Comprehensive evaluation method of high-speed deep groove ball bearing NVH

Abstract

This invention discloses a comprehensive evaluation method for NVH (Noise, Vibration, and Harshness) of high-speed deep groove ball bearings. The method first involves multi-dimensional data acquisition and preprocessing to establish a multi-dimensional evaluation index system and dataset covering bearing geometric and vibration parameters. Then, it constructs a dual-channel feature matrix for energy and impact, and completes independent evaluation of the dual channels using the quartile-robin robust normalized entropy weight method to obtain a comprehensive bearing vibration score. Subsequently, a geometry-dynamic mapping prediction model is established, and a nonlinear comprehensive evaluation function is built to calculate the final score. Based on the score, the bearing is classified into four levels: excellent, qualified, and unqualified, and a diagnostic report containing defect analysis, rework suggestions, and level conclusions is output. This invention solves the problems of traditional bearing NVH evaluation methods, such as single-dimensionality evaluation, subjective weight allocation, lack of quantitative evaluation, and insufficient parameter correlation analysis. It achieves accurate and comprehensive quantitative evaluation of bearing NVH performance, providing a reliable technical basis for quality control and optimization design of high-end bearings.

Description

Technical Field

[0001] This invention relates to the field of bearing performance evaluation technology, and in particular to a method for evaluating NVH (noise, vibration, and acoustic roughness) of high-speed deep groove ball bearings. Background Technology

[0002] In high-speed rotating machinery, the NVH performance of bearings is a key factor affecting the overall performance and reliability of the machine. High-speed deep groove ball bearings are widely used in automobiles, aerospace, precision machinery and other fields, and their NVH performance directly affects the comfort, safety and service life of the products.

[0003] In the prior art, patent document CN119023264A discloses a bearing noise testing method, which drives the test shaft to rotate through a drive mechanism, and uses a soundproof cover, vibration sensor and noise testing mechanism to isolate background noise and collect vibration and noise data of the bearing respectively; at the same time, it is equipped with an oil cooling lubrication system and a multi-directional force loading mechanism to simulate axial and radial forces under real working conditions, thereby accurately solving the NVH testing problem of drive motor bearings. However, it only considers the single dimension of vibration velocity and cannot comprehensively reflect the NVH performance of the bearing. Another prior art, CN110987472A, discloses a bearing housing noise treatment system for NVH testing of automotive electric drive systems. This patent discloses a bearing housing noise treatment system for NVH testing of pure electric drive systems. By setting up a basement and a horizontal actuation mechanism on the floor of the anechoic chamber, the bearing housing is placed underground, which isolates the structure from sound transmission and optimizes the sound field environment. This effectively solves the problem of NVH data distortion caused by background noise interference and improves the accuracy of testing. However, the disadvantages are that the construction of the underground structure is complex and costly, and the horizontal actuation mechanism may limit the ability to simulate multi-directional vibration, making it difficult to fully reproduce the coupled noise characteristics under complex working conditions.

[0004] While existing technologies can evaluate bearing performance from a single dimension, they have the following drawbacks in practical applications: (1) Single evaluation dimension: Existing technologies evaluate bearing performance only from the single dimension of vibration or geometric accuracy, which cannot fully reflect the comprehensive performance of NVH. NVH performance is a comprehensive reflection of multiple factors such as vibration, noise, and geometric accuracy. Single-dimensional evaluation cannot accurately reflect the actual NVH level of the bearing. (2) Lack of quantitative comprehensive evaluation: Existing technologies mostly adopt qualitative evaluation or single index evaluation, lacking scientific quantitative comprehensive evaluation methods. They cannot comprehensively analyze the evaluation results of different dimensions and are difficult to provide a comprehensive performance assessment. (3) Subjective weight allocation: Some existing technologies adopt subjective weight allocation methods, which lack scientific basis. The evaluation results are greatly affected by subjective factors and have low reliability. (4) Insufficient correlation analysis: Existing technologies do not deeply analyze the correlation between vibration performance and geometric accuracy, cannot reveal the influence mechanism of NVH performance, and are difficult to guide the optimized design of bearings.

