A method for evaluating performance degradation of a rolling bearing
By combining dual-channel data acquisition and cyclic stationary rapid spectral correlation feature extraction with a coupled hidden Markov model, the timeliness and accuracy of rolling bearing performance degradation assessment were solved, enabling a comprehensive and scientific assessment of rolling bearing performance degradation.
Patent Information
- Application Number
- CN202210827524.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-13
- Publication Date
- 2026-02-10
- Estimated Expiration
- 2042-07-13
Smart Images
Figure CN115200865B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for evaluating the performance degradation of rolling bearings, belonging to the field of bearing performance evaluation technology. Background Technology
[0002] Rolling bearing performance degradation assessment technology can train a model to form a health baseline using only the data features of the bearing under normal operating conditions. Real-time data feature vectors are input into the training model, and the real-time health status of the equipment is judged by comparing its output value with the health baseline, thereby truly realizing intelligent prediction of the bearing's operating status. In bearing performance degradation assessment technology, three key factors determine the quality of the assessment: (1) effective feature extraction; (2) efficient and accurate intelligent algorithms; and (3) a robust definition of performance degradation indicators. Traditional time-domain and frequency-domain feature extraction methods often fail to effectively extract non-stationary and nonlinear feature vectors when rolling bearings fail. Currently, feature extraction mainly involves simply extracting the spectral features of the original vibration signal, which cannot reflect the weak fault stage of the rolling bearing, resulting in the inability to timely and accurately understand the bearing's condition in the early stage of performance degradation during subsequent degradation performance assessment. In addition, current degradation performance assessments are all conducted using single-channel data, which cannot fully reflect the bearing's operating status, leading to the feature bias and incompleteness caused by single-channel information acquisition during subsequent assessments, affecting the accuracy of the assessment. Summary of the Invention
[0003] The purpose of this invention is to provide a method for evaluating the performance degradation of rolling bearings, so as to solve the problem that current rolling bearing performance degradation evaluations cannot timely and accurately understand the bearing in the early stage of performance degradation.
[0004] To address the aforementioned technical problems, this invention provides a method for evaluating the performance degradation of rolling bearings. This evaluation method includes the following steps:
[0005] 1) Obtain vibration data of rolling bearings in each channel under normal operating conditions. Vibration data of rolling bearings under normal operating conditions refers to data that does not exceed the alarm threshold, or data that shows stable operation in historical trends.
[0006] 2) The vibration data of each channel under normal working conditions is segmented and processed. Cyclic stationary fast spectrum correlation features are extracted from each segment of data to obtain the fast spectrum correlation features of each channel. This forms a fast spectrum correlation map. The energy distribution values of the fast spectrum correlation map on the horizontal axis are summed as the feature vector of the rolling bearing under normal working conditions.
[0007] 3) Establish a coupled hidden Markov knowledge base model and train the coupled hidden Markov knowledge base model using the feature vectors of the rolling bearing under normal working conditions.
[0008] 4) Acquire the vibration data of the rolling bearing to be evaluated in real time, and extract the cyclic stationary fast spectrum correlation features to obtain the test feature vector. Input the vector into the trained coupled hidden Markov knowledge base model to obtain the performance degradation index curve of the rolling bearing to be evaluated.
[0009] This invention employs a cyclic stationary fast spectral correlation feature extraction method, which can more effectively extract the non-stationary and nonlinear features of rolling bearing signals during weak fault stages. This provides effective data support for subsequent performance degradation assessment, enabling accurate and timely understanding of the initial stage of rolling bearing performance degradation and providing a reliable data source for the normal operation of rolling bearings. Furthermore, coupled Hidden Markov Models are used to further fuse the cyclic stationary features, effectively improving the accuracy of performance degradation assessment.
[0010] Furthermore, the vibration data obtained in step 1) is dual-channel data, that is, two channels, one horizontal and one vertical, are arranged at each monitoring point to obtain horizontal vibration data and vertical vibration data; the coupled hidden Markov knowledge base model established in step 3) adopts a double-chain coupled hidden Markov.
[0011] This invention, by acquiring dual-channel signals from the same source as the bearing, can more comprehensively reflect the bearing's operating status, avoiding the bias and incompleteness of features caused by single-channel information acquisition in subsequent evaluation processes; and by employing a dual-chain coupled Hidden Markov model to further fuse cyclic stationary features, effectively improving the accuracy of performance degradation assessment.
[0012] Furthermore, the cyclic stationary fast spectral correlation feature extraction process in step 2) is as follows:
[0013] Perform a short-time Fourier transform on the segmented signal according to set parameters, including a window function. and its window length Maximum cycle frequency and frequency of use ;
[0014] The calculation formula used is as follows: Fast spectral correlation is performed based on the short-time Fourier transform results.
