostic decision vector of the overall feature parameters; Based on the pointing characteristics of the comprehensive diagnostic decision vector, the fault state type of the hydraulic piston pump is determined.

2. The intelligent fault diagnosis method for hydraulic piston pumps based on pressure signals as described in claim 1, characterized in that, The process of arranging the pressure signals of the hydraulic piston pump according to a helical structure and then interpolating the arranged signals to obtain an optimized pressure image of the hydraulic piston pump includes: The pressure signal of the hydraulic piston pump is collected during operation. The starting point of the spiral structure is taken as the center of the image plane, and the unfolding direction of the spiral path in the spiral structure is determined by the equal angle increment of the spiral structure. The pressure signals are distributed to equal-angle positions along the spiral path in chronological order to obtain a spiral distribution of the pressure signals. In the spiral distribution, the pressure signal corresponding to the equiangular position is mapped to the image grayscale value to obtain the initial image of the spiral grayscale distribution in the hydraulic plunger pump; Identify blank areas in the initial image caused by path gaps in the spiral path; Fill in the missing pixel values ​​in the blank areas of the image to obtain an optimized pressure image of the hydraulic piston pump.

3. The intelligent fault diagnosis method for hydraulic piston pumps based on pressure signals as described in claim 2, characterized in that, In the spiral distribution, mapping the pressure signals corresponding to the equiangular positions into image grayscale values ​​to obtain an initial image of the spiral grayscale distribution in the hydraulic piston pump includes: Using the time-domain characteristics of the pressure signal and the combined characterization of the position and amplitude characteristics in the spiral distribution, a pressure signal characterization space for the spiral distribution is constructed. In the pressure signal characterization space, the pressure signals at the isoangular positions are mapped onto a preset grayscale base to obtain grayscale characterization values ​​with visual discrimination at the isoangular positions. According to the topological structure of the spiral path, the grayscale representation value is projected onto the image coordinate position of the image plane; Along the spiral path, the grayscale representation value is reconstructed using the visual saliency distribution of the pressure signal to obtain an initial image of the spiral grayscale distribution.

4. The intelligent fault diagnosis method for hydraulic piston pumps based on pressure signals as described in claim 1, characterized in that, The step of performing visual enhancement processing on the optimized pressure image to obtain the enhanced pressure image of the hydraulic piston pump includes: By analyzing the visual characteristics of the optimized stress image, structural feature information of the optimized stress image at different scales is obtained. The structural feature information is reconstructed to obtain the structure-aware enhanced map of the optimized pressure image; Based on the structure-aware enhanced map, adaptive contrast enhancement is performed on the optimized pressure image to obtain the visual mode of the operating status related to the hydraulic piston pump in the optimized pressure image. The visual mode of the running state is subjected to detail restoration processing to obtain the essential feature texture information of the pressure signal in the optimized pressure image; The essential feature texture information is input into the optimized pressure image to obtain the enhanced pressure image of the hydraulic plunger pump.

5. The intelligent fault diagnosis method for hydraulic piston pumps based on pressure signals as described in claim 1, characterized in that, The step of performing local texture feature analysis on the enhanced pressure image to obtain local feature parameters of the image gradient changes in the hydraulic piston pump includes: The local texture features in the enhanced stress image are scanned region by region to obtain the spatial variation pattern of the enhanced stress image; Based on the changing trend of the spatial variation pattern, the directional gradient distribution of texture orientation characteristics in the enhanced stress image is determined; Regional aggregation of gradient change descriptors in the directional gradient distribution yields a texture primitive feature map of the enhanced stress image. The gradient statistics in the texture primitive feature map are used as local feature parameters of the enhanced stress image.

6. The intelligent fault diagnosis method for a hydraulic piston pump based on pressure signals as described in claim 1, characterized in that, The construction of overall characteristic parameters of the operating state of the hydraulic piston pump, using the global characteristic parameters as the upper-level framework and the local characteristic parameters as the lower-level benchmark, includes: The global feature parameters are configured as the top-level representation of the hydraulic piston pump, and the local feature parameters are configured as the bottom-level representation of the hydraulic piston pump. The spatial distribution of the bottom-level representation and the feature dimensions of the top-level representation are synchronously aligned to obtain the preliminary feature relationship of the hydraulic piston pump. Verify the feature integrity in the preliminary feature relationship to obtain the feature association relationship of the hydraulic piston pump; Based on the aforementioned feature association, the local feature parameters are embedded as nodes into the overall framework of the global feature parameters to obtain the overall feature parameters of the operating state of the hydraulic piston pump.

7. The intelligent fault diagnosis method for a hydraulic piston pump based on pressure signals as described in claim 6, characterized in that, The calculation formula for the preliminary characteristic relationship is as follows: ； In the formula, For the first The first underlying representation and the first The initial relationship strength between the top-level representations The local feature parameter is the first one. The spatial location of the first 1 eigenvalue, The first of the global feature parameters 1 eigenvalue, It is a preset small positive number.

8. A hydraulic piston pump intelligent fault diagnosis system based on pressure signals, used to implement the hydraulic piston pump intelligent fault diagnosis method based on pressure signals as described in claim 1, the system comprising: An image construction module is used to arrange the pressure signals of the hydraulic piston pump in a spiral structure and to perform interpolation processing on the arranged signals to obtain an optimized pressure image of the hydraulic piston pump. An image enhancement module is used to perform visual enhancement processing on the optimized pressure image to obtain an enhanced pressure image of the hydraulic plunger pump; The global feature module is used to analyze the multi-scale features in the enhanced pressure image to obtain the global feature parameters of the image brightness statistical features in the hydraulic piston pump, including: Using visual information at different scales in the enhanced stress image, an image pyramid structure of the enhanced stress image is constructed; In the image pyramid structure, the brightness statistics of the brightness distribution characteristics in the enhanced pressure image are extracted; Cross-scale correlation analysis is performed on the brightness statistics to obtain the correlation matrix of the enhanced stress image; Based on the correlation matrix, tensor synthesis is performed on the overall operating state of the hydraulic piston pump to obtain the brightness feature vector of the hydraulic piston pump. By normalizing the different dimensions of the brightness feature vector, the global feature parameters of the hydraulic plunger pump are obtained; The local feature module is used to perform local texture feature analysis on the enhanced pressure image to obtain local feature parameters of the image gradient change in the hydraulic plunger pump; The overall feature module is used to construct the overall feature parameters of the operating state of the hydraulic piston pump, with the global feature parameters as the upper-level framework and the local feature parameters as the lower-level reference. The fault diagnosis module is used to perform multi-level comparisons of the overall characteristic parameters with preset standard characteristic thresholds, and determine the fault state of the hydraulic piston pump based on the comparison results, including: The preset standard feature thresholds are divided into macro-consistency thresholds and micro-difference thresholds according to feature level; The global feature components in the overall feature parameters are compared with the macro-consistency threshold at the first level to obtain the consistency evaluation result of the global feature components. The local feature components in the overall feature parameters are compared with the micro-difference threshold at a second level to obtain the difference evaluation result of the local feature components; Cross-validate the consistency assessment results and the difference assessment results to obtain the comprehensive diagnostic decision vector of the overall feature parameters; Based on the pointing characteristics of the comprehensive diagnostic decision vector, the fault state type of the hydraulic piston pump is determined.

Images