Rolling bearing health assessment method based on physically guided trend synchronization dynamic weighting

Abstract

The application belongs to the technical field of rolling bearing health assessment. A rolling bearing health assessment method based on physical guiding trend synchronous dynamic weighting is proposed. The oil particle image sequence of the bearing life cycle is obtained, the reference features representing wear accumulation and multiple auxiliary features representing different wear mechanisms are extracted through pretreatment, and a multi-dimensional feature matrix is constructed. Then, the matrix is segmented by using a sliding window, the effective degradation time is determined by the local change trend of the reference features, and the trend synchronization score of the auxiliary features and the reference features is calculated. According to the score and the anchoring constraint, the weight of each feature is dynamically determined and weighted fusion is carried out to generate a fusion health index sequence. Finally, a time series change point detection algorithm is applied to analyze the sequence to accurately divide the wear state stage of the rolling bearing. The method effectively fuses multiple sources of wear information through a dynamic weight mechanism, improves the accuracy and physical interpretability of health assessment, and provides a new idea for bearing condition monitoring.

Description

Technical Field

[0001] This invention relates to the field of rolling bearing health assessment technology, and specifically to a rolling bearing health assessment method based on physical guidance trend synchronous dynamic weighting. Background Technology

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

[0003] Oil wear analysis is an indispensable key technology for condition monitoring and health management of rotating machinery. Under harsh operating conditions of long-term heavy loads, variable speeds, and strong impacts, the core component, rolling bearings, inevitably undergoes a progressive degradation process from healthy break-in, early pitting, to severe spalling. The ability to accurately identify the health evolution stages throughout the entire life cycle using oil characteristics, especially the timely detection of subtle early wear, directly determines the proactiveness of fault prevention and the optimization of operation and maintenance costs. Once early abnormal condition identification fails, wear will accelerate exponentially, ultimately inducing catastrophic downtime accidents, causing huge economic losses and safety hazards.

[0004] Currently, wear condition assessment methods based on multiple oil features still have significant limitations. At the feature level, existing schemes use single macroscopic indicators such as the Index of Particle Coverage Area (IPCA) or simple statistical thresholds for discrimination. These indicators can only reflect the accumulation of wear but are difficult to capture the evolution of wear mechanisms, resulting in insufficient characterization of early, minor faults. At the fusion algorithm level, most existing multi-feature fusion strategies follow a purely data-driven global optimization paradigm, lacking effective constraints from physical mechanisms. This "black box" fusion mechanism struggles to distinguish between effective degradation information and random background interference, easily misjudging non-degradable background fluctuation noise as fault features during the stable operation phase of the bearing, leading to incorrect weight allocation and "misfitting." The resulting assessment results often have poor robustness, ambiguous distinctions between "healthy" and "abnormal" critical points, and a lack of physical interpretability in weight allocation, making it difficult to meet the high reliability diagnostic needs of industrial sites. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a rolling bearing health assessment method based on physical-guided trend synchronous dynamic weighting. By synchronously analyzing abrasive images, five key features characterizing different wear mechanisms—wear rate, wear degree, cutting index, fatigue index, and oxidation index—are extracted to construct a high-dimensional wear feature space. Based on this, a physical-guided trend synchronous dynamic weighting fusion strategy is introduced to achieve adaptive fusion of multi-source features during temporal evolution, highlighting sensitivity to changes in wear mechanisms. Finally, combined with a stage-based algorithm, the method achieves staged identification of the wear state throughout the rolling bearing's entire lifespan, providing reliable technical support for predictive maintenance and operational decisions for mechanical equipment.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for assessing the health of rolling bearings based on physical guidance trend synchronous dynamic weighting.

[0007] A method for assessing the health of rolling bearings based on physical guidance trends and synchronous dynamic weighting includes the following process: Obtain a sequence of oil abrasive images throughout the entire life cycle of a rolling bearing; The oil abrasive image sequence is preprocessed and features are extracted to construct a multi-dimensional feature matrix containing reference features and multiple auxiliary features. The reference features are abrasive concentration features that characterize the accumulation of wear volume, and the auxiliary features are at least one of wear rate features, large abrasive proportion features, cutting abrasive proportion features, fatigue abrasive proportion features, and oxidation abrasive proportion features that characterize different wear mechanisms. The multidimensional feature matrix is ​​segmented based on the sliding window mechanism to obtain multiple local feature fragments; For each local feature segment, the effective degradation time index set is determined by using the local change trend of the reference feature, and the trend synchronization score between each auxiliary feature and the reference feature is calculated based on the effective degradation time index set. Based on the trend synchronicity score and the preset anchoring constraints, the fixed weights of the reference features and the dynamic weights of the auxiliary features in each local feature segment are determined. By using fixed weights and dynamic weights, reference features and auxiliary features in local feature segments are weighted and fused to obtain a fused health index sequence. The wear state stage of rolling bearings is determined by segmenting the fused health index sequence based on the time-series variable point detection algorithm.

[0008] In one implementation of the first aspect of the present invention, calculating the trend synchronicity score between each auxiliary feature and the reference feature based on the effective degradation time index set includes: ; in, The value represents the correlation coefficient of the nth reference sample on the jth feature dimension; k represents the sampling point index. Represents the set of sampling points; This represents the maximum feature sequence value at the k-th sampling point; This represents the mean of the largest feature sequence under the j-th feature dimension; This represents the feature value of the nth reference sample at the kth sampling point; This represents the mean of the feature sequence of the nth reference sample; This represents the set of sampling points for the nth reference sample.