[0015]
[0016]
[0017]
[0018] in This is the result of the short-time Fourier transform. For discrete frequencies, The cycle frequency, This is a block shift of the short-time Fourier transform. , express conjugate, The center time index of the window function.
[0019] Furthermore, step 2) also includes summing the energy of the obtained fast spectrum as a step to extract the correlation features of the three-dimensional cyclic stationary fast spectrum.
[0020] Furthermore, the method also includes a step of normalizing the extracted three-dimensional cyclic stationary fast spectrum correlation features.
[0021] This invention eliminates the influence of size through normalization, facilitating subsequent analysis and calculation.
[0022] Furthermore, in step 4), the maximum likelihood probability output by the trained coupled hidden Markov knowledge base model is... As an indicator for evaluating the performance degradation of rolling bearings.
[0023] Furthermore, the method also includes underflow prevention processing for the obtained performance degradation evaluation index, using the following formula:
[0024]
[0025] in To prevent performance degradation evaluation metrics after underflow treatment, For data length, This represents the number of channels.
[0026] Through the above-described treatment, the present invention can prevent Data underflow is eliminated, and the effects of data length T and number of channels C are eliminated, improving the reliability of the metrics.
[0027] Furthermore, the method also includes processing the performance degradation evaluation index after the anti-overflow treatment to obtain the final performance index, which is:
[0028]
[0029] in This represents the final performance index of the bearing at time t, where β is a weighting coefficient with a value between 0 and 1. This is an evaluation index for the performance degradation of the bearing after anti-overflow treatment at time t.
[0030] This invention uses a moving average to process the performance degradation evaluation index after anti-overflow treatment, so that the final performance index can reflect the early performance degradation of the rolling bearing and reduce the number of false alarms when based on threshold alarms, thereby improving the accuracy of the evaluation.
[0031] Furthermore, the weighting coefficient β is 0.5. Attached Figure Description
[0032] Figure 1 This is a flowchart of the rolling bearing performance degradation assessment method of the present invention;
[0033] Figure 2-a This is a schematic diagram of the vertical channel signal collected under normal bearing operating conditions in this embodiment;
[0034] Figure 2-b This is a schematic diagram of the parallel channel signals collected under normal bearing operating conditions in this embodiment;
[0035] Figure 3 This is a schematic diagram of the performance degradation index curve of the rolling bearing throughout its entire life cycle according to the present invention.
[0036] Figure 4 This is a schematic diagram of the amplitude index degradation curve of the horizontal channel of a rolling bearing throughout its entire life cycle in the prior art;
[0037] Figure 5 This is a schematic diagram of the kurtosis degradation curve of rolling bearings in the horizontal direction throughout their entire life cycle in the prior art. Detailed Implementation
[0038] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings.
[0039] Examples of Rolling Bearing Performance Degradation Assessment Methods
[0040] This invention first acquires dual-channel vibration data of a rolling bearing under normal operating conditions; then, it segments the acquired dual-channel data under normal operating conditions and performs cyclic stationary fast spectrum correlation feature extraction on each segment to obtain homogeneous multi-channel feature vectors under normal operating conditions; these feature vectors are used as training vectors for a coupled hidden Markov model to train a knowledge base model of the rolling bearing under normal operating conditions; the vibration data of each channel of the rolling bearing to be evaluated is acquired in real time, and cyclic stationary fast spectrum correlation feature extraction is performed to obtain test feature vectors, which are then input into the trained coupled hidden Markov knowledge base model under normal operating conditions to obtain performance degradation index curves. The specific implementation process of this method is as follows: Figure 1 As shown, the specific implementation process is as follows.
[0041] 1. Obtain vibration data of rolling bearings under normal operating conditions.
[0042] This invention acquires vibration data by installing vibration sensors on rolling bearings. To more comprehensively reflect the bearing's operating state and avoid the bias and incompleteness caused by single-channel information acquisition during subsequent evaluation, this invention employs multi-channel information acquisition. This involves acquiring vibration data from different directions at the same monitoring point. For example, in this embodiment, two channels (horizontal and vertical) can be simultaneously arranged at the same monitoring point on the rolling bearing (two vibration sensors in different directions at the same monitoring point) to acquire dual-channel data under normal operating conditions at the same monitoring point, i.e., acquiring vibration data from different directions at the same monitoring point. According to the above acquisition method, the vibration data acquired in the vertical channel in this embodiment is as follows: Figure 2-a As shown, the vibration data of the horizontal channel are as follows: Figure 2-b As shown.
[0043] To ensure the accuracy of subsequent model training, this invention requires vibration data under normal operating conditions. Normal operating conditions refer to data that does not exceed the alarm threshold, or data that shows stable operation in historical trends.
[0044] 2. The acquired vibration data is segmented, and the cyclic stationary fast spectrum correlation features of each segment are extracted to obtain the feature vector of the rolling bearing under normal operating conditions.
