Fusion decision-based equipment wear stage identification method and system

Abstract

The invention belongs to the technical field of mechanical equipment wear state monitoring. The method comprises the following steps of: extracting wear rate characteristics, wear degree characteristics, cutting index characteristics, fatigue index characteristics and oxidation index characteristics in an abrasive particle sample image, and respectively clustering to obtain five groups of discrete stage labels; carrying out normalization processing on the five groups of discrete stage labels to obtain a normalized data matrix; according to the data matrix, weights corresponding to all the features are calculated in multiple modes, and the average value of the weights calculated in the multiple modes serves as the final weight of all the features; according to the final weight of each feature and the normalized data matrix, calculating a comprehensive score sequence of all the abrasive particle sample images; performing clustering according to the comprehensive score sequence to obtain a plurality of optimal clustering centers, and determining a current wear stage according to the optimal clustering centers; and more comprehensive and more accurate division of the wear stages is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical equipment wear status monitoring, and in particular to a method and system for identifying equipment wear stages based on fusion decision-making. Background Art

[0002] The statements in this section merely provide background art related to the present invention and do not necessarily constitute prior art.

[0003] Wear-related failures account for over 75% of all mechanical equipment failures. Wear of mechanical components is a dynamic and gradual process, progressing from minor wear to normal wear, and finally to severe wear and failure. This process is typically divided into several stages based on changes in wear information: the run-in period, the stabilization period, the acceleration period, and the failure period. Lubricating oil is a key element in ensuring the healthy and stable operation of mechanical equipment, carrying wear particles generated by various components. The characteristic information of abrasive particles in the lubricating oil can be used to determine the changes in component wear stages in real time, thereby monitoring, analyzing, and predicting the wear condition of the equipment.

[0004] Early wear monitoring mainly relies on ferroscopes or particle counters to extract information on abrasive concentration and size, and to reflect "quantitative changes" by calculating the wear rate and wear degree (such as the wear particle coverage area index IPCA). Subsequently, researchers proposed extended indicators such as the total wear debris concentration index (TIDC) and the cumulative dimensionless wear debris concentration index (ADIDC) based on IPCA to make up for the limitations of a single indicator in identifying the pitting stage of gears. In addition, the number of large abrasive particles (LPQ) and its size gradient changes were also introduced to enhance sensitivity to the later wear stages. However, whether it is the wear rate or the wear degree, these macroscopic quantitative indicators are still difficult to reveal the qualitative changes in the wear mechanism when the values ​​are the same or similar, resulting in the "same quantity, different quality" phenomenon, making it difficult to accurately distinguish different wear stages.

[0005] With the development of microscopic imaging and image processing technology, microscopic features such as the shape, texture, and color of wear particles can be efficiently collected and quantitatively analyzed, providing new ideas for revealing wear mechanisms. However, while traditional macroscopic quantitative indicators such as wear rate and wear degree can reflect the "quantitative change" of wear, they cannot reveal the "qualitative change" driven by different mechanisms when the values ​​are similar, and are prone to "same quantity, different quality" blind spots. Although microscopic imaging technology has enabled the efficient acquisition of microscopic mechanism features such as the shape, texture, and color of wear particles, existing research is mostly limited to wear particle classification and fails to systematically apply mechanism information to the wear stage classification, resulting in insufficient overall identification accuracy. Summary of the Invention

[0006] In order to address the shortcomings of the existing technology, the present invention provides a method and system for identifying equipment wear stages based on fusion decision-making, which realizes a more comprehensive and accurate division of wear stages. It not only avoids the limitation of "same quantity and different quality" of traditional wear indicators, but also provides an explainable mechanism support for online monitoring and fault warning.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for identifying equipment wear stages based on fusion decision-making.

[0008] A method for identifying equipment wear stages based on fusion decision-making includes the following steps: Extract wear rate features, wear degree features, cutting index features, fatigue index features and oxidation index features from multiple abrasive sample images; The wear rate feature, wear degree feature, cutting index feature, fatigue index feature and oxidation index feature are clustered separately to obtain five groups of discrete stage labels; Normalize the five sets of discrete stage labels to obtain a normalized data matrix; According to the data matrix, weights corresponding to the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature are calculated respectively in a variety of ways, and the average of the weights calculated in various ways is the final weight of each feature; According to the final weight of each feature and the normalized data matrix, the comprehensive score sequence of all wear particle sample images is calculated; Clustering is performed according to the comprehensive score sequence to obtain multiple optimal cluster centers, and the current wear stage is determined according to the optimal cluster centers.

[0009] In one implementation of the first aspect of the present invention, before extracting the wear rate features, wear degree features, cutting index features, fatigue index features, and oxidation index features from the plurality of abrasive sample images, a preprocessing process is further included, including: The wear particle sample image is processed in sequence by grayscale processing, background subtraction processing, binarization processing and morphological processing.

[0010] In one implementation of the first aspect of the present invention, the wear rate characteristic is the actual area of ​​the abrasive region, the wear degree characteristic is the number of abrasives whose equivalent circle diameter exceeds the set threshold, the cutting index characteristic is the proportion of cutting abrasives, the fatigue index characteristic is the proportion of fatigue abrasives, and the oxidation index characteristic is the proportion of oxidized abrasives.

[0011] As a further limitation of the first aspect of the present invention, the area, equivalent circular diameter, aspect ratio, roundness, and curvature standard deviation of each abrasive particle are calculated; When the equivalent circle diameter is greater than or equal to the first set threshold, and the roundness is greater than or equal to the second set threshold, it is a normal abrasive particle, otherwise it goes to the next step of judgment; When the curvature standard deviation is less than or equal to the third set threshold, and the roundness is greater than or equal to the fourth set threshold, it is an oxidized abrasive particle, otherwise it proceeds to the next step of judgment; When the aspect ratio is greater than or equal to the fifth set threshold and the roundness is less than or equal to the second set threshold, it is a cutting abrasive particle; otherwise, it is a fatigue sliding abrasive particle.

[0012] In an implementation of the first aspect of the present invention, the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature are clustered using a Mean-shift algorithm to divide the wear stages.

[0013] In one implementation of the first aspect of the present invention, the entropy weight method, the CRITIC method, and the coefficient of variation method are used to calculate the weights corresponding to the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature, respectively; The final weight of the wear rate feature is the mean of the entropy weight, CRITIC weight, and coefficient of variation weight of the wear rate feature; The final weight of the wear degree feature is the mean of the entropy weight, CRITIC weight, and coefficient of variation weight of the wear degree feature; The final weight of the cutting index feature is the mean of the entropy weight, CRITIC weight, and coefficient of variation weight of the cutting index feature; The final weight of the fatigue index feature is the mean of the entropy weight, CRITIC weight, and coefficient of variation weight of the fatigue index feature; The final weight of the oxidation index feature is the average of the entropy weight, CRITIC weight, and coefficient of variation weight of the oxidation index feature.