[0009] In one implementation of the first aspect of the present invention, determining the fixed weight of the reference feature and the dynamic weight of each auxiliary feature in each local feature segment based on the trend synchronicity score and a preset anchoring constraint includes: ; in, Represents the weight value of the j-th feature dimension; Represents the nth reference sample at the th... Correlation coefficients across each feature dimension; The total number of feature dimensions; The correlation coefficient of the nth sample on the jth feature dimension; The traversal index representing the feature dimension; This represents a smoothing term to prevent the denominator from being zero.

[0010] In one implementation of the first aspect of the present invention, reference features and auxiliary features in a local feature segment are weighted and fused using fixed weights and dynamic weights to obtain a fused health index sequence, including: ; in, This represents the health indicator value of the nth sample at the kth sampling point; Weighting coefficients representing reference features; This represents the reference feature value of the nth sample at the kth sampling point; Representative to D Sum the results across one auxiliary feature dimension; The weight coefficients represent the j-th auxiliary feature dimension; This represents the auxiliary feature value of the nth sample at the kth sampling point in the jth dimension; D represents the total number of feature dimensions.

[0011] In one implementation of the first aspect of the present invention, the fused health index sequence is segmented based on a time-series variable point detection algorithm to determine the wear state stage of the rolling bearing, including: Initialization treats the fused health index sequence as a complete data segment; Traverse all possible split positions in the complete data segment and calculate the loss function value after splitting the data segment into two parts at each split position. The loss function value is the sum of squares of the differences between the data in each part and the mean of that part. The optimal breakpoint is selected at the position with the minimum loss function value, and the complete data segment is divided into two sub-segments. The segmented sub-segments are recursively traversed, calculated, and selected until a preset number of critical breakpoints are detected. The time intervals segmented by the key breakpoints are mapped to different wear state stages in chronological order.

[0012] In one implementation of the first aspect of the present invention, the time intervals segmented by the key breakpoints are mapped to different wear state stages according to the time sequence, including: The time interval before the first critical breakpoint is mapped to a healthy state, where the fusion health index sequence corresponding to the healthy state remains near the low baseline. The time interval between the first and second critical breakpoints is mapped to the early wear state, where the fusion health index sequence corresponding to the early wear state shows an initial upward trend. The time interval between the second and third critical breakpoints is mapped to a severe wear state, where the fusion health index sequence corresponding to the severe wear state exhibits a rapidly rising nonlinear growth characteristic. The time interval after the third critical breakpoint is mapped to the failure state, where the fusion health index sequence corresponding to the failure state reaches a high extreme value.

[0013] In one implementation of the first aspect of the present invention, the oil abrasive particle image sequence is preprocessed and features are extracted to construct a multidimensional feature matrix containing reference features and multiple auxiliary features, including: The original color images in the oil abrasive image sequence are converted to grayscale to obtain grayscale images; Background subtraction and threshold segmentation are performed on the grayscale image to obtain a binarized abrasive grain image; Morphological operations are performed on the binarized abrasive grain images to restore the integrity of the abrasive grain regions and extract the abrasive grain contours; Abrasive particle concentration features are calculated by statistically counting the number of foreground pixels based on abrasive particle contours. Morphological parameters such as area, equivalent circle diameter, aspect ratio, roundness, and standard deviation of curvature of a single abrasive grain are extracted based on the abrasive grain profile. Based on morphological parameters, a discriminant tree model is used to classify abrasive particles, and the number of each type of abrasive particle is counted to calculate wear rate characteristics, large abrasive particle proportion characteristics, cutting abrasive particle proportion characteristics, fatigue abrasive particle proportion characteristics, and oxidation abrasive particle proportion characteristics. The calculated abrasive particle concentration feature is used as the reference feature, and other calculated features are used as auxiliary features to construct a multidimensional feature matrix.

[0014] Secondly, the present invention provides a method for assessing the health of rolling bearings based on physical guidance trend synchronous dynamic weighting.

[0015] A method for assessing the health of rolling bearings based on physical guidance trends and synchronous dynamic weighting includes the following process: The image acquisition unit is configured to acquire a sequence of oil abrasive images throughout the entire life cycle of the rolling bearing; The feature construction unit is configured to: preprocess and extract features from the oil abrasive image sequence, and construct a multi-dimensional feature matrix containing reference features and multiple auxiliary features. The reference features are abrasive concentration features that characterize the accumulation of wear volume, and the auxiliary features are at least one of wear rate features, large abrasive proportion features, cutting abrasive proportion features, fatigue abrasive proportion features, and oxidation abrasive proportion features that characterize different wear mechanisms. The segmentation unit is configured to segment the multidimensional feature matrix based on a sliding window mechanism to obtain multiple local feature segments; The synchronous computing unit is configured to: for each local feature segment, determine the effective degradation time index set by utilizing the local change trend of the reference feature, and calculate the trend synchronization score between each auxiliary feature and the reference feature based on the effective degradation time index set; The weight determination unit is configured to: determine the fixed weight of the reference feature and the dynamic weight of each auxiliary feature in each local feature segment based on the trend synchronicity score and the preset anchoring constraints. The index fusion unit is configured to use fixed weights and dynamic weights to perform weighted fusion of reference features and auxiliary features in local feature segments to obtain a fused health index sequence. The stage division unit is configured to: segment the fused health index sequence based on the time-series variable point detection algorithm to determine the wear state stage of the rolling bearing.