[0005] To address the technical problems in existing bearing NVH evaluation technologies, such as single dimensions, lack of quantitative comprehensive evaluation, subjective weight allocation, and insufficient correlation analysis, this invention aims to provide a multi-dimensional comprehensive evaluation method for NVH of high-speed deep groove ball bearings. Through the fusion of multi-dimensional indicators and scientific weight allocation, a comprehensive and accurate NVH evaluation can be achieved. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention aims to provide a comprehensive evaluation method for the NVH (Noise, Vibration, and Harshness) of high-speed deep groove ball bearings. This method first comprehensively integrates key geometric and dynamic parameters such as vibration velocity, vibration acceleration, and the roundness, waviness, and roughness of the inner and outer raceways, establishing a comprehensive evaluation index library. It then introduces the entropy weight method for in-depth analysis of each dimension, achieving a scientific and quantitative allocation of weights. A dual-channel fusion comprehensive evaluation model is established, transforming scattered index data into a quantitative comprehensive score, thus achieving a precise "portrait" of the bearing's NVH performance and revealing the contribution of microscopic geometric features of each dimension to macroscopic NVH performance. This innovative method significantly improves the accuracy and comprehensiveness of the evaluation results, successfully bridging the gap between theoretical analysis and engineering application. This achievement provides a solid scientific basis for the optimized design, process improvement, and quality control of high-end bearings, and has significant engineering application value for promoting the quieter development of new energy vehicle drive systems.

[0007] To achieve the above objectives, this invention provides the following technical solution: A comprehensive evaluation method for NVH of high-speed deep groove ball bearings includes the following steps: Step 1, Multidimensional data acquisition and preprocessing: Acquire external multidimensional data, and construct a multidimensional evaluation index system based on the multidimensional data, including inner and outer ring rolling roundness, waviness and roughness, vibration velocity, vibration acceleration, kurtosis and crest factor, to construct a corresponding dataset; Step 2, Constructing an "energy-impact" dual-channel feature matrix: The reciprocal of the displacement with respect to time during bearing rotation is used as an index to measure the bearing vibration energy, named channel A; the distribution probability of the velocity with respect to time and the amplitude, as well as the crest intensity, are used as indices to measure the impact of bearing vibration, and a two-channel A / B feature matrix is ​​constructed based on the dataset constructed in Step 1; Step 3, Performing independent entropy weight evaluation within the dual channels: First, based on robust normalization of the interquartile range, then calculate the information entropy e jThe process involves several steps: Step 1: Calculating the information utility value and weights, and finally, a comprehensive score. A higher comprehensive score indicates a worse product quality. Step 4: Establishing a "geometric-dynamic" mapping prediction model. Step 5: Performing full-scenario dynamic residual diagnosis and analysis based on the prediction model established in Step 4. Step 6: Constructing a nonlinear comprehensive evaluation function to calculate the final score. Step 7: Based on the final score calculated in Step 6, classifying and outputting the product grades into top-grade, superior, qualified, and unqualified products. The final output is a bearing NVH diagnostic report. This report analyzes the bearing's original dataset and residuals to determine if defects are possible. For minor defects that can be repaired, it provides reasonable rework suggestions and concludes the bearing NVH grade classification.

[0008] As a further improvement of the present invention, the specific method of multidimensional data acquisition and preprocessing in step one is as follows: First, the geometric parameters of the bearing are measured and an initial signal set G1{Ciri,Wzi,Rai,Cire,Wze,Rae,ec} is constructed using the rolling roundness, waviness, roughness of the inner and outer rings of the bearing and the contaminant coefficient of the lubricating medium. Second, the vibration velocity index of the bearing is measured and a dataset G2{VL,VM,VH} is constructed using the vibration velocity VL / VM / VH of the bearing. Finally, the vibration acceleration, kurtosis and crest factor of the bearing are measured to construct a dataset G3{Z,K,CF}.

[0009] As a further improvement of the present invention, the specific details of the "energy-impact" dual-channel feature matrix constructed in step two are as follows: Vibration energy channel A, using the bearing vibration velocity measurement dataset G2 obtained in step one, constructs a feature matrix XG2=[VL, VM, VH] reflecting the dynamic operating performance of the bearing; Vibration impact channel B, using the bearing vibration velocity measurement dataset G3 obtained in step one, constructs a feature matrix XG3=[Z, K, CF] reflecting the dynamic operating performance of the bearing; Wherein, the above feature matrices XG2 and XG3 both have at least 9 rows, and are at least 3 times the statistical dimension.

[0010] As a further improvement of the present invention, the specific method for evaluating the independent entropy weights within the dual channels in step three is as follows: (1) Based on robust normalization of the interquartile range, specifically: a. For each column of data in the feature matrices of channel A and channel B, calculate the median Xmedian, the first quartile Q1, the third quartile Q3, and the interquartile range IQR = Q3 - Q1; b. Perform data normalization transformation. Since Xmedian is the median, Z will inevitably have negative values ​​here, and will be centered at 0; c. To satisfy the basic conditions for subsequent calculations using the entropy weight method, it is necessary to... The value is shifted to the non-negative interval, i.e. Where min(Z) is the minimum value of data in b after normalization transformation. For a data point that is a minimum value, it is defined here as... This allows us to construct a completely new matrix. d. Calculate the specific gravity The newly constructed matrix Divide each data point in the table by the sum of the data in the column containing that value, and calculate the proportion of the i-th sample to that index under the j-th index. : (2) Calculate the information entropy ej. Calculate the information entropy of the j-th phase index: Where k is related to the sample size m, and is defined as follows: Note: ,but (3) Calculate the information utility value in The larger the value, the higher the discrimination of the indicator and the more important it is; (4) Calculate the weight by normalizing the information utility value to obtain the final weight of the j-th indicator: (5) Calculate the comprehensive score to obtain the bearing vibration energy score based on the feature matrices XG2 and XG3. and vibration and impact scores : The evaluation is based on a comprehensive score; the higher the comprehensive score, the worse the evaluation is considered.