[0045] Since the vibration data obtained in step 1 under normal operating conditions is multi-channel data, it is necessary to extract corresponding features for each channel separately. The feature extraction process is the same for each channel. First, the obtained vibration data under normal operating conditions needs to be divided into multiple segments according to time. For ease of calculation, this invention segments the vibration data with each segment having a length of 1024 sampling points. Then, cyclic stationary fast spectrum correlation feature extraction is used for each segment. The cyclic stationary fast spectrum correlation feature extraction process is as follows:
[0046] 1) Perform a short-time Fourier transform on each segment of data.
[0047] Initialization signal Short-time Fourier Transform window function and its window length Maximum cycle frequency Frequency of use Parameters, namely:
[0048]
[0049] In the formula and This indicates the block shift of the STFT.
[0050] 2) Extract fast spectral correlation features based on the short-time Fourier transform results.
[0051] initialization , , , , Calculate the fast spectral correlation using the center-time index of the window function:
[0052] For p=0,1,2,…,P
[0053]
[0054]
[0055]
[0056] End
[0057] For discrete frequencies, The cycle frequency, This is a block shift of the short-time Fourier transform. , express . conjugate.
[0058] 3) The obtained fast spectrum correlation features are simplified.
[0059] Step 2) yields fast spectrum correlation features for different channels. To make fault feature extraction more intuitive using these features, this invention simplifies the obtained fast spectrum correlation features for the two channels. The two channels' fast spectrum correlation features can form a two-dimensional fast spectrum correlation map B. This invention will simplify the two-dimensional fast spectrum correlation map... The summation of the energy distribution values on the horizontal axis is used as the simplified feature vector, expressed as:
[0060]
[0061] in These represent the start and cutoff frequencies of the frequency band, respectively.
[0062] 4) Normalize the simplified feature vectors.
[0063] The eigenvectors obtained in step 3) are normalized to eliminate the influence of size, facilitating subsequent analysis and calculation: Let The final normalized feature vector is as follows:
[0064]
[0065] 3. Establish a coupled hidden Markov knowledge base model and train the coupled hidden Markov knowledge base model using the feature vectors of the rolling bearing under normal operating conditions.
[0066] Since this invention acquires multi-channel vibration data of rolling bearings, a multi-chain Coupled Hidden Markov (CPMM) model is used to establish the Coupled Hidden Markov knowledge base model. For example, this embodiment collects dual-channel data; therefore, this invention uses a dual-chain CPMM to establish the CPMM knowledge base model to further integrate cyclic stationary features, thereby effectively improving the accuracy of performance degradation assessment.
[0067] 4. Real-time acquisition of vibration data of the rolling bearing to be evaluated, and extraction of cyclic stationary fast spectrum correlation features to obtain test feature vectors. Input these features into a trained coupled hidden Markov knowledge base model to obtain the performance degradation index curve of the rolling bearing to be evaluated.
[0068] First, real-time vibration data of the rolling bearing to be evaluated is acquired; this real-time vibration data is also dual-channel data. Then, following the process in step 2, the obtained real-time vibration data undergoes fast spectral correlation feature extraction, followed by simplification and normalization to obtain the corresponding feature vector. This feature vector is used as the test feature vector. The obtained test feature vector is then input into the coupled hidden Markov knowledge base model trained in step 3. This coupled hidden Markov knowledge base model outputs the corresponding maximum likelihood probability. The performance degradation can be assessed using this maximum likelihood probability.
[0069] To prevent maximum likelihood probability Data underflow is a concern. Before using this metric for evaluation, this invention requires data underflow prevention processing. In this embodiment, the underflow prevention processing can represent the maximum likelihood probability in logarithmic form and divide it by the data length T and the number of channels C to eliminate the influence of the data length T and the number of channels C, i.e.:
[0070]
[0071] in To prevent performance degradation evaluation metrics after underflow treatment, For data length, The number of channels is T = 1024 and C = 2 in this embodiment.
[0072] To further improve the reliability of the assessment, enabling it to reflect early performance degradation of the bearing and reduce the number of false alarms when using threshold-based alarms, this invention also improves the obtained indicators. Further weighted moving average processing is performed, and the processed index is used as the final performance degradation index of the rolling bearing.
[0073] The final performance degradation index of rolling bearings can be expressed as:
[0074]
[0075] in The value represents the bearing's performance index at time t, and β is a weighting coefficient with a value between 0 and 1. β represents the weighting coefficient of the weighted moving average to the current measurement value, and after repeated testing, it is set to 0.5 in this invention.