[0016] Thirdly, the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and executed by the rolling bearing health assessment method based on physical guidance trend synchronous dynamic weighting of the first aspect of the present invention.

[0017] Fourthly, the present invention provides a computer device, comprising: a processor and a computer-readable storage medium; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the rolling bearing health assessment method based on physical guidance trend synchronous dynamic weighting of the first aspect of the present invention.

[0018] Compared with the prior art, the beneficial effects of the present invention are: This invention overcomes the bottleneck of pure data-driven models being susceptible to noise interference. It innovatively establishes a trend masking mechanism with oil concentration as a "physical anchor point." By calculating feature synchronicity only during the effective degradation stage, this method can actively eliminate non-degradation fluctuations during the stable operation period of the bearing, effectively solving the problem of misfitting background noise in traditional attention mechanisms. It exhibits extremely high evaluation robustness under complex working conditions. This invention possesses excellent dynamic perception capabilities and can adapt to the non-stationary degradation characteristics of the bearing throughout its entire life cycle. Through a dynamic weighting strategy under physical constraints, the model can keenly capture subtle feature changes in the early stages of wear and automatically adjust the weight allocation during the severe wear period to adapt to the complex situation of multi-feature coupling. Combined with a time-series change point detection algorithm, this invention achieves accurate staged classification of wear state evolution, providing dynamic decision support covering the entire life cycle for condition-based maintenance.

[0019] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0021] Figure 1 A flowchart illustrating a rolling bearing health assessment method based on physical-guided trend synchronous dynamic weighting, provided as an exemplary embodiment of the present invention; Figure 2 A schematic diagram of a preprocessing flow for abrasive grain images is provided as an exemplary embodiment of the present invention; Figure 3 A schematic diagram of concentration evolution trend provided for an exemplary embodiment of the present invention; Figure 4 A schematic diagram illustrating the evolution trend of the proportion of large abrasive particles provided as an exemplary embodiment of the present invention; Figure 5 A schematic diagram illustrating the evolution trend of the proportion of cutting abrasive particles provided as an exemplary embodiment of the present invention; Figure 6A schematic diagram illustrating the evolution trend of fatigue wear particle proportion provided for an exemplary embodiment of the present invention; Figure 7 A schematic diagram illustrating the evolution trend of the proportion of oxide abrasive particles provided as an exemplary embodiment of the present invention; Figure 8 A schematic diagram of weight allocation provided for an exemplary embodiment of the present invention; Figure 9 A schematic diagram illustrating the division of health status stages provided in an exemplary embodiment of the present invention; Figure 10 A schematic diagram of a rolling bearing health assessment system based on physical guidance trend synchronous dynamic weighting, provided as 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. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0024] This implementation proposes a rolling bearing health assessment method based on physical-guided trend synchronous dynamic weighting, such as... Figure 1 The process shown includes the following steps: S101: Data Acquisition.

[0025] Based on accelerated wear experiments throughout the bearing's life cycle, this invention employs a high-throughput online monitoring system for oil abrasive images to achieve real-time perception of the entire bearing degradation process. This system can acquire the oil characteristics of the bearing at different health stages in real time and comprehensively obtain multi-dimensional key information including the microscopic morphology of abrasive particles. On this basis, the system accurately correlates and labels the streaming data by combining the actual physical wear state of the bearing, thus constructing a bearing life cycle oil abrasive dataset with high confidence labels.

[0026] More specifically, based on the bearing accelerated life test platform, this invention integrates a CMOS1 imaging system, which can directly capture the annular distribution characteristics of oil wear particles. The experiment ran for a total of 482 minutes, during which image sequences were acquired at 48 time points. The data at each sampling point contained two types of key images: annular images reflecting the macroscopic distribution of magnetic adsorption and dispersed particle images characterizing the microscopic morphology. These time-series data completely reproduced the entire life cycle evolution of the bearing from break-in, stable wear, abnormal degradation to severe failure, providing core data support for the subsequent construction of wear state stage classification based on oil characteristics.

[0027] S102: Image preprocessing.

[0028] To extract the features of abrasive grains from the image, it is necessary to eliminate interference from other factors in the image and preprocess the abrasive grain image. The processing flow is as follows: Figure 2 As shown.

[0029] The original color abrasive grain image contains information redundancy in its RGB three channels, and the color information has a weak correlation with key features such as abrasive grain morphology and density. Therefore, a weighted average method is first used to convert the image to grayscale, and the grayscale values ​​are... The calculation formula is: (1); In this representation, R, G, and B represent the intensities of the red, green, and blue channels, respectively, and their weights reflect the human eye's sensitivity to different colors. After grayscale processing, each pixel in the image is represented by only one grayscale value (0-255), making it easier to distinguish brighter areas.

[0030] Abrasive grain images are often affected by background interference such as uneven lighting, equipment reflection, and oil flow, which can introduce noise and affect the recognition of abrasive grains. To address this, a differential operation method is used to subtract the background image from the original image, thereby suppressing background interference and enhancing the saliency of the abrasive grain region.