[0011] As a further improvement of the present invention, the specific method for establishing the "geometry-dynamic" mapping prediction model in step four is as follows: (1) Constructing training datasets Based on previous sample libraries, construct datasets G4{Ciri,Wzi,Rai,Cire,Wze,Rae} containing the roundness, waviness, and roughness of the inner and outer raceways of the bearing, and dataset G5{Wzi,Rai,Wze,Rae,ec} containing the waviness, roughness, and lubricant contaminant coefficient of the inner and outer raceways of the bearing. (2) Constructing prediction models based on multiple linear regression a. Vibration energy prediction model, establishing the mapping relationship between dataset G4{Ciri,Wzi,Rai,Cire,Wze,Rae} and bearing vibration energy score Sv' based on multiple linear regression: in It is a constant; when all measured data for the current bearing are 0, the corresponding... Expected value; The error term includes unconsidered factors and random noise; b. The vibration and shock prediction model establishes a mapping relationship between the dataset G5{Wzi,Rai,Wze,Rae,ec} and the bearing vibration and shock score Sz' based on multiple linear regression. Similarly: in It is a constant; when all measured data for the current bearing are 0, the corresponding... Expected value; The error term includes unconsidered factors and random noise; (3) Output the predicted value. Using the multiple linear regression model constructed in step (2) above for vibration energy and vibration impact, input the parameters of each index in the bearing dataset G1 collected in step one to obtain the predicted scores Sv' and Sz' of bearing vibration energy and vibration impact.

[0012] As a further improvement of the present invention, the full-scene dynamic residual diagnosis and analysis in step five is as follows: (1) Calculate the residual, which is calculated using the following formula: The formula for calculating the vibration energy residual is: The formula for calculating vibration and impact residuals is: (2) The diagnostic logic adopts a four-quadrant diagnostic logic, which calculates the residual between the observed value and the predicted value. , The bearing condition is precisely classified into four categories: dynamic instability, steady-state imbalance, hidden impact, and excellent matching, as follows: Scenario 1: Double positive residuals, dynamic instability type, not only poor geometric parameters, but also defects in assembly and lubrication, seriously unqualified; Scenario 2: Only positive energy residuals, steady-state imbalance type, the bearing's macroscopic motion is unstable, moderately unqualified; Scenario 3: Only positive impact residuals, hidden impact type, high-quality transient impact, seriously unqualified; Scenario 4: Double negative or zero residuals, excellent matching type, both geometric parameters and assembly are very good, qualified product.

[0013] As a further improvement of the present invention, the nonlinear comprehensive evaluation function is constructed in step six as follows: (1) Calculation of basic score a. Reverse normalization, the higher the score, the better. b. The formula for calculating the basic score is: For high-speed deep groove ball bearings, the effects of vibration acceleration and NVH are more significant, therefore, it is specified in the calculation that... , (2) Determine the consistency correction coefficient K; Based on the diagnostic results in step five, construct a scoring reward and punishment mechanism: a. Dynamic instability type, K=0.3, moderate punishment; b. Steady-state imbalance type, K=0.7, mild punishment; c. Hidden impact type, K=0.1, severe punishment; d. Excellent matching type, K=1.1, reward (3) Calculate the final score. .

[0014] As a further improvement to the present invention, the grading method in step seven is as follows: Premium grade products: Premium grade: Qualified products: Defective products: .

[0015] The beneficial effects of this invention are as follows: By constructing a multi-dimensional evaluation index system that includes geometric accuracy and dynamic vibration parameters, this invention solves the problem of single evaluation dimensions in existing technologies; by adopting a dual-channel independent entropy weight evaluation method, it achieves scientific quantitative allocation of weights, eliminating the interference of subjective factors on the evaluation results; by establishing a "geometry-dynamic" mapping prediction model and combining it with full-scenario dynamic residual diagnosis, it deeply reveals the intrinsic relationship between geometric parameters and NVH performance, achieving accurate defect location and classification; through a nonlinear comprehensive evaluation function and a grade division mechanism, it forms a complete quantitative comprehensive evaluation system, capable of outputting a comprehensive diagnostic report including defect analysis and rework suggestions, providing scientific and reliable technical support for the quality control, process improvement, and optimization design of high-speed deep groove ball bearings. Subsequent improvement steps further refine the specific processes of data acquisition, feature construction, model training, and diagnostic evaluation, improving the operability of the method and the accuracy of the evaluation results. Detailed Implementation

[0016] The present invention will now be described in further detail with reference to specific embodiments.