[0014] In one implementation of the first aspect of the present invention, K-means clustering is performed based on the comprehensive score sequence, the optimal number of clusters K is calculated, K cluster centers are obtained, the cluster centers are sorted, and adjacent center points are taken as stage boundaries to obtain the current wear stage; The current wear stages include: running-in wear stage, stable wear stage, oxidation wear stage, transition stage, accelerated wear stage and failure stage.

[0015] In a second aspect, the present invention provides an equipment wear stage identification system based on fusion decision-making.

[0016] An equipment wear stage identification system based on fusion decision-making, comprising: The feature extraction unit is configured to extract a wear rate feature, a wear degree feature, a cutting index feature, a fatigue index feature, and an oxidation index feature from a plurality of abrasive sample images; The feature clustering unit is configured to: cluster the wear rate feature, wear degree feature, cutting index feature, fatigue index feature, and oxidation index feature respectively to obtain five groups of discrete stage labels; The data normalization unit is configured to: normalize the five sets of discrete stage labels to obtain a normalized data matrix; a weight allocation unit configured to: calculate weights corresponding to the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature respectively according to the data matrix in a plurality of ways, and the average of the weights calculated in the various ways is the final weight of each feature; The comprehensive score calculation unit is configured to: calculate a comprehensive score sequence of all wear particle sample images according to the final weight of each feature and the normalized data matrix; The wear stage determination unit is configured to: perform clustering according to the comprehensive score sequence to obtain multiple optimal cluster centers, and determine the current wear stage according to the optimal cluster centers.

[0017] In a third aspect, the present invention provides a computer device comprising: a processor and a computer-readable storage medium; a processor adapted to execute a computer program; A computer-readable storage medium having a computer program stored therein, wherein when the computer program is executed by the processor, the method for identifying equipment wear stages based on fusion decision-making as described in the first aspect of the present invention is implemented.

[0018] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program is suitable for being loaded by a processor and executing the equipment wear stage identification method based on fusion decision as described in the first aspect of the present invention.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The present invention innovatively proposes a method for identifying equipment wear stages based on fusion decision-making. By introducing and comprehensively analyzing the three major wear mechanism indicators of cutting, fatigue and oxidation, it can not only clearly distinguish the dynamic characteristics of the running-in, stable, accelerated and severe wear stages, but also deeply reveal the interactive evolution of each mechanism on the time axis: the peak of the initial cutting abrasive particles and their rapid decay map the formation of the surface lubricating film during the running-in period, the occasional oscillation of the proportion of oxidation wear and fatigue abrasive particles in the stable stage indicates a precursor to instability, the synchronous fermentation of multiple mechanisms in the acceleration period reveals the combined effect of crack initiation and coordinated cutting of hard particles, and the outbreak of multiple indicators in the severe wear period directly points to the critical state of macroscopic destruction. Such a mechanism-based wear stage division method not only significantly improves the accuracy and robustness of stage division, but also provides physical explainability and decision support for online early warning, life prediction and fault location.

[0020] The present invention organically integrates macro-quantitative indicators (wear rate, wear degree) and micro-mechanism indicators (cutting index, fatigue index, oxidation index), and captures "quantitative changes" and "qualitative mechanism changes" simultaneously under the same analysis framework. Compared with the method that relies only on single or dual macro-indicators, it can reveal the essence of material failure from the micro-mechanism level, and completely solve the stage confusion caused by the traditional "same quantity, different quality"; the Mean-Shift algorithm is used to cluster each indicator separately, and weighted integration is performed under the entropy weight method-CRITIC method-coefficient of variation method (EWM-CRITIC-DWM) multi-criteria decision fusion framework, which significantly improves the recognition accuracy of key stages such as running-in, stabilization, acceleration and failure; at the same time, by dynamically analyzing the weight evolution of each feature in different stages, the stage division results are given a clear physical mechanism explanation, and the traceability of online early warning and diagnosis is enhanced.

[0021] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0023] Figure 1 A schematic diagram of an equipment wear stage identification method based on fusion decision-making provided by an exemplary embodiment of the present invention; Figure 2 A schematic diagram of the principle of online monitoring of wear particles provided by an exemplary embodiment of the present invention; Figure 3 A schematic diagram of a wear particle image preprocessing process according to an exemplary embodiment of the present invention; Figure 4 A schematic diagram of the principle of the Mean shift algorithm provided by an exemplary embodiment of the present invention is shown below. Figure 4 (A) represents the process of step 1. Figure 4 (B) in the figure represents the process of step 2. Figure 4 (C) in the figure represents the process of step 3. Figure 4 (D) in the figure represents the process of step 4. Figure 4 (E) in the figure represents the process of step 5. Figure 4 (F) in the figure represents the process of step 6; Figure 5 A schematic diagram of the changes in the five major characteristics during the wear process provided by an exemplary embodiment of the present invention, wherein PQ-O represents the proportion of oxidized abrasive particles, PQ-C represents the proportion of cutting abrasive particles, PQ-F represents the proportion of fatigue abrasive particles, and IPCA represents the abrasive coverage area index. Represents the proportion of large abrasive particles; Figure 6 A schematic diagram of a comprehensive phase division result provided by an exemplary embodiment of the present invention; Figure 7 A schematic diagram of the division of wear stages based on wear rate provided in accordance with an exemplary embodiment of the present invention; Figure 8 A schematic diagram of dividing wear stages based on wear degree provided in accordance with an exemplary embodiment of the present invention; Figure 9 Schematic diagram of wear particle images at different times provided by an exemplary embodiment of the present invention. Figure 9 (a) represents the schematic diagram of the abrasive particles at 20 minutes. Figure 9 (b) in the figure represents the schematic diagram of the wear particles at 70 minutes. Figure 9 (c) in the figure represents the schematic diagram of the wear particles at 370 minutes. Figure 9 (d) in the figure represents the schematic diagram of wear particles at 480 minutes; Figure 10 A schematic diagram of an equipment wear stage identification system based on fusion decision-making provided by an exemplary embodiment of the present invention; Figure 11 A schematic diagram of a computer device provided for an exemplary embodiment of the present invention; Among them, 1. rolling element; 2. abrasive particles; 3. peristaltic pump; 4. first image sensor; 5. second image sensor; 6. abrasive particle ring image; 7. dispersed abrasive particle image. DETAILED DESCRIPTION

[0024] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0025] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0026] As mentioned in the background, existing wear evaluation metrics, such as wear rate and wear degree, can only quantitatively reflect the "quantitative change" in wear degree, but fail to reveal the "qualitative change in the mechanism" during the wear process. Consequently, they are unable to meet the requirements for high-precision wear stage identification. In light of this, this implementation proposes a method for equipment wear stage identification based on fusion decision-making, which includes the following steps: S101: extracting wear rate features, wear degree features, cutting index features, fatigue index features, and oxidation index features from multiple abrasive sample images; S102: clustering the wear rate feature, wear degree feature, cutting index feature, fatigue index feature, and oxidation index feature respectively to obtain five groups of discrete stage labels; S103: normalizing the five groups of discrete stage labels to obtain a normalized data matrix; S104: Calculating weights corresponding to the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature using a plurality of methods according to the data matrix, and averaging the weights calculated using the various methods as the final weight of each feature; S105: Calculating a comprehensive score sequence of all wear particle sample images based on the final weight of each feature and the normalized data matrix; S106: Clustering is performed according to the comprehensive score sequence to obtain multiple optimal cluster centers, and the current wear stage is determined according to the optimal cluster centers.