[0031] Since the abrasive grains themselves are dark gray, the grayscale difference between them and the background area is small, and their outlines are not clear enough in the grayscale image, which is not conducive to the extraction of morphological features. Therefore, the Otsu global threshold segmentation algorithm is further adopted to automatically determine the optimal segmentation threshold and achieve the separation of the foreground (abrasive grains) from the background: (2); In the formula, For pixel grayscale values, The threshold is calculated by the algorithm. This step effectively removes the background, improves the recognizability of the abrasive grain contour, and lays the foundation for subsequent statistical extraction of features such as grain size and shape.

[0032] To improve the accuracy of feature extraction, morphological operations were introduced based on the initial preprocessing. To address potential issues such as holes or discontinuities within the abrasive grains after binarization, dilation and erosion operations were sequentially applied to restore the integrity of the abrasive grain regions. Subsequently, based on the grayscale difference between the abrasive grain edges and the background, the contour of each abrasive grain was extracted and numbered, providing fundamental data for subsequent shape parameter analysis.

[0033] S103: Multidimensional feature construction.

[0034] To overcome the limitations of traditional wear monitoring methods that rely solely on macroscopic indicators to reflect quantitative changes and accurately identify wear mechanisms, this invention selects five key features from multiple dimensions and mechanisms to achieve a more comprehensive and accurate assessment of equipment wear conditions. These five features are: wear rate, proportion of large abrasive particles, proportion of cutting abrasive particles, proportion of fatigue abrasive particles, and proportion of oxidized abrasive particles. Through the synergistic analysis of these features, the rate changes, severity, and differentiated manifestations of various wear mechanisms during the wear process can be effectively captured, thus providing a more mechanism-related basis for wear condition identification.

[0035] The abrasive coverage area index (IPCA) converts the material loss rate into a visible spatial coverage index by statistically analyzing the area of ​​all foreground pixels in an abrasive ring image. The foreground pixel coverage area is obtained by counting the number of all foreground pixels. (3); in, Foreground pixel coverage area The number of foreground pixels in the abrasive ring image, when the first... When the grayscale value of each pixel is 255, When the grayscale value is 1; when the grayscale value is 0, It is 0.

[0036] Further combining image pixel resolution This can be converted into actual physical area: (4); Abrasive grains with an equivalent circular diameter exceeding 30 μm are defined as large abrasive grains, and a significant increase in their number usually indicates an accelerated wear stage. Proportion of large abrasive grains The calculation formula is: (5); in, The proportion of large abrasive particles For a large number of abrasive grains, This represents the total number of abrasive grains.

[0037] Besides normal sliding wear, the main wear mechanisms include cutting wear, fatigue wear, and oxidative wear, which are quantified by the proportion of each type of abrasive grain. First, the following morphological features need to be extracted as classification criteria: The abrasive area A is the actual area of ​​a single abrasive grain based on pixel statistics.

[0038] (6); in, The area of ​​a single abrasive grain. To disperse the number of individual abrasive grain pixels in the abrasive grain image, when the first... When the grayscale value of each pixel is 255, When the grayscale value is 1; when the grayscale value is 0, It is 0.

[0039] The equivalent circle diameter D is the diameter of a circle with the same area as the abrasive grain.

[0040] (7); in, This is the equivalent diameter of a single abrasive grain.

[0041] The aspect ratio AR is the ratio of the long side to the short side of the smallest bounding rectangle of the abrasive grain. When AR≈1, the abrasive grain is approximately square; when AR>1, the abrasive grain is slender or flat.

[0042] (8); Roundness R describes how close the shape is to a circle. An ideal circle has a roundness of 1, meaning the abrasive grain is almost perfectly round. A roundness less than 1 indicates that the abrasive grain shape deviates from a circle, and the closer it is to 0, the more irregular the shape.

[0043] (9); in, The perimeter of the abrasive grain can be obtained from its profile.

[0044] The standard deviation of curvature C reflects the degree of fluctuation in the profile curvature and is used to identify spherical oxide abrasive particles.

[0045] Based on the above characteristics, a four-level discriminant tree model is established to classify all abrasive particles. The specific criteria are as follows: If D ≥ threshold 1 and R ≥ threshold 2, it is determined to be normal abrasive grains; otherwise, if C ≤ threshold 3 and R ≥ threshold 4, it is determined to be oxidized abrasive grains; otherwise, if AR ≥ threshold 5 and R ≤ threshold 6, it is determined to be cutting abrasive grains; otherwise, it is determined to be fatigue sliding abrasive grains.

[0046] More specifically, five key features—concentration, proportion of large abrasive particles, proportion of cutting abrasive particles, proportion of fatigue abrasive particles, and proportion of oxidized abrasive particles—are extracted from the preprocessed abrasive particle images to construct a full lifecycle oil feature dataset. The evolution trends of these five oil abrasive particle features are as follows: Figure 3 , Figure 4 , Figure 5 , Figure 6 and Figure 7 As shown.

[0047] S104: Multidimensional feature weighted fusion.

[0048] To address the limitations of purely data-driven fusion methods in the degradation process of rolling bearings, which suffer from a lack of physical constraints, making it difficult to distinguish feature contributions and susceptible to noise interference, this invention proposes a Physically Guided Trend Synchronization Dynamic Weighted (PG-TSDW) algorithm. Within a sliding window framework, this algorithm innovatively introduces a "physical anchor point" strategy to establish a degradation benchmark and combines it with a "trend synchronization" mechanism to quantify the dynamic correlation between features, aiming to construct a bearing health index (HI) that combines clear physical interpretability with strong robustness.