[0017] The comprehensive evaluation method for NVH of high-speed deep groove ball bearings in this embodiment includes the following steps: Step 1, Multidimensional data acquisition and preprocessing: Acquire external multidimensional data, and construct a multidimensional evaluation index system based on the multidimensional data, including inner and outer ring rolling roundness, waviness and roughness, vibration velocity, vibration acceleration, kurtosis and crest factor, to construct a corresponding dataset; Step 2, Construct an "energy-impact" dual-channel feature matrix: Use the reciprocal of displacement with respect to time to describe the bearing rotation as an index to measure the bearing vibration energy, and name it Channel A; Use the distribution probability of velocity with respect to time and amplitude and crest intensity to describe the bearing rotation as an index to measure the bearing vibration impact, and construct a two-channel A / B feature matrix based on the dataset constructed in Step 1; Step 3, Perform independent entropy weight evaluation within the dual channels: First, based on robust normalization of the interquartile range, then calculate the information entropy e jThe process involves several steps: Step 1: Calculating the information utility value and weights, and finally, a comprehensive score. A higher comprehensive score indicates a worse product quality. Step 4: Establishing a "geometric-dynamic" mapping prediction model. Step 5: Performing full-scenario dynamic residual diagnosis and analysis based on the prediction model established in Step 4. Step 6: Constructing a nonlinear comprehensive evaluation function to calculate the final score. Step 7: Based on the final score calculated in Step 6, classifying and outputting the product grades into top-grade, superior, qualified, and unqualified products. The final output is a bearing NVH diagnostic report. This report analyzes the bearing's original dataset and residuals to determine if defects are possible. For minor defects that can be repaired, it provides reasonable rework suggestions and concludes the bearing NVH grade classification.

[0018] In this embodiment, step one constructs a multi-dimensional evaluation index system by simultaneously collecting the bearing's geometric accuracy parameters and dynamic vibration parameters, overcoming the limitations of existing technologies that evaluate from only a single dimension, and comprehensively reflecting the overall performance of the bearing's NVH (Noise, Vibration, and Harshness) performance. Step two separates the vibration characteristics into two independent channels: energy and impact, corresponding to the bearing's macroscopic steady-state vibration and microscopic transient impact characteristics, respectively, achieving refined analysis of the vibration signal. Step three employs an entropy weighting method based on robust quartile normalization to avoid the influence of outliers on the evaluation results, while objectively calculating the weights of each index through information entropy, solving the subjective problem of weight allocation in existing technologies. Steps four and five establish a mapping relationship between geometric parameters and dynamic performance and perform residual diagnosis, which can distinguish between geometric defects and non-geometric defects such as assembly and lubrication defects, revealing the influence mechanism of NVH performance. Steps six and seven transform the multi-dimensional evaluation results into intuitive quantitative scores and grade conclusions through nonlinear comprehensive evaluation functions and grade division, and output targeted diagnostic reports, realizing quantitative comprehensive evaluation and decision support for bearing NVH performance.

[0019] Furthermore, the specific methods for multidimensional data acquisition and preprocessing in step one are as follows: First, measure the bearing's geometric parameters and construct an initial signal set G1{Ciri,Wzi,Rai,Cire,Wze,Rae,ec} based on the bearing's inner and outer ring rolling roundness, waviness, roughness, and contaminant coefficient of the lubricating medium. Second, measure the bearing's vibration velocity indices and construct a dataset G2{VL,VM,VH} based on the bearing's vibration velocities VL / VM / VH. Finally, measure the bearing's vibration acceleration, kurtosis, and crest factor to construct a dataset G3{Z,K,CF}.

[0020] In this step, geometric parameters were measured using a Taylor-Hopson FormTalysurfi600 surface profilometer; vibration parameters were acquired using a PCB352C33 piezoelectric accelerometer, mounted radially on the bearing housing. Each parameter was measured three times and the average value was taken to reduce random errors. The lubricant contaminant coefficient (ec) was determined using a particle counter. This step, by constructing datasets of different dimensions through classification, achieved standardized management of the raw data, providing a unified data foundation for the subsequent construction of the dual-channel feature matrix. Furthermore, incorporating the lubricant contaminant coefficient further enriched the evaluation dimensions and improved the comprehensiveness of the evaluation results.