[0027] In S101 of this implementation, specifically, the following steps are included: Wear particle images can provide rich wear particle characteristics, such as quantity, size, shape, texture, etc., and thus can characterize different wear indicators. In order to obtain wear particle images during component wear, a wear particle image sensor is used to obtain wear particle images, which can simultaneously obtain wear particle characteristics reflecting the number and morphology of wear particles.

[0028] like Figure 2The figure shows the specific equipment used, including rolling elements 1, abrasive particles 2, a peristaltic pump 3, a first image sensor 4, and a second image sensor 5. Ultimately, abrasive particle ring images 6 and dispersed abrasive particle images 7 are obtained. Abrasive particles 2 generated by bearing wear are carried by lubricating oil and pumped into the sensor's flow channel at a flow rate of 120 ml / min by peristaltic pump 3. They flow through its deposition area, which consists of an oil pool and a glass observation window. A high-gradient magnetic field is formed in the tiny gap between the inner and outer annular electromagnets, attracting the passing ferromagnetic abrasive particles and forming an abrasive particle ring. CMOS camera 1 (i.e., first image sensor 4) captures the abrasive particle ring image. Subsequently, the electromagnets are disconnected, allowing the lubricating oil to disperse the abrasive particle ring. At this point, CMOS camera 2 (i.e., second image sensor 5) captures the dynamic dispersed abrasive particle image. Connecting the sensors to the equipment's lubrication system enables real-time acquisition of abrasive particle images during operation. The captured abrasive particle images (including abrasive particle ring image 6 and dispersed abrasive particle image 7) are transmitted to a computer for subsequent processing and analysis.

[0029] In order to extract the features of wear particles in the image, it is necessary to eliminate the interference of other factors in the image and preprocess the wear particle image. The processing flow is as follows: Figure 3 As shown in the figure. After the wear particles are imaged, the RGB channels of the color wear particle image contain too much redundant information, which is not conducive to subsequent analysis. In particular, the morphology and concentration characteristics of the wear particles are not highly correlated with the color information. Therefore, grayscale conversion can be used to reduce the data dimension, highlight the brightness differences of the wear particles, and improve processing efficiency. Use the weighted average method to convert the color image into a grayscale image: (1); in, 、 、 The weights represent the red, green, and blue channel intensities, respectively. These weights reflect the human eye's sensitivity to different colors. After grayscale processing, each pixel in the image is represented by a single grayscale value (0-255), making brighter areas easier to distinguish.

[0030] Background artifacts in wear particle images (such as lighting, equipment reflections, and lubricant flow) can interfere with wear particle detection, leading to noise and misjudgment. Background subtraction helps enhance the saliency of wear particle regions, enabling more accurate feature extraction. By performing a differential operation between the wear particle image and the background image, only the wear particle region is retained, eliminating interference from background lighting and reflections.

[0031] Since the wear particles themselves appear dark gray, their grayscale is similar to that of the wear particle-free area after background subtraction, resulting in unclear wear particle boundaries and difficulty in directly analyzing their morphological features. Therefore, a global threshold segmentation is performed on the image, using the Otsu algorithm to automatically calculate the global optimal threshold and divide the image into foreground (wear particles) and background: (2); in, is the grayscale value, The gray value threshold is calculated, and the accurate segmentation of the abrasive area is achieved through threshold segmentation. The background is completely removed, the clarity of the abrasive outline is improved, and it is convenient for further extraction and statistics of particle size, shape and other features.

[0032] In order to improve the accuracy of feature extraction, further morphological operations are required on the above-mentioned preliminary preprocessed dispersed abrasive image 7. Due to illumination or binarization processing, there are holes or discontinuities inside the abrasive, which affect the calculation of the abrasive area and the extraction of shape features. Dilation and corrosion operations are used to fill the holes and restore the integrity of the pixels in the abrasive area. Then, according to the difference in grayscale value between the abrasive boundary and the background, the outline of each abrasive is extracted and numbered, providing a basis for subsequent shape feature analysis.

[0033] In S102 of this implementation, specifically, the following steps are included: Wear rate, wear degree, and wear mechanism together constitute the "process-state-essence" trinity wear stage identification system, providing complete information support for accurate stage division. At the same time, these three types of indicators also need to be described by specific wear particle characteristics, as follows: The index of wear particle coverage area (IPCA) characterizes the wear rate. The IPCA can be viewed as an optical projection of the wear rate. By quantifying the wear particle deposition area, it transforms the abstract "material loss rate" into an observable "spatial coverage change." After image preprocessing, all connected regions within the wear particle ring image—that is, areas where foreground pixels are connected—are identified. By counting all foreground pixels, the foreground pixel coverage area is calculated.

[0034] (3); Among them, FPCA is the foreground pixel coverage area in the wear particle ring image, is the number of foreground pixels in the wear particle ring image. When the grayscale value of a pixel is 255, When the gray value is 0, is 0.

[0035] Convert the pixel area to the actual area of ​​the wear particle. If the pixel resolution of the image is , then the actual area of ​​the abrasive region can be expressed as: (4).

[0036] A significant increase in the number of large abrasive particles indicates that the wear has entered the accelerated wear stage. Therefore, the proportion of large abrasive particles is used as an indicator to evaluate the degree of wear. Abrasive particles with an equivalent circle diameter of more than 30 μm are considered large abrasive particles, and their proportion is: (5); in, is the proportion of large abrasive particles, For large abrasive particles, is the total number of abrasive particles. The calculation method of the equivalent circle diameter will be introduced below.

[0037] In addition to normal wear, the wear mechanism can be divided into four types: adhesive wear, fatigue wear, cutting wear and oxidation wear.

[0038] Adhesive wear often results from lubricant degradation due to contamination, or from the breakdown of the lubricating film, leading to direct metal-to-metal contact, due to insufficient lubrication or extreme conditions such as high load, high speed, and high temperature. Adhesion occurs when friction pairs slide against each other, and the resulting adhesion points shear and fracture. The sheared material either breaks off, forming wear debris, or migrates to another surface. Severe adhesion can tear the surface, producing flat, coarse fragments, often with scratches aligned with the sliding direction.

[0039] When two contact surfaces are in pure rolling or rolling and sliding friction, the small contact area will be slightly deformed due to the load. After multiple stress cycles, the surface will be damaged at a certain point to form fatigue fragmentation, and small pieces of material will peel off in the deformation area. As the cracks gradually expand, large fatigue metal particles will be produced, which can be up to hundreds of microns in length, causing fatigue wear. Once it starts, it will continue to grow until the surface completely fails and ends in catastrophic damage.