[0049] Let the standardized multidimensional feature matrix be... ,in For signal length, The feature quantity is defined as follows: Given the direct causal relationship between the cumulative increment of the total oil concentration feature and the bearing wear volume, this paper designates it as a physical anchor point (or reference feature), denoted as... ,the remaining Each feature is defined as an auxiliary feature, denoted as . ,in, .

[0050] To capture the time-varying characteristics of the degradation process, a length of [length missing] was used. The sliding window is used to segment the data, for the first... A window, whose local feature fragments are defined as follows: .

[0051] A core principle of this invention is that effective correlation between features should be established during the wear-accelerating phase, rather than during periods of stable operation or random fluctuations. Therefore, this invention introduces a degradation trend masking mechanism for the first... Reference feature sequence within a window Calculate its first difference to identify local trends, and define the time index set for "concentration showing an upward trend". for: (10); in, Represents the discrete difference operator; Representing the The first window The reference feature (concentration) values ​​for each sampling point; L represents the sliding window length; this set It can explicitly filter out background noise and retain only the time step in which physical indicators show that the bearing condition is deteriorating.

[0052] To quantify the contribution of auxiliary features to the degradation process, this invention proposes the Trend Synchronous Correlation Coefficient (TSCC). Unlike traditional global correlation, TSCC only applies to the extracted effective degradation stages. (i.e., the time index set of "concentration showing an upward trend") Calculations are performed within )

[0053] For the The first window Each auxiliary feature has a synchronicity score with the reference feature. The calculation method is as follows: (11); in, The value represents the correlation coefficient of the nth reference sample on the jth feature dimension; k represents the sampling point index. Represents the set of sampling points; This represents the maximum feature sequence value at the k-th sampling point; This represents the mean of the largest feature sequence under the j-th feature dimension; This represents the feature value of the nth reference sample at the kth sampling point; This represents the mean of the feature sequence of the nth reference sample; The set of sampling points representing the nth reference sample. The larger the value, the stronger the synchronous deterioration trend of this auxiliary feature and the concentration index at the time of accelerated wear.

[0054] To ensure the physical interpretability of the fusion index, a semi-fixed weight allocation strategy is adopted. Weight vector of each window The following is confirmed: Anchoring constraint: The weights of the reference features are fixed to ensure the reliability of the baseline trend. (12); in, The fixed weighting coefficient represents the reference feature (abrasive particle concentration).

[0055] Competitive allocation: The remaining 0.5 weighted budget is dynamically allocated based on the normalized synchronicity score of the auxiliary features. (13); in, Represents the weight value of the j-th feature dimension; Represents the nth reference sample at the th... Correlation coefficients across each feature dimension; The total number of feature dimensions; The correlation coefficient of the nth sample on the jth feature dimension; The traversal index representing the feature dimension; This represents a smoothing term to prevent the denominator from being zero.

[0056] Finally, the first Each window in local time Integrated Health Index Obtained through a weighted linear combination: (14); in, This represents the health indicator value of the nth sample at the kth sampling point; Weighting coefficients representing reference features; This represents the reference feature value of the nth sample at the kth sampling point; Representative to D Sum the results across one auxiliary feature dimension; The weight coefficients represent the j-th auxiliary feature dimension; This represents the auxiliary feature value of the nth sample at the kth sampling point in the jth dimension; D represents the total number of feature dimensions.

[0057] Formula (14) ensures that the generated HI is mainly driven by the amount (concentration) of physical wear, and only adaptively incorporates it when auxiliary features (such as the proportion of wear particles of a specific size) corroborate the deterioration trend, thereby effectively suppressing the interference of irrelevant features and improving the robustness of diagnosis.

[0058] The Physically Guided Trend Synchronization Dynamic Weighted (PG-TSDW) algorithm proposed in this invention is used to fuse multidimensional oil feature data. Abrasive particle concentration is used as the physical anchor point, and other features are dynamically weighted according to their synchronicity with the upward trend of abrasive particle concentration. Finally, multidimensional feature fusion is achieved. The weight allocation is illustrated in the diagram below. Figure 8 As shown in the diagram, the FSW (Time Series Segmentation) algorithm is used to divide the entire lifecycle training oil feature data into four stages: healthy state, early wear, severe wear, and failure state. The stage division diagram is shown below. Figure 9 As shown.

[0059] S105: Adaptive division of wear state based on time-series variable point detection.

[0060] The constructed fusion Health Index (HI) curve reflects the continuous degradation trajectory of bearing performance over time. To support the prediction of bearing remaining life and maintenance decisions, this continuous time series needs to be divided into discrete wear stages with clear physical meaning. A time series segmentation algorithm is used to detect three key breakpoints (corresponding to four state boundaries) in the Health Index (HI) sequence, and then the segmented intervals are mapped to discrete states in chronological order. The evolution of bearing health exhibits stage-specific characteristics: the healthy stage (HI) is stable, early wear (HI) increases slowly, severe wear (HI) increases rapidly, and the failure stage (HI) approaches its limit. This characteristic creates abrupt change points between different stages, which serve as the basis for segmentation.

[0061] Breakpoint detection relies on the Binseg algorithm (binary segmentation based on L2 loss) of the FSW system, which locates breakpoints by "recursive segmentation + loss minimization": the HI sequence is initialized as a complete segment, all possible segmentation positions are traversed for each segment, the sum of the L2 losses of the two segments after segmentation is calculated, and the point with the minimum loss is selected as the optimal breakpoint, until 3 breakpoints (corresponding to 4 segments) are obtained.