[0021] Furthermore, the specific details of the "energy-impact" dual-channel feature matrix constructed in step two are as follows: Vibration energy channel A, using the bearing vibration velocity measurement dataset G2 obtained in step one, constructs a feature matrix XG2=[VL, VM, VH] reflecting the dynamic operating performance of the bearing; Vibration impact channel B, using the bearing vibration velocity measurement dataset G3 obtained in step one, constructs a feature matrix XG3=[Z, K, CF] reflecting the dynamic operating performance of the bearing; Wherein, the above feature matrices XG2 and XG3 both have at least 9 rows, and are at least 3 times the statistical dimension.

[0022] In this embodiment, 30 6205-2RS bearings from the same batch can be selected as statistical samples. Therefore, both feature matrices XG2 and XG3 are 30 rows and 3 columns, which meets the statistical sample size requirements. This step ensures a sufficient sample size for statistical analysis by limiting the minimum number of rows in the feature matrix, avoiding the bias in evaluation results caused by small samples. At the same time, it assigns different vibration characteristic indicators to different channels, laying the foundation for subsequent dual-channel independent entropy weight evaluation, which can more accurately reflect the impact of different vibration characteristics on NVH performance.

[0023] Furthermore, the specific method for evaluating the independent entropy weights within the dual channels in step three is as follows: (1) Based on robust normalization of the interquartile range, specifically: a. For each column of data in the feature matrices of channel A and channel B, calculate the median Xmedian, the first quartile Q1, the third quartile Q3, and the interquartile range IQR = Q3 - Q1; b. Perform data normalization transformation. Since Xmedian is the median, Z will inevitably have negative values ​​here, and will be centered at 0; c. To satisfy the basic conditions for subsequent calculations using the entropy weight method, it is necessary to... The value is shifted to the non-negative interval, i.e. Where min(Z) is the minimum value of data in b after normalization transformation. For a data point that is a minimum value, it is defined here as... This allows us to construct a completely new matrix. d. Calculate the specific gravity The newly constructed matrix Divide each data point in the table by the sum of the data in the column containing that value, and calculate the proportion of the i-th sample to that index under the j-th index. : (2) Calculate the information entropy ej. Calculate the information entropy of the j-th phase index: Where k is related to the sample size m, and is defined as follows: Note: ,but (3) Calculate the information utility value in The larger the value, the higher the discrimination of the indicator and the more important it is; (4) Calculate the weight by normalizing the information utility value to obtain the final weight of the j-th indicator: (5) Calculate the comprehensive score to obtain the bearing vibration energy score based on the feature matrices XG2 and XG3. and vibration and impact scores : The evaluation is based on a comprehensive score; the higher the comprehensive score, the worse the evaluation is considered.

[0024] In this embodiment, the weights of each indicator in channel A are calculated as follows: , , Weights of each indicator in Channel B: , , It is evident that the mid-frequency vibration velocity VM and vibration kurtosis K have the most significant impact on the bearing's NVH performance, which is consistent with the vibration characteristics of bearings under high-speed operating conditions. This step employs interquartile range robust normalization instead of traditional range normalization, effectively reducing the impact of outliers on data distribution and improving the robustness of data preprocessing. By independently calculating entropy weights and comprehensive scores through dual channels, the contribution of vibration energy and vibration shock to NVH performance can be evaluated separately, providing a more accurate identification of different types of bearing defects compared to single-channel evaluation.

[0025] Furthermore, the specific method for establishing the "geometry-dynamic" mapping prediction model in step four is as follows: (1) Constructing training datasets Based on previous sample databases, construct datasets G4{Ciri,Wzi,Rai,Cire,Wze,Rae} containing the roundness, waviness, and roughness of the inner and outer raceways of the bearing, and dataset G5{Wzi,Rai,Wze,Rae,ec} containing the waviness, roughness, and lubricant contaminant coefficient of the inner and outer raceways of the bearing. (2) Constructing prediction models based on multiple linear regression a. Vibration energy prediction model, establishing the mapping relationship between dataset G4{Ciri,Wzi,Rai,Cire,Wze,Rae} and bearing vibration energy score Sv' based on multiple linear regression: in It is a constant; when all measured data for the current bearing are 0, the corresponding... Expected value; The error term includes unconsidered factors and random noise; b. The vibration and shock prediction model establishes a mapping relationship between the dataset G5{Wzi,Rai,Wze,Rae,ec} and the bearing vibration and shock score Sz' based on multiple linear regression. Similarly: in It is a constant; when all measured data for the current bearing are 0, the corresponding... Expected value; The error term includes unconsidered factors and random noise; (3) Output the predicted value. Using the multiple linear regression model constructed in step (2) above for vibration energy and vibration impact, input the parameters of each index in the bearing dataset G1 collected in step one to obtain the predicted scores Sv' and Sz' of bearing vibration energy and vibration impact.