[0040] Cutting wear is generally caused by broken or misaligned machine parts, or by abrasive particles trapped in the lubricant. When machine parts are misaligned, the harder friction surface will penetrate the softer surface, causing cutting, similar to the chips produced by machine tool processing. These chips are generally 2-5μm wide and 25-100μm long. When hard particles are embedded in a soft surface, they cut into the soft surface, creating chips that are often small, generally only a few microns long, but still slender.

[0041] Oxidative wear is a serious problem for mechanical equipment. It is caused by the presence of acidic oxidation products and moisture in the oil, or the formation of oxides on the metal surface when exposed to air. These chemical reactions damage the surface of the machine and produce very fine spherical particles, which form welds and pitting. This highly destructive wear often leads to secondary wear such as adhesion and fatigue wear.

[0042] The wear mechanism is the fundamental cause of abrasive particle generation, directly determining the type of abrasive particles and their characteristics, such as shape and size. Conversely, differences in abrasive particle characteristics can be used to classify abrasive particles, thereby characterizing the wear mechanism and inferring the generation mechanism of abnormal abrasive particles from their characteristics. Therefore, by extracting the current abrasive particle morphology and calculating the proportion of each abrasive particle type (normal abrasive particles, fatigue abrasive particles, cutting abrasive particles, and oxidation abrasive particles), the wear mechanism at a specific moment can be quantified.

[0043] Abrasive particle area A: Similar to the acquisition of IPCA, the abrasive particle area is obtained by calculating the number of pixels occupied by the abrasive particles.

[0044] (6); in, is the area of ​​a single abrasive particle, is the number of pixels of a single wear particle in the dispersed wear particle image. When the grayscale value of a pixel is 255, When the gray value is 0, is 0.

[0045] Equivalent circle diameter D: The diameter of a circle with the same area as the abrasive grain.

[0046] (7); in, is the equivalent diameter of a single abrasive particle.

[0047] Aspect Ratio (AR): The ratio of the abrasive particle's length, L, to its width, W. Based on the particle's outline, its minimum enclosing rectangle (LR) is calculated to obtain the height and width of its bounding box. The larger value represents the particle's length, L, and the smaller value represents its width, W. When AR ≈ 1, the abrasive particle is approximately square; when AR > 1, the abrasive particle is elongated or flat.

[0048] (8); Roundness R is an important indicator that describes the degree to which the abrasive particle shape is close to a circle. The roundness of an ideal circle is 1, indicating that the abrasive particle is completely close to a perfect circle. A roundness less than 1 indicates that the abrasive particle shape deviates from a circle, and the closer it is to 0, the more irregular the shape.

[0049] (9); Where P is the circumference of the abrasive particle, which can be obtained from the contour of the abrasive particle.

[0050] Curvature standard deviation C: The curvature of a pixel in an abrasive particle is the angle between it and the two consecutive pixels before and after it. The edge contour of a circular abrasive particle is relatively smooth, so the curvature changes more gently. However, due to the existence of concave or convex corners, the curvature distribution of abrasive particles of other shapes is uneven, and even obvious peaks may appear. By calculating the standard deviation of the curvature of each abrasive particle, circular abrasive particles can be distinguished from other abrasive particles.

[0051] In order to identify different types of abrasive particles based on their characteristics, a three-level discriminant tree model for four types of wear particle classification was established. All abrasive particles were divided into four types: normal abrasive particles, oxidized abrasive particles, cutting abrasive particles, and fatigue sliding abrasive particles. Specifically, the following are included: When the equivalent circle diameter Greater than or equal to the first set threshold , and the roundness Greater than or equal to the second set threshold If , it is normal abrasive particles, otherwise it goes to the next step of judgment; When the standard deviation of curvature Less than or equal to the third set threshold , and the roundness Greater than or equal to the fourth set threshold If , it is oxidized abrasive particles, otherwise it goes to the next step of judgment; When the aspect ratio Greater than or equal to the fifth set threshold , and the roundness Less than or equal to the second set threshold When , it is a cutting abrasive particle, otherwise, it is a fatigue sliding abrasive particle.

[0052] In this implementation, since the diameter of normal abrasive particles is generally small, and the roundness is quite different from that of oxidized abrasive particles of the same small particle size, the equivalent circle diameter and roundness of normal abrasive particles and other abrasive particles can be used to improve the accuracy of the judgment of oxidized abrasive particles by combining the curvature standard deviation and roundness. Compared with fatigue abrasive particles, cutting abrasive particles have a larger aspect ratio and smaller roundness. Through this classification rule, the four types of abrasive particles can be well identified.

[0053] In order to accurately describe the wear trend of components, the Mean-Shift algorithm is now used to divide the wear stages. Mean-Shift is a non-parametric clustering algorithm used to identify high-density areas in data distribution, especially suitable for nonlinear distribution data. Its core idea is to achieve data clustering by iteratively adjusting the data points to move in the direction of the probability density function gradient, and finally converge to the local density peak. The algorithm principle is as follows Figure 4 shown.

[0054] Step 1: If Figure 4 As shown in (A), the first sampling point is selected as the center point , determine an area with radius r, called the neighborhood, and the points in the neighborhood belong to the same cluster , calculate the distance from all points in the neighborhood to the center point The vector of is superimposed to obtain the vector shift, which represents the gradient direction of the data distribution. The red dots represent unlabeled and the gray dots represent labeled. Step 2: If Figure 4 As shown in (B), the center point Move along the sum vector shift to get the second center point (Right now ), continue to repeat steps 1 and 2 until |shift| is less than the convergence value , then execute step 3, where the red dot indicates unmarked and the gray dot indicates marked; Step 3: If Figure 4 As shown in (C), check the current cluster If the distance to other existing clusters is less than the threshold d, the two clusters are merged; if there is no cluster that meets the merging conditions, a new cluster is created, in which the red dots represent unmarked and the gray dots represent marked; Step 4: If Figure 4 As shown in (D), steps 1 to 3 are repeated for all unlabeled data points until all points are labeled. The clusters are shown in the figure. The corresponding center point , where red dots represent unmarked and gray dots represent marked; Step 5: Figure 4 As shown in (E), the frequency of each point being accessed by different clusters is calculated, and the cluster with the largest access frequency is taken as the final cluster to which it belongs. The clusters are shown in the figure. and clusters , where red dots represent unmarked and gray dots represent marked; Step 6: Figure 4 As shown in (F) in the figure, the division results of all data points are obtained (C1, C2, C3 and C4), where the orange points represent the points in cluster C1, the blue points represent the points in cluster C2, the green points represent the points in cluster C3, and the red points represent the points in cluster C4.

[0055] Mean-shift clustering is performed on the five quantitative features of the three types of wear indicators to obtain five groups of discrete stage labels: (10); in, 、 、 、 and They represent the first group of discrete phase labels, the second group of discrete phase labels, the third group of discrete phase labels, the fourth group of discrete phase labels, and the fifth group of discrete phase labels respectively; Indicates the The sample in Features The stage number to which it belongs (which may be different from the total number of stages for other features); K is the number of comprehensive stages. This paper focuses on how to achieve comprehensive stage division of the five features based on the existing five sets of single-feature stage labels, that is, to obtain a global stage label set.