[0062] The L2 loss formula is: (15); in, This represents the L2 loss value with k as the split point; Represents the location of the candidate split point; Represents the starting index of the current data segment; This represents the end of the index for the current data segment; Represents the health index of the i-th sampling point; This represents the mean HI value of the first half after segmentation. This represents the mean HI value of the second half of the segment.

[0063] After obtaining the breakpoint, the `classify_stages` function completes the state mapping: based on the unidirectional irreversible nature of bearing failure, the time interval is mapped into four states "from morning to night": Healthy Stage, Time interval During this stage, the bearing operates smoothly, with abrasive particle concentration and auxiliary characteristics fluctuating at low levels, and the HI curve remaining near the low baseline; early wear... Time interval As surface micro-scraping occurs, tiny abrasive particles begin to appear in the oil, and the HI curve shows an initial upward trend, marking the beginning of degradation; severe wear... Time interval Accumulated wear leads to increased damage to the contact surface, a significant increase in large abrasive particles, and the HI curve exhibits a rapidly rising nonlinear growth characteristic; failure stage, Time interval When bearings fail, vibration intensifies and abrasive particle concentration reaches saturation or extreme values, the HI curve reaches a high level and may be accompanied by violent fluctuations.

[0064] The stage division method of this invention is used to divide the accelerated test data of rolling bearings into stages, specifically dividing the entire life cycle into four stages: 0-360 minutes is the healthy state; 360-420 minutes is the early wear state; 420-450 minutes is the severe wear state; and 450-480 minutes is the failure state. In the abrasive image at 200 minutes, the image shows very few abrasive grains, and the annular pattern is almost blank; in the abrasive image at 380 minutes, the image shows fine fatigue abrasive grains or a small amount of cutting abrasive grains; in the abrasive image at 420 minutes, the image shows more large abrasive grains, and the adsorption band in the annular pattern becomes significantly wider; in the abrasive image at 450 minutes, the image shows a large number of large abrasive grains, and the color of the annular band becomes significantly darker.

[0065] In summary, the wear stage segmentation method proposed in this invention was used to verify and analyze bearing accelerated life test data. The model accurately identified the critical state of "early wear." This result demonstrates that the intelligent stage segmentation model, which integrates physical priors, can sensitively capture the weak signals indicating the transition of bearings from healthy and stable operation to abnormal degradation, effectively revealing the early "qualitative change" characteristics in the wear evolution process. This experiment not only confirms the superiority of the proposed method compared to traditional monitoring methods but also proves that it can significantly detect potential faults in advance, providing crucial and timely scientific decision support for implementing condition-based maintenance (CBM) and optimizing proactive intervention strategies.

[0066] Figure 10 A method for assessing the health of rolling bearings based on physical guidance trends and synchronous dynamic weighting is presented, including the following process: The image acquisition unit 1001 is configured to acquire a sequence of oil abrasive images throughout the entire life cycle of a rolling bearing. The feature construction unit 1002 is configured to: preprocess and extract features from the oil abrasive image sequence, and construct a multi-dimensional feature matrix containing reference features and multiple auxiliary features. The reference features are abrasive concentration features that characterize the accumulation of wear volume, and the auxiliary features are at least one of wear rate features, large abrasive proportion features, cutting abrasive proportion features, fatigue abrasive proportion features, and oxidation abrasive proportion features that characterize different wear mechanisms. The segmentation unit 1003 is configured to segment the multidimensional feature matrix based on a sliding window mechanism to obtain multiple local feature segments; The synchronous computing unit 1004 is configured to: for each local feature segment, determine the effective degradation time index set by utilizing the local change trend of the reference feature, and calculate the trend synchronization score between each auxiliary feature and the reference feature based on the effective degradation time index set; The weight determination unit 1005 is configured to: determine the fixed weight of the reference feature and the dynamic weight of each auxiliary feature in each local feature segment based on the trend synchronicity score and the preset anchoring constraints. The index fusion unit 1006 is configured to: use fixed weights and dynamic weights to perform weighted fusion of reference features and auxiliary features in local feature segments to obtain a fused health index sequence; The stage division unit 1007 is configured to: segment the fused health index sequence based on the time-series variable point detection algorithm to determine the wear state stage of the rolling bearing.

[0067] It is understood that the aforementioned units can be individually or entirely combined into one or more other units, or some of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effect of the present invention. The aforementioned units are based on logical functional division. In practical applications, the function of one unit can be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of the present invention, the system may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0068] According to another embodiment of the present invention, the system of this embodiment can be constructed by running a computer program (including program code) capable of performing 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, loaded into the aforementioned computing device through the computer-readable recording medium, and run therein.

[0069] 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, communication interface 1102, and computer-readable storage medium 1103 can be connected via a bus or other means.

[0070] 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, which include program instructions. The processor 1101 is used to execute the program instructions stored in the computer-readable storage medium 1103.

[0071] The processor 1101 is the computing and control core of the electronic device. It is suitable for implementing one or more instructions, specifically for loading and executing one or more instructions to achieve the corresponding method flow or corresponding function.