[0026] In this embodiment, 30 samples are divided into a training set (21 samples) and a test set (9 samples) in a 7:3 ratio, and the regression coefficients are solved using the least squares method. The goodness of fit of the vibration energy prediction model is then obtained after training. Goodness of fit of vibration and shock prediction model The average prediction error on the test set was less than 5%, indicating that the model has good prediction accuracy and generalization ability. This step, by constructing prediction models for vibration energy and vibration impact respectively, accurately established the mapping relationship between different geometric parameters and different vibration characteristics. The multiple linear regression model has high computational efficiency and strong interpretability, and can clearly quantify the influence of each geometric parameter on NVH performance, providing a theoretical basis for subsequent residual diagnosis and bearing optimization design.

[0027] Furthermore, the full-scene dynamic residual diagnosis and analysis in step five are as follows: (1) Calculate the residual, which is calculated using the following formula: The formula for calculating the vibration energy residual is: The formula for calculating vibration and impact residuals is: (2) The diagnostic logic adopts a four-quadrant diagnostic logic, which calculates the residual between the observed value and the predicted value. , The bearing condition is precisely classified into four categories: dynamic instability, steady-state imbalance, hidden impact, and excellent matching, as follows: Scenario 1: Double positive residuals, dynamic instability type, not only poor geometric parameters, but also defects in assembly and lubrication, seriously unqualified; Scenario 2: Only positive energy residuals, steady-state imbalance type, the bearing's macroscopic motion is unstable, moderately unqualified; Scenario 3: Only positive impact residuals, hidden impact type, high-quality transient impact, seriously unqualified; Scenario 4: Double negative or zero residuals, excellent matching type, both geometric parameters and assembly are very good, qualified product.

[0028] This embodiment determines the residual threshold based on the 3σ principle: , That is, when When it is determined to be a positive energy residual, when The time was determined to be a positive impact residual. Calculations were performed on a specific bearing under test. , This is a hidden impact-type defect. This step, by calculating the residual between the actual score and the predicted score, can distinguish between inherent defects caused by geometric parameters and non-inherent defects caused by process factors such as assembly and lubrication. The four-quadrant diagnostic logic realizes the accurate classification and severity assessment of bearing defects, solves the problem that existing technologies cannot locate the root cause of defects, and provides a clear direction for bearing rework and process improvement.

[0029] Furthermore, the construction of the nonlinear comprehensive evaluation function in step six is ​​as follows: (1) Calculation of basic score a. Reverse normalization, the higher the score, the better. b. The formula for calculating the basic score is: For high-speed deep groove ball bearings, the effects of vibration acceleration and NVH are more significant, therefore, it is specified in the calculation that... , (2) Determine the consistency correction coefficient K; Based on the diagnostic results in step five, construct a scoring reward and punishment mechanism: a. Dynamic instability type, K=0.3, moderate punishment; b. Steady-state imbalance type, K=0.7, mild punishment; c. Hidden impact type, K=0.1, severe punishment; d. Excellent matching type, K=1.1, reward (3) Calculate the final score. .

[0030] In this embodiment, the extreme values ​​are determined based on the statistical results of 1000 qualified bearing samples: , , , The calculations for the above-mentioned concealed impact bearings yielded... , Basic score After multiplying by the correction factor K=0.1, the final score is... This step converts the scores into an intuitive form of "higher scores mean better performance" through reverse normalization, which aligns with evaluation practices in engineering applications. Differentiated weighting coefficients are set based on the characteristics of high-speed deep groove ball bearings, highlighting the critical impact of vibration and shock on NVH performance. The introduction of a consistency correction coefficient to construct a scoring reward and penalty mechanism further reflects the differences in the impact of different defect types on the overall bearing performance, improving the accuracy and rationality of the comprehensive evaluation results.

[0031] Furthermore, the grading method in step seven is as follows: Premium grade products: Premium grade: Qualified products: Defective products: .

[0032] This embodiment combines the industry standard JB / T10336-2020 "Methods for Measuring Vibration (Velocity) of Rolling Bearings" and enterprise quality requirements to determine the grade threshold: , , The aforementioned concealed impact bearing scored 0.0778 (<60), classifying it as a non-conforming product and recommending its scrapping. This step, by setting clear grade thresholds, transforms the quantified final score into a standardized quality grade, facilitating product grading management and quality control for enterprises. Furthermore, by combining defect analysis and rework recommendations from the diagnostic report, a complete closed loop from evaluation to decision-making is achieved, enhancing the method's engineering practical value.

[0033] This embodiment can also adjust the evaluation parameters according to different application scenarios: for ultra-high precision bearings used in aerospace, the grade threshold can be increased by 10%, and evaluation indicators such as vibration displacement and sound pressure level can be added; for bearings used in general machinery, the sample size requirement can be appropriately reduced to m≥15 to improve evaluation efficiency. In addition, this method can further improve the accuracy of the prediction model by increasing the number of training samples and introducing machine learning algorithms (such as random forest and support vector machine), making it suitable for NVH evaluation of rolling bearings of different types and specifications.