[0056] To mitigate the influence of subjective experience on the final stage classification, this paper uses three classic objective weighting methods (entropy weighting, CRITIC method, and coefficient of variation method) to automatically calculate weights for the five features. The entropy weighting method focuses on information entropy differences, the CRITIC method considers both contrast and redundancy, and the coefficient of variation method emphasizes dispersion. By averaging these three weightings, we can leverage the core strengths of each method while mitigating the limitations of each (such as the entropy weighting method's neglect of correlation, the CRITIC method's sensitivity to extreme values, and the coefficient of variation method's overweighting of indicators with low mean values). This results in a more robust, comprehensive, and subjective weighting scheme, ensuring the objectivity and interpretability of the final wear stage classification results.

[0057] The following process is based on the data matrix after the five features are normalized , is the normalized value of the i-th sample on the j-th feature, between [0,1].

[0058] (1) Entropy weight method.

[0059] First calculate the weight of the jth feature on the i-th sample: (11); in, represents the normalized value of the kth sample on the jth feature, .

[0060] Then calculate the information entropy of each feature : (12); Among them, when Time Definition , is the number of samples.

[0061] Calculating information redundancy : (13).

[0062] Calculate entropy weight after normalization : (14); in, Indicates the The information redundancy of the features, .

[0063] (2) CRITIC method.

[0064] Based on the normalized matrix Z, calculate the standard deviation of the j-th feature: (15); in, For the The standard deviation of the features, For the The average value of the features.

[0065] Calculate the Pearson correlation coefficient between pairs of features : (16); in, For the The average value of the features.

[0066] calculate The amount of information of a feature : (17).

[0067] Calculate CRITIC weight after normalization : (18).

[0068] in, Representative The amount of information of a feature.

[0069] (3) Coefficient of variation method.

[0070] Based on formula (15) The mean of the features and standard deviation , the coefficient of variation can be calculated : (19); Calculate DWM weight after normalization : (20); in, For the The coefficient of variation of a feature.

[0071] In order to take into account the advantages of each weighting method, a simple averaging method is used for weight fusion, and the final comprehensive weight vector is: (twenty one); Among them, formula (21) satisfies , given the comprehensive weight and the normalized data matrix Z, calculate the The comprehensive score of the samples: (twenty two).

[0072] The comprehensive score sequence obtained based on formula (22) , the comprehensive score sequence Perform K-means clustering, calculate the optimal number of clusters K, and obtain K cluster centers. Sorting the cluster centers, take the adjacent center points as the stage boundaries to complete the discretization.

[0073] To validate the practicality of the proposed method, a series of wear tests were conducted on bearings using an accelerated life test platform. Wear particles were sampled using a wear particle image sensor, generating wear particle images. Through image processing and feature extraction, three indicators—wear rate, wear severity, and wear mechanism—were derived and integrated to assess the bearing's wear stage. Furthermore, the combined information from these three indicators revealed key wear mechanisms at different wear stages.

[0074] This study is in Figure 2 Bearing roller wear experiments were conducted on the accelerated bearing life test platform shown in the figure. The test bearings were XTAB-A7204AC angular contact ball bearings, with specific parameters shown in Table 1. The test platform was driven by a variable-frequency speed-controlled motor, and a hydraulic cylinder applied a steady unidirectional force in the radial direction of the test bearing. To provide necessary lubrication and monitor wear in real time, an oil bath lubrication system was installed at the test bearing, with a stable lubricant flow rate of 120 ml / min. Furthermore, a wear particle image sensor was integrated into the lubrication circuit to capture images of wear particles in the lubricant in real time at a 10-minute sampling interval.

[0075] Table 1: Bearing accelerated life test parameters

[0076] In order to obtain the evolution law of the wear stage within the complete life cycle of the bearing in a relatively short period of time, an accelerated full-life test of the bearing is carried out to simulate the long-term service process of the bearing under actual working conditions. By applying higher loads and speeds, the failure process of the bearing is accelerated.

[0077] Before starting the experiment, clean the test bearing and install it on the test platform to ensure that the force loading module is well aligned. Start the variable frequency speed regulating motor, increase the speed to the target speed in stages, and then gradually apply the radial load to the set value through the hydraulic cylinder. After the speed and load are stable for 5 minutes, the start time of the experiment is officially recorded (t=0), and the wear particle image sensor and vibration sensor are started synchronously. During the entire experiment, the speed and radial load remain constant, and the lubricating oil bath system continues to provide stable lubrication. When the vibration signal amplitude measured by the vibration accelerometer installed on the bearing seat continues to exceed hour( is the maximum amplitude of the bearing during stable operation), the bearing is considered to have failed and the test is terminated at this time.

[0078] The experiment lasted 482 minutes, and 48 sets of wear particle images were collected. After image processing, the wear particle features were extracted as shown in Table 2.

[0079] Table 2: Abrasive particle characteristics

[0080] Based on the three-level discriminant tree model established above, the abrasive particles are divided into normal abrasive particles, cutting abrasive particles, oxidized abrasive particles and fatigue abrasive particles according to the shape characteristics of the abrasive particles. The proportion of cutting abrasive particles, oxidized abrasive particles and fatigue abrasive particles is used as the wear mechanism index, IPCA is used as the wear rate index, and the proportion of large abrasive particles is used as the wear degree index. The changes of the five types of characteristics in the wear process are shown in the figure below. Figure 5 As shown in the figure, the wear rate and wear extent trends are generally similar, but in the later stages, the wear extent increases in a step-like manner, better reflecting the severity of the wear. Furthermore, the changing trends of the three types of wear mechanisms differ, which well reflects the changes in the wear mechanism during the wear process.

[0081] Based on the Mean-shift clustering algorithm, the five major features are divided into stages, and the objective weighting method is applied to integrate the division results of the five features to obtain the final stage division results as follows: Figure 5 shown.