[0072] Processor 1101 is configured to perform the following procedure: Obtain a sequence of oil abrasive images throughout the entire life cycle of a rolling bearing; The oil abrasive image sequence is preprocessed and features are extracted to construct a multi-dimensional feature matrix containing reference features and multiple auxiliary features. The reference features are abrasive concentration features that characterize the accumulation of wear volume, and the auxiliary features are at least one of wear rate features, large abrasive proportion features, cutting abrasive proportion features, fatigue abrasive proportion features, and oxidation abrasive proportion features that characterize different wear mechanisms. The multidimensional feature matrix is ​​segmented based on the sliding window mechanism to obtain multiple local feature fragments; For each local feature segment, the effective degradation time index set is determined by using the local change trend of the reference feature, and the trend synchronization score between each auxiliary feature and the reference feature is calculated based on the effective degradation time index set. Based on the trend synchronicity score and the preset anchoring constraints, the fixed weights of the reference features and the dynamic weights of the auxiliary features in each local feature segment are determined. By using fixed weights and dynamic weights, reference features and auxiliary features in local feature segments are weighted and fused to obtain a fused health index sequence. The wear state stage of rolling bearings is determined by segmenting the fused health index sequence based on the time-series variable point detection algorithm.

[0073] This 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 here may include both built-in storage media in the electronic device and extended storage media supported by the electronic device. The computer-readable storage medium provides storage space for storing the processing system of the electronic device.

[0074] Furthermore, this storage space also contains one or more instructions suitable for loading and execution by the processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory; alternatively, it can also be at least one computer-readable storage medium located remotely from the aforementioned processor.

[0075] 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 perform the following process: Obtain a sequence of oil abrasive images throughout the entire life cycle of a rolling bearing; The oil abrasive image sequence is preprocessed and features are extracted to construct a multi-dimensional feature matrix containing reference features and multiple auxiliary features. The reference features are abrasive concentration features that characterize the accumulation of wear volume, and the auxiliary features are at least one of wear rate features, large abrasive proportion features, cutting abrasive proportion features, fatigue abrasive proportion features, and oxidation abrasive proportion features that characterize different wear mechanisms. The multidimensional feature matrix is ​​segmented based on the sliding window mechanism to obtain multiple local feature fragments; For each local feature segment, the effective degradation time index set is determined by using the local change trend of the reference feature, and the trend synchronization score between each auxiliary feature and the reference feature is calculated based on the effective degradation time index set. Based on the trend synchronicity score and the preset anchoring constraints, the fixed weights of the reference features and the dynamic weights of the auxiliary features in each local feature segment are determined. By using fixed weights and dynamic weights, reference features and auxiliary features in local feature segments are weighted and fused to obtain a fused health index sequence. The wear state stage of rolling bearings is determined by segmenting the fused health index sequence based on the time-series variable point detection algorithm.

[0076] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can implement the described functions using different methods for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0077] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. 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 or transmitted through 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, fiber optic cable, digital cable) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0078] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for assessing the health of rolling bearings based on physical-guided trend-synchronized dynamic weighting, characterized in that, The process includes the following: Obtain a sequence of oil abrasive images throughout the entire life cycle of a rolling bearing; The oil abrasive image sequence is preprocessed and features are extracted to construct a multi-dimensional feature matrix containing reference features and multiple auxiliary features. The reference features are abrasive concentration features that characterize the accumulation of wear volume, and the auxiliary features are at least one of wear rate features, large abrasive proportion features, cutting abrasive proportion features, fatigue abrasive proportion features, and oxidation abrasive proportion features that characterize different wear mechanisms. The multidimensional feature matrix is ​​segmented based on the sliding window mechanism to obtain multiple local feature fragments; For each local feature segment, the effective degradation time index set is determined by using the local change trend of the reference feature, and the trend synchronization score between each auxiliary feature and the reference feature is calculated based on the effective degradation time index set. Based on the trend synchronicity score and the preset anchoring constraints, the fixed weights of the reference features and the dynamic weights of the auxiliary features in each local feature segment are determined. By using fixed weights and dynamic weights, reference features and auxiliary features in local feature segments are weighted and fused to obtain a fused health index sequence. The fused health index sequence is segmented based on a time-series variable point detection algorithm to determine the wear state stages of the rolling bearing, including: Initialization treats the fused health index sequence as a complete data segment; Traverse all possible split positions in the complete data segment and calculate the loss function value after splitting the data segment into two parts at each split position. The loss function value is the sum of squares of the differences between the data in each part and the mean of that part. The optimal breakpoint is selected at the position with the minimum loss function value, and the complete data segment is divided into two sub-segments. The segmented sub-segments are recursively traversed, calculated, and selected until a preset number of critical breakpoints are detected. The time intervals segmented by the key breakpoints are mapped to different wear state stages in chronological order.

2. The rolling bearing health assessment method based on physical guidance trend synchronous dynamic weighting as described in claim 1, characterized in that, The trend synchronicity score between each auxiliary feature and the reference feature is calculated based on the effective degradation time index set, including: ； in, The value represents the correlation coefficient of the nth reference sample on the jth feature dimension; k represents the sampling point index. Represents the set of sampling points; This represents the maximum feature sequence value at the k-th sampling point; This represents the mean of the largest feature sequence under the j-th feature dimension; This represents the feature value of the nth reference sample at the kth sampling point; This represents the mean of the feature sequence of the nth reference sample; This represents the set of sampling points for the nth reference sample.