[0034] In summary, this invention employs a technical solution that combines multi-dimensional index fusion, dual-channel independent entropy weight evaluation, "geometric-dynamic" mapping prediction, and residual diagnosis. This solution addresses the technical problems of existing bearing NVH evaluation methods, such as single-dimensional evaluation, subjective weight allocation, lack of quantitative comprehensive evaluation, and insufficient correlation analysis. It achieves a comprehensive, accurate, and scientific quantitative evaluation of the NVH performance of high-speed deep groove ball bearings, accurately pinpointing defect types and root causes, and outputting a complete diagnostic report including grade conclusions and rework suggestions. This provides solid technical support for the quality control, process improvement, and optimized design of high-end bearings, and has significant engineering application value.

[0035] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A comprehensive evaluation method for NVH of high-speed deep groove ball bearings, characterized in that: Includes the following steps: Step 1, Multidimensional data acquisition and preprocessing: Collect external multidimensional data, and construct a multidimensional evaluation index system based on the multidimensional data, including inner and outer rolling roundness, waviness and roughness, vibration velocity, vibration acceleration, kurtosis and crest factor, in order to construct the corresponding dataset; Step 2, construct the "energy-impact" dual-channel feature matrix: the reciprocal of the displacement with respect to time during bearing rotation is used as an indicator to measure the energy of bearing vibration, and it is named channel A; the distribution probability of the amplitude and the peak intensity of the velocity with respect to time during bearing rotation are used as indicators to measure the impact of bearing vibration, and the two-channel A / B feature matrix is ​​constructed based on the dataset constructed in step 1. Step 3: Perform independent entropy weight evaluation within the dual channels: First, based on robust normalization using the interquartile range, then calculate the information entropy e. j The system considers information utility value and weight, and finally calculates a comprehensive score. The higher the comprehensive score, the worse the assessment. Step 4: Establish a "geometry-dynamic" mapping prediction model; Step 5: Perform dynamic residual diagnosis and analysis across the entire scenario based on the prediction model established in Step 4; Step 6: Construct a nonlinear comprehensive evaluation function to calculate the final score; Step 7: Based on the final score calculated in Step 6, classify and output the grades, which are divided into top-grade products, superior products, qualified products, and unqualified products. The final output is a bearing NVH diagnostic report. This bearing NVH diagnostic report analyzes whether the bearing may have defects based on the original collected data set and the calculated residuals. For the diagnosed bearings with minor defects that can be repaired, reasonable rework suggestions are given; and the bearing NVH grade classification conclusion is also provided.

2. The comprehensive evaluation method for NVH of high-speed deep groove ball bearings according to claim 1, characterized in that: The specific method for multidimensional data acquisition and preprocessing in step one is as follows: First, measure the bearing's geometric parameters and construct an initial signal set G1{Ciri,Wzi,Rai,Cire,Wze,Rae,ec} based on the bearing's inner and outer ring rolling roundness, waviness, roughness, and contaminant coefficient of the lubricating medium. Second, measure the bearing's vibration velocity index and construct a dataset G2{VL,VM,VH} based on the bearing's vibration velocity VL / VM / VH. Finally, measure the bearing's vibration acceleration, kurtosis, and crest factor to construct a dataset G3{Z,K,CF}.

3. The comprehensive evaluation method for NVH of high-speed deep groove ball bearings according to claim 2, characterized in that: The specific details of the "energy-impact" dual-channel feature matrix constructed in step two are as follows: Vibration energy channel A, using the vibration velocity measurement dataset G2 of the bearing obtained in step one, constructs a feature matrix XG2=[VL, VM, VH] that reflects the dynamic operating performance of the bearing; Vibration and impact channel B uses the bearing vibration velocity measurement dataset G3 obtained in step one to construct a feature matrix XG3=[Z, K, CF] that reflects the dynamic operating performance of the bearing. Among them, the aforementioned feature matrices XG2 and XG3 each have at least 9 rows and are at least 3 times the statistical dimension.

4. The comprehensive evaluation method for NVH of high-speed deep groove ball bearings according to claim 3, characterized in that: The specific method for conducting independent entropy weight evaluation within the dual channels in step three is as follows: (1) Based on robust normalization of interquartile range, specifically: a. For each column of data in the feature matrix of channel A and channel B, calculate the median Xmedian, the first quartile Q1, the third quartile Q3, and the interquartile range IQR = Q3 - Q1; b. Perform data normalization transformation Since Xmedian is the median, Z here must be negative, and it is centered around 0. c. To satisfy the basic conditions for subsequent calculations using the entropy weight method, it is necessary to... The value is shifted to the non-negative interval, i.e. Where min(Z) is the minimum value of data in b when the data is normalized. Given a minimum value, here defined as 10⁻⁶, a completely new matrix can be constructed. ; d. Calculate the specific gravity The newly constructed matrix Divide each data point in the table by the sum of the data in the column containing that value, and calculate the proportion of the i-th sample to that index under the j-th index. : (2) Calculate information entropy e j Calculate the information entropy of the j-th phase index: Where k is related to the number of samples m, it is defined as k = 1 / ln(m). Note that if Pij = 0, then Pij * ln(Pij) = 0. (3) Calculate the information utility value in The larger the value, the higher the distinguishability of the indicator, and the more important it is; (4) Calculate the weights Normalize the information utility value to obtain the final weight of the j-th indicator: (5) Calculate the overall score Obtain bearing vibration energy scores based on feature matrices XG2 and XG3. and vibration and impact scores : , The evaluation is based on a comprehensive score; the higher the comprehensive score, the worse the evaluation is.