[0082] As can be seen, the entire wear process can be divided into six stages, corresponding to the evolving wear mechanisms throughout the operating time. Stage 1 (0-30 minutes) is the run-in wear stage, during which all five characteristics have high values. High-speed meshing and hard inclusion cutting dominate, generating a large amount of cutting abrasive particles. Meanwhile, the surface lubricating film has not yet fully formed, resulting in a rapid increase in abrasive particle concentration followed by a plateau. This typically reflects the combined process of surface micropolishing and lubricating layer construction during the run-in period. Stage 2 (40-280 minutes) is the stable wear stage, during which all five characteristics decrease and then stabilize. This indicates that the surface lubricating film has been largely established, with minor abrasive wear and trace particle scratches predominating at the contact interface. Stage 3 (280-290 minutes) is the oxidative wear stage, with a significant peak in the proportion of oxidized abrasive particles and a slight increase in the remaining four characteristics. This phenomenon can be attributed to the accumulation of temperature at the friction interface or local lubrication failure, which leads to the formation and peeling of surface oxide film, thus becoming dominated by oxidation wear, marking a sudden change in the mechanism from the stable wear period to the accelerated wear period; the fourth stage (300-370 min) is the transition stage or the initial stage of acceleration, during which the proportion of oxidized abrasive particles and other mechanism characteristics temporarily drop back to a low level, which may be due to the shedding and cleaning of surface oxidation products and the local repair of the lubricating film, but at the same time it indicates that the accumulation of secondary damage to the surface caused by fatigue and cutting has begun to appear, laying the foundation for subsequent accelerated wear; the fifth stage The accelerated wear stage (380-450 minutes) shows a rapid increase in wear rate and severity, with a significant simultaneous increase in the proportion of fatigue and oxidation abrasive particles. This reflects the coupled effects of fatigue crack propagation and repeated oxide film flaking. Microscopic pits and crack networks on the material surface intensify, further increasing the equipment wear rate. Stage 6 (470-480 minutes) is the failure stage, with various characteristics rising sharply: the proportion of fatigue and oxidation abrasive particles reaches its highest point, while cutting abrasive particles also increase significantly, but to a relatively small degree, reflecting the complete failure of the lubrication system and large-scale surface shedding. At this point, wear enters a state of uncontrolled wear, indicating that the equipment or component is nearing its lifespan and must be shut down for immediate maintenance or replacement to avoid more serious system failure.

[0083] If only the wear rate or wear degree is used to divide the stages, the mean-shift clustering results are as follows: Figure 6 and Figure 7 shown. pass Figure 7As can be seen, the wear rate is divided into three stages: 0-40 minutes and 370-460 minutes are Stage 1, 50-360 minutes is Stage 2, and 470-480 minutes is Stage 3. Based on the bathtub curve, we know that the first 40 minutes should be the running-in period, with the formation of the surface lubrication layer and the accelerated polishing of the material surface micro-meshing, during which the wear rate has a brief peak. However, because the wear rate during the acceleration period (370-450 minutes) is comparable to that of the running-in period, data from different periods are classified into the same stage. This is because the wear rate during the acceleration period may gradually increase due to sudden fracture or lubrication failure, reaching a similar level as the running-in period. The wear rate indicator only focuses on the wear rate result, and the numerical value of the wear rate alone makes it difficult to distinguish the differences in the wear mechanism. It fails to reflect the different causes of the high wear rate during the running-in and acceleration periods, resulting in overlapping characteristics of the wear rate stages.

[0084] pass Figure 8 It can be seen that the wear degree is divided into three stages, 0-30min and 420-450min are stage 1, 40-420min is stage 2, and 460-480min is stage 3. The same problem exists as the wear rate indicator. The 420-450min that should be the acceleration stage is divided into the running-in period. The reason is that the degree of wear is also the result of the wear mechanism. During the running-in period, during the initial surface engagement and the establishment of the lubricating film, occasional microscopic surface collapse or local hard point shedding will produce a large number of large abrasive particles in a short time. During the acceleration period, large abrasive particles continue to be produced due to lubrication failure, material fracture, etc. As a result, the instantaneous values ​​of the proportion of large abrasive particles in the running-in period and the acceleration period are close, and thus fall into the same stage.

[0085] Since both the wear rate and the wear degree are result mappings guided by the wear mechanism, the stage division based on the combination of these two indicators will also lead to misjudgment due to overlapping stage characteristics.

[0086] It can be seen that the introduction of wear mechanism is necessary in the division of wear stages. It can not only reveal the essential differences in material failure at different stages from a microscopic level, but also effectively make up for the "collision value" blind spot brought by a single macroscopic indicator; through dynamic monitoring and quantitative analysis of major wear mechanism indicators such as cutting, fatigue, and oxidation, a multidimensional feature space can be constructed, thereby fully reflecting the intrinsic mechanism properties of each stage in the clustering model and improving the accuracy and stability of stage identification.

[0087] During the running-in period (0–30 min), the weights of the five types of features are equal, indicating that during the initial polishing and lubricating film formation process, various mechanisms such as cutting, fatigue and oxidation all contribute equally to surface damage; after entering the stable period (40–280 min), the weight of PQ-O (percentage of oxidized abrasive particles) rises rapidly to 60%–73%, while the weights of the other four features remain at a low level of 7%–10%, indicating that the oxidation mechanism dominates the stable wear stage; in the transition period (280–300 min), the weight of PQ-O still plays a dominant role, indicating that although short-term cutting and fatigue effects begin to appear, the oxidation process still occupies a major position, providing a precursor for the subsequent mechanism mutation; subsequently, during the accelerated wear period (300–450 min), the weight of PQ-O gradually decreases, and the weights of IPCA and PQ-F rise synchronously to the range of 20%–50%. It also reached 34%, corresponding to the compound aggravation of fatigue crack growth effect; to the failure period (after 450min), IPCA, , PQ-F, and PQ-C all show secondary peaks, while the PQ-O weight drops back to 16%–26%, signaling the coordinated action of multiple mechanisms within the macro-destruction stage. This evolution fully demonstrates the key role of wear mechanism indicators in delineating different stages and provides an indispensable physical basis for high-precision identification of wear stages.

[0088] The introduction of wear mechanism indicators not only improves the accuracy of wear stage classification but also reveals the microscopic failure mechanisms and evolution patterns during the wear process. For example, by real-time monitoring of the instantaneous distribution and changing trends of three wear mechanism indicators—cutting wear, fatigue wear, and oxidation wear—it is possible to accurately determine the current dominant wear type.

[0089] like Figure 9 As shown in the figure, in the first 30 minutes of the experiment, the proportion of cutting abrasive particles is very high and gradually decreases. It can be determined that the material surface is experiencing high-speed meshing or hard inclusion cutting. At this time, the proportion of fatigue abrasive particles and oxidized abrasive particles is low, and the concentration of wear debris increases rapidly in the early stage and then tends to be flat. It can be seen that this is a typical running-in period. Figure 9 The presence of cutting abrasive particles can be clearly seen in Figure (a).

[0090] From 40 to 360 minutes, the wear rate remains at a low level, and the wear degree and the proportion of cutting abrasive particles fluctuate slightly but are also low. It can be considered that this is the stable wear stage. Figure 9The abrasive coverage area in (b) is significantly reduced, and the particles are primarily small. However, between 280 and 290 minutes, the proportion of oxidized abrasive particles suddenly increases, indicating that oxidative wear occurs, possibly due to temperature accumulation at the friction interface or local lubrication failure, leading to the rapid formation and flaking of a surface oxide film. Around this time, the degree of wear, the proportion of cutting abrasive particles, and the proportion of fatigue abrasive particles all show a slight upward trend. It can be inferred that the cutting and fatigue wear mechanisms also begin to work together, with microcrack initiation on the material surface and the cutting effect of hard particles gradually emerging. The combined action of this complex mechanism indicates that the stable wear stage is transitioning to the accelerated wear stage, an early signal of a sudden change in the wear mechanism.