3. The rolling bearing health assessment method based on physical guidance trend synchronous dynamic weighting as described in claim 1, characterized in that, Based on the trend synchronicity score and preset anchoring constraints, the fixed weights of reference features and the dynamic weights of auxiliary features in each local feature segment are determined, including: ； in, This represents the weight value of the j-th feature dimension; Represents the nth reference sample at the th... Correlation coefficients across each feature dimension; The total number of feature dimensions; The correlation coefficient of the nth sample on the jth feature dimension; The traversal index representing the feature dimension; This represents a smoothing term to prevent the denominator from being zero.

4. The rolling bearing health assessment method based on physical guidance trend synchronous dynamic weighting as described in claim 1, characterized in that, Using fixed and dynamic weights, reference features and auxiliary features in local feature segments are weighted and fused to obtain a fused health index sequence, including: ； in, This represents the health indicator value of the nth sample at the kth sampling point; Weighting coefficients representing reference features; This represents the reference feature value of the nth sample at the kth sampling point; Representative to D Sum the results across one auxiliary feature dimension; The weight coefficients represent the j-th auxiliary feature dimension; This represents the auxiliary feature value of the nth sample at the kth sampling point in the jth dimension; D represents the total number of feature dimensions.

5. The rolling bearing health assessment method based on physical guidance trend synchronous dynamic weighting as described in claim 1, characterized in that, The time intervals segmented by the key breakpoints are mapped into different wear state stages according to chronological order, including: The time interval before the first critical breakpoint is mapped to a healthy state, where the fusion health index sequence corresponding to the healthy state remains near the low baseline. The time interval between the first and second critical breakpoints is mapped to the early wear state, where the fusion health index sequence corresponding to the early wear state shows an initial upward trend. The time interval between the second and third critical breakpoints is mapped to a severe wear state, where the fusion health index sequence corresponding to the severe wear state exhibits a rapidly rising nonlinear growth characteristic. The time interval after the third critical breakpoint is mapped to the failure state, where the fusion health index sequence corresponding to the failure state reaches a high extreme value.

6. The rolling bearing health assessment method based on physical guidance trend synchronous dynamic weighting as described in claim 1, characterized in that, The oil abrasive particle image sequence was preprocessed and its features extracted to construct a multidimensional feature matrix containing reference features and multiple auxiliary features, including: The original color images in the oil abrasive image sequence are converted to grayscale to obtain grayscale images; Background subtraction and threshold segmentation are performed on the grayscale image to obtain a binarized abrasive grain image; Morphological operations are performed on the binarized abrasive grain images to restore the integrity of the abrasive grain regions and extract the abrasive grain contours; Abrasive particle concentration features are calculated by statistically counting the number of foreground pixels based on abrasive particle contours. Morphological parameters such as area, equivalent circle diameter, aspect ratio, roundness, and standard deviation of curvature of a single abrasive grain are extracted based on the abrasive grain profile. Based on morphological parameters, a discriminant tree model is used to classify abrasive particles, and the number of each type of abrasive particle is counted to calculate wear rate characteristics, large abrasive particle proportion characteristics, cutting abrasive particle proportion characteristics, fatigue abrasive particle proportion characteristics, and oxidation abrasive particle proportion characteristics. The calculated abrasive particle concentration feature is used as the reference feature, and other calculated features are used as auxiliary features to construct a multidimensional feature matrix.

7. A method for assessing the health of rolling bearings based on physical-guided trend synchronous dynamic weighting, characterized in that, The process includes the following: The image acquisition unit is configured to acquire a sequence of oil abrasive images throughout the entire life cycle of the rolling bearing; The feature construction unit is configured to: preprocess and extract features from the oil abrasive image sequence, and construct a multi-dimensional feature matrix containing reference features and multiple auxiliary features. The reference features are abrasive concentration features that characterize the accumulation of wear volume, and the auxiliary features are at least one of wear rate features, large abrasive proportion features, cutting abrasive proportion features, fatigue abrasive proportion features, and oxidation abrasive proportion features that characterize different wear mechanisms. The segmentation unit is configured to segment the multidimensional feature matrix based on a sliding window mechanism to obtain multiple local feature segments; The synchronous computing unit is configured to: for each local feature segment, determine the effective degradation time index set by utilizing the local change trend of the reference feature, and calculate the trend synchronization score between each auxiliary feature and the reference feature based on the effective degradation time index set; The weight determination unit is configured to: determine the fixed weight of the reference feature and the dynamic weight of each auxiliary feature in each local feature segment based on the trend synchronicity score and the preset anchoring constraints. The index fusion unit is configured to use fixed weights and dynamic weights to perform weighted fusion of reference features and auxiliary features in local feature segments to obtain a fused health index sequence. The stage division unit is configured to: segment the fused health index sequence based on a time-series change point detection algorithm to determine the wear state stages of the rolling bearing, including: Initialization treats the fused health index sequence as a complete data segment; Traverse all possible split positions in the complete data segment and calculate the loss function value after splitting the data segment into two parts at each split position. The loss function value is the sum of squares of the differences between the data in each part and the mean of that part. The optimal breakpoint is selected at the position with the minimum loss function value, and the complete data segment is divided into two sub-segments. The segmented sub-segments are recursively traversed, calculated, and selected until a preset number of critical breakpoints are detected. The time intervals segmented by the key breakpoints are mapped to different wear state stages in chronological order.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1 to 6, the rolling bearing health assessment method based on physical guidance trend synchronous dynamic weighting.

9. A computer device, characterized in that, include: Processor and computer-readable storage media; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the rolling bearing health assessment method based on physical guidance trend synchronous dynamic weighting as described in any one of claims 1 to 6.

Images