5. The comprehensive evaluation method for NVH of high-speed deep groove ball bearings according to claim 3, characterized in that: The specific method for establishing the "geometry-dynamic" mapping prediction model in step four is as follows: (1) Constructing the training dataset Based on previous sample databases, a dataset G4{C} was constructed that includes the roundness, waviness, and roughness of the raceways of the inner and outer rings of bearings. iri W zi ,R ai C ire W ze ,R ae } and a dataset G5{W that includes the waviness, roughness, and lubricant contaminant coefficients of the inner and outer ring raceways of the bearing. zi ,R ai W ze ,R ae ,e c }; (2) Constructing a prediction model based on multiple linear regression a. Vibration energy prediction model, based on multiple linear regression to establish a dataset G4{C iri W zi ,R ai C ire W ze ,R ae Mapping relationship between} and bearing vibration energy score Sv': in It is a constant; when all measured data for the current bearing are 0, the corresponding... Expected value; This is the error term, which includes unconsidered factors and random noise; b. Vibration and shock prediction model, based on multiple linear regression to establish a dataset G5{W zi ,R ai W ze ,R ae ,e c The mapping relationship between} and the bearing vibration impact score Sz' is similarly as follows: in It is a constant; when all measured data for the current bearing are 0, the corresponding... Expected value; This is the error term, which includes unconsidered factors and random noise; (3) Output predicted value Using the multiple linear regression model constructed in step (2) above for vibration energy and vibration impact, input the parameters of each index in the bearing dataset G1 collected in step one to obtain the predicted scores Sv' and Sz' of bearing vibration energy and vibration impact.

6. The comprehensive evaluation method for NVH of high-speed deep groove ball bearings according to claim 5, characterized in that: The full-scene dynamic residual diagnosis and analysis in step five are as follows: (1) Calculate the residual using the following formula: The formula for calculating the vibration energy residual is: ΔS v =S v -S v ' The formula for calculating vibration and impact residuals is: ΔS z =S z -S z ' (2) The diagnostic logic adopts a four-quadrant diagnostic logic, which calculates the residual ΔS between the observed value and the predicted value. v ,ΔS z The bearing condition is precisely classified into four categories: dynamic instability, steady-state imbalance, hidden impact, and excellent matching, as detailed below: Scenario 1: Double positive residuals, dynamic instability type, not only poor geometric parameters, but also defects in assembly and lubrication, etc., which are seriously unqualified; Scenario 2: Positive energy residual only; steady-state imbalance type, unstable macroscopic motion of the bearing, moderate non-compliance; Scenario 3: Only positive residual impact; concealed impact type, high-quality transient impact, severely unqualified; Scenario 4: Double negative or zero residuals; excellent matching type, good geometric parameters and assembly, qualified product.

7. The comprehensive evaluation method for NVH of high-speed deep groove ball bearings according to claim 6, characterized in that: The construction of the nonlinear comprehensive evaluation function in step six is ​​as follows: (1) Calculation of basic score a. Reverse normalization, the higher the score, the better. b. The formula for calculating the basic score is: S coer =a*S vf +β*S zf ; For high-speed deep groove ball bearings, the effects of vibration acceleration and NVH are more significant. Therefore, α=0.4 and β=0.6 are specified in the calculation. (2) The consistency correction coefficient K is determined; based on the diagnostic results in step five, a scoring and reward / penalty mechanism is constructed. a. Dynamically unstable type, K=0.3, moderate penalty; b. Steady-state imbalance type, K=0.7, mild penalty; c. Covert impact type, K=0.1, severe penalty; d. Excellent match, K=1.1, reward (3) Final score calculation, S core_final =K*S core .

8. The comprehensive evaluation method for NVH of high-speed deep groove ball bearings according to claim 7, characterized in that: The grading method in Step 7 is as follows: Special-grade product: S core_final ≥ T1, First-grade product: T2 ≤ S core_final < T1, Qualified product: T3 ≤ S core_final < T2, Unqualified product: S core_final < T3.

Images