[0091] Within 370 to 460 minutes, the wear rate increased significantly, the proportion of fatigue abrasive particles fluctuated periodically and continued to accumulate, and the degree of wear, the proportion of cutting abrasive particles, and the proportion of oxidized abrasive particles all increased to varying degrees, which means that multiple wear mechanisms have been out of control at the same time: fatigue cracks continue to initiate and expand under the action of cyclic stress, hard particles continue to cut the surface and produce more large-particle wear debris, and the peeling of the high-temperature oxide film also aggravates the brittle failure of the material. At this time, obvious microscopic pits and crack networks appear on the bearing surface, and the friction pair begins to experience adhesive wear and adhesive peeling, indicating that the system has entered an accelerated period, which is characterized by a continuous increase in wear rate and a rapid accumulation of surface damage. It can be seen from Figure 9 In (c), it can be seen that the abrasive coverage area increases significantly and large abrasive particles appear.

[0092] In the final stage of 470–480 min, all indicators increased sharply: the wear rate reached its peak, the proportion of cutting, fatigue and oxidation abrasive particles exploded simultaneously, and the degree of wear exceeded the historical high. It can be considered that the bearing has undergone macroscopic damage and is close to the failure limit. In this critical failure stage, the wear mechanism is further superimposed, forming a chain reaction - crack propagation causes the material blocks to fall off, the cutting action of large particles aggravates the surface roughness, and the repeated peeling of the oxide film induces new cracks. These cracks spread rapidly under high load and high temperature environment, eventually leading to the collapse of the structural integrity of the bearing. Figure 9 In (d), the abrasive coverage area increases dramatically, and fatigued large abrasive particles appear.

[0093] In summary, by introducing and comprehensively analyzing the three major wear mechanism indicators of cutting, fatigue, and oxidation, we can not only clearly distinguish the dynamic characteristics of the running-in, stable, accelerated, and severe wear stages, but also deeply reveal the interactive evolution of each mechanism on the timeline: the peak of initial cutting abrasive particles and their rapid decay reflect the formation of the surface lubricating film during the running-in period, the occasional oscillation of oxidation wear and fatigue abrasive particles in the stable stage indicates a precursor to instability, the simultaneous fermentation of multiple mechanisms during the acceleration period reveals the combined effect of crack initiation and coordinated cutting of hard particles, and the outbreak of multiple indicators during the severe wear period directly points to the critical state of macroscopic damage. This mechanism-based wear stage division method not only significantly improves the accuracy and robustness of stage division, but also provides physical explainability and decision support for online early warning, life prediction, and fault location.

[0094] Figure 10 A system for identifying equipment wear stages based on fusion decision making is shown, comprising: The feature extraction unit 1001 is configured to extract wear rate features, wear degree features, cutting index features, fatigue index features, and oxidation index features from a plurality of abrasive sample images; The feature clustering unit 1002 is configured to: cluster the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature respectively to obtain five groups of discrete stage labels; The data normalization unit 1003 is configured to: perform normalization processing on the five groups of discrete stage labels to obtain a normalized data matrix; The weight allocation unit 1004 is configured to: calculate the weights corresponding to the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature respectively according to the data matrix using a plurality of methods, and the average of the weights calculated in the various methods is the final weight of each feature; The comprehensive score calculation unit 1005 is configured to: calculate a comprehensive score sequence of all wear particle sample images according to the final weight of each feature and the normalized data matrix; The wear stage determining unit 1006 is configured to: perform clustering according to the comprehensive score sequence to obtain multiple optimal cluster centers, and determine the current wear stage according to the optimal cluster centers.

[0095] It is understandable that each of the above-mentioned units can be separately or completely combined into one or several other units to form a unit, or one (or some) of the units can be further divided into multiple functionally smaller units to form a unit, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above-mentioned units are divided based on logical functions. In actual applications, the functions of one unit can also be implemented by multiple units, or the functions of multiple units can be implemented by one unit. In other embodiments of the present application, the system may also include other units. In actual applications, these functions can also be implemented with the assistance of other units and can be implemented by the collaboration of multiple units.

[0096] According to another embodiment of the present application, the system described in this embodiment can be constructed by running a computer program (including program code) capable of executing the steps involved in the corresponding method of the present invention on a general-purpose computing device such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). The computer program can be recorded on, for example, a computer-readable recording medium, and loaded into the above-mentioned computing device through the computer-readable recording medium and run therein.

[0097] Figure 11 A computer device is shown, which includes a processor 1101, a communication interface 1102, and a computer-readable storage medium 1103. The processor 1101, the communication interface 1102, and the computer-readable storage medium 1103 may be connected via a bus or other means.

[0098] Among them, the communication interface 1102 is used to receive and send data, the computer-readable storage medium 1103 can be stored in the memory of the electronic device, the computer-readable storage medium 1103 is used to store computer programs, the computer programs include program instructions, and the processor 1101 is used to execute the program instructions stored in the computer-readable storage medium 1103.

[0099] The processor 1101 is the computing core and control core of the electronic device, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions to implement corresponding method processes or corresponding functions.

[0100] The processor 1101 is configured to perform the following process: Extract wear rate features, wear degree features, cutting index features, fatigue index features and oxidation index features from multiple abrasive sample images; The wear rate feature, wear degree feature, cutting index feature, fatigue index feature and oxidation index feature are clustered separately to obtain five groups of discrete stage labels; Normalize the five sets of discrete stage labels to obtain a normalized data matrix; According to the data matrix, weights corresponding to the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature are calculated respectively in a variety of ways, and the average of the weights calculated in various ways is the final weight of each feature; According to the final weight of each feature and the normalized data matrix, the comprehensive score sequence of all wear particle sample images is calculated; Clustering is performed according to the comprehensive score sequence to obtain multiple optimal cluster centers, and the current wear stage is determined according to the optimal cluster centers.

[0101] The present invention also provides a computer-readable storage medium, which is a memory device in an electronic device for storing programs and data. It is understood that the computer-readable storage medium herein may include both built-in storage media in the electronic device and, of course, extended storage media supported by the electronic device. The computer-readable storage medium provides storage space that stores the processing system of the electronic device.

[0102] Furthermore, the storage space also stores one or more instructions suitable for being loaded and executed by the processor. These instructions may be one or more computer programs (including program code). It should be noted that the computer-readable storage medium herein may be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device; alternatively, it may be at least one computer-readable storage medium located remotely from the processor.

[0103] In one embodiment, the computer-readable storage medium stores one or more instructions; the processor loads and executes the one or more instructions stored in the computer-readable storage medium to implement the following process: Extract wear rate features, wear degree features, cutting index features, fatigue index features and oxidation index features from multiple abrasive sample images; The wear rate feature, wear degree feature, cutting index feature, fatigue index feature and oxidation index feature are clustered separately to obtain five groups of discrete stage labels; Normalize the five sets of discrete stage labels to obtain a normalized data matrix; According to the data matrix, weights corresponding to the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature are calculated respectively in a variety of ways, and the average of the weights calculated in various ways is the final weight of each feature; According to the final weight of each feature and the normalized data matrix, the comprehensive score sequence of all wear particle sample images is calculated; Clustering is performed according to the comprehensive score sequence to obtain multiple optimal cluster centers, and the current wear stage is determined according to the optimal cluster centers.

[0104] The present invention also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the following process: Extract wear rate features, wear degree features, cutting index features, fatigue index features and oxidation index features from multiple abrasive sample images; The wear rate feature, wear degree feature, cutting index feature, fatigue index feature and oxidation index feature are clustered separately to obtain five groups of discrete stage labels; Normalize the five sets of discrete stage labels to obtain a normalized data matrix; According to the data matrix, weights corresponding to the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature are calculated respectively in a variety of ways, and the average of the weights calculated in various ways is the final weight of each feature; According to the final weight of each feature and the normalized data matrix, the comprehensive score sequence of all wear particle sample images is calculated; Clustering is performed according to the comprehensive score sequence to obtain multiple optimal cluster centers, and the current wear stage is determined according to the optimal cluster centers.

[0105] Those skilled in the art will appreciate that the units and algorithmic steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technical personnel may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0106] The above embodiments can be implemented in whole or in part using software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. A computer program product comprises one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions according to the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted via a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data processing device such as a server or data center that integrates one or more available media. Available media can include magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives).

[0107] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for identifying equipment wear stages based on fusion decision making, characterized in that: The following processes are included: Extract wear rate features, wear degree features, cutting index features, fatigue index features and oxidation index features from multiple abrasive sample images; The wear rate feature, wear degree feature, cutting index feature, fatigue index feature and oxidation index feature are clustered separately to obtain five groups of discrete stage labels; Normalize the five sets of discrete stage labels to obtain a normalized data matrix; According to the data matrix, weights corresponding to the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature are calculated respectively in a variety of ways, and the average of the weights calculated in various ways is the final weight of each feature; According to the final weight of each feature and the normalized data matrix, the comprehensive score sequence of all wear particle sample images is calculated; Clustering is performed according to the comprehensive score sequence to obtain multiple optimal cluster centers, and the current wear stage is determined according to the optimal cluster centers.

2. The equipment wear stage identification method based on fusion decision-making according to claim 1 is characterized in that: Before extracting the wear rate features, wear degree features, cutting index features, fatigue index features, and oxidation index features from multiple abrasive sample images, a pre-processing process is also included, including: The wear particle sample image is processed in sequence by grayscale processing, background subtraction processing, binarization processing and morphological processing.

3. The equipment wear stage identification method based on fusion decision-making according to claim 1 is characterized in that: The wear rate characteristic is the actual area of ​​the abrasive region, the wear degree characteristic is the number of abrasive particles whose equivalent circle diameter exceeds the set threshold, the cutting index characteristic is the proportion of cutting abrasive particles, the fatigue index characteristic is the proportion of fatigue abrasive particles, and the oxidation index characteristic is the proportion of oxidized abrasive particles.

4. The equipment wear stage identification method based on fusion decision-making according to claim 3 is characterized in that: Calculate the area, equivalent circle diameter, aspect ratio, roundness and curvature standard deviation of each abrasive particle; When the equivalent circle diameter is greater than or equal to the first set threshold, and the roundness is greater than or equal to the second set threshold, it is a normal abrasive particle, otherwise it goes to the next step of judgment; When the curvature standard deviation is less than or equal to the third set threshold, and the roundness is greater than or equal to the fourth set threshold, it is an oxidized abrasive particle, otherwise it proceeds to the next step of judgment; When the aspect ratio is greater than or equal to the fifth set threshold and the roundness is less than or equal to the second set threshold, it is a cutting abrasive particle; otherwise, it is a fatigue sliding abrasive particle.

5. The equipment wear stage identification method based on fusion decision-making according to claim 1 is characterized in that: The wear rate features, wear degree features, cutting index features, fatigue index features and oxidation index features are clustered using the Mean-shift algorithm to divide the wear stages.

6. The equipment wear stage identification method based on fusion decision-making according to claim 1 is characterized in that: The entropy weight method, CRITIC method and coefficient of variation method were used to calculate the weights of wear rate feature, wear degree feature, cutting index feature, fatigue index feature and oxidation index feature respectively; The final weight of the wear rate feature is the mean of the entropy weight, CRITIC weight, and coefficient of variation weight of the wear rate feature; The final weight of the wear degree feature is the mean of the entropy weight, CRITIC weight, and coefficient of variation weight of the wear degree feature; The final weight of the cutting index feature is the mean of the entropy weight, CRITIC weight, and coefficient of variation weight of the cutting index feature; The final weight of the fatigue index feature is the mean of the entropy weight, CRITIC weight, and coefficient of variation weight of the fatigue index feature; The final weight of the oxidation index feature is the average of the entropy weight, CRITIC weight, and coefficient of variation weight of the oxidation index feature.

7. The equipment wear stage identification method based on fusion decision-making according to any one of claims 1 to 6, characterized in that: Perform K-means clustering based on the comprehensive score sequence, calculate the optimal number of clusters K, obtain K cluster centers, sort the cluster centers, take adjacent center points as stage boundaries, and obtain the current wear stage; The current wear stages include: running-in wear stage, stable wear stage, oxidation wear stage, transition stage, accelerated wear stage and failure stage.

8. An equipment wear stage identification system based on fusion decision-making, characterized in that: include: The feature extraction unit is configured to extract a wear rate feature, a wear degree feature, a cutting index feature, a fatigue index feature, and an oxidation index feature from a plurality of abrasive sample images; The feature clustering unit is configured to: cluster the wear rate feature, wear degree feature, cutting index feature, fatigue index feature, and oxidation index feature respectively to obtain five groups of discrete stage labels; The data normalization unit is configured to: perform normalization processing on the five sets of discrete stage labels to obtain a normalized data matrix; a weight allocation unit configured to: calculate weights corresponding to the wear rate feature, the wear degree feature, the cutting index feature, the fatigue index feature, and the oxidation index feature respectively according to the data matrix in a plurality of ways, and the average of the weights calculated in the various ways is the final weight of each feature; The comprehensive score calculation unit is configured to: calculate a comprehensive score sequence of all wear particle sample images according to the final weight of each feature and the normalized data matrix; The wear stage determination unit is configured to: perform clustering according to the comprehensive score sequence to obtain multiple optimal cluster centers, and determine the current wear stage according to the optimal cluster centers.

9. A computer device, characterized in that: include: a processor and a computer-readable storage medium; a processor adapted to execute a computer program; A computer-readable storage medium having a computer program stored therein, wherein when the computer program is executed by the processor, the method for identifying equipment wear stages based on fusion decision-making according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which is suitable for being loaded by a processor and executing the equipment wear stage identification method based on fusion decision according to any one of claims 1 to 7.

Images