[0076] In this embodiment, the real-time vibration data of the rolling bearing to be evaluated is subjected to fast spectral correlation feature extraction. The extracted features are input into the knowledge base model trained in step 3. The knowledge base model outputs the corresponding maximum likelihood probability. The obtained maximum likelihood probability is then processed for underflow prevention and reliability. The final performance degradation index is obtained by connecting all the output performance degradation indices into a curve to form the bearing's full life cycle performance degradation curve, such as... Figure 3 As shown. Figure 4 The amplitude index degradation curve of the horizontal channel of the rolling bearing throughout its entire life cycle is presented in the prior art, and the amplitude is used as the performance degradation index. Figure 5 This section presents the kurtosis degradation curves for the entire lifespan of rolling bearings in the horizontal direction in existing technologies, using kurtosis as a performance degradation index. (Comparison) Figure 3 and Figure 4 , Figure 5 Compared with traditional evaluation indicators, the present invention can more effectively reflect the three degradation stages of bearings: normal operation, minor faults, and complete failure, and can more accurately and comprehensively evaluate the degradation performance of rolling bearings.
[0077] This invention acquires dual-channel signals from the same source in the bearing, which can more comprehensively reflect the bearing's operating status and avoid the feature bias and incompleteness caused by single-channel information acquisition in subsequent evaluation processes. It employs a cyclostationary fast spectral correlation feature extraction method, which can more effectively extract the non-stationary and nonlinear features of the rolling bearing's weak fault stage signals, providing effective data support for subsequent performance degradation assessment. Simultaneously, it uses a double-chain coupled Hidden Markov Model to further fuse the cyclostationary features, effectively improving the accuracy of performance degradation assessment. Therefore, this invention, compared to traditional amplitude and kurtosis indices, more scientifically and effectively reflects the performance degradation process of rolling bearings.
[0078] The rolling bearing performance degradation assessment method of the present invention can be implemented by a computer program. For example, it can be designed as a rolling bearing performance degradation assessment device, including a processor and a memory. The processor executes the computer program stored in the memory to implement the method of the above-described method embodiments of the present invention. That is, the method in the above method embodiments should be understood to be a flow of rolling bearing performance degradation assessment method implemented by computer program instructions. These computer program instructions can be provided to the processor, such that the processor executes these instructions to produce the functions specified for implementing the above-described method flow.
[0079] In this embodiment, the processor refers to a processing device such as a microprocessor (MCU) or a programmable logic device (FPGA); the memory refers to a physical device used to store information, which typically involves digitizing the information and then storing it using media that utilizes electrical, magnetic, or optical methods. Examples include: various types of memory that store information using electrical energy, such as RAM and ROM; various types of memory that store information using magnetic energy, such as hard disks, floppy disks, magnetic tapes, magnetic core memory, bubble memory, and USB flash drives; and various types of memory that store information using optical methods, such as CDs or DVDs. Of course, there are other types of memory, such as quantum memories and graphene memories.
[0080] The device consisting of the aforementioned memory, processor, and computer program is implemented in a computer by the processor executing the corresponding program instructions. The processor can run various operating systems, such as Windows, Linux, Android, and iOS.
[0081] As an alternative implementation, the device may also include a display for showing the selection results for staff reference.
[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Many changes and modifications can be made without departing from the scope of the invention. Therefore, the above detailed description is intended to be illustrative rather than restrictive, and it should be understood that the following claims (including all equivalents) are intended to define the spirit and scope of the invention. These embodiments should be understood as illustrative only and not intended to limit the scope of protection of the invention. After reading the description of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent changes and modifications also fall within the scope defined by the claims of this invention.
Claims
1. A method for evaluating the performance degradation of rolling bearings, characterized in that, The evaluation method includes the following steps: 1) Obtain vibration data of rolling bearings in each channel under normal operating conditions. Vibration data of rolling bearings under normal operating conditions refers to data that does not exceed the alarm threshold, or data that shows stable operation in historical trends. 2) The vibration data of each channel under normal working conditions is segmented and processed. Cyclic stationary fast spectrum correlation features are extracted from each segment of data to obtain the fast spectrum correlation features of each channel. This forms a fast spectrum correlation map. The energy distribution values of the fast spectrum correlation map on the horizontal axis are summed as the feature vector of the rolling bearing under normal working conditions. 3) Establish a coupled hidden Markov knowledge base model and train the coupled hidden Markov knowledge base model using the feature vectors of the rolling bearing under normal working conditions. 4) Acquire the vibration data of the rolling bearing to be evaluated in real time, and extract the cyclic stationary fast spectrum correlation features to obtain the test feature vector. Input the vector into the trained coupled hidden Markov knowledge base model to obtain the performance degradation index curve of the rolling bearing to be evaluated.
2. The method for evaluating the performance degradation of rolling bearings according to claim 1, characterized in that, The vibration data obtained in step 1) is dual-channel data, that is, two channels, horizontal and vertical, are arranged at each monitoring point to obtain horizontal vibration data and vertical vibration data; the coupled hidden Markov knowledge base model established in step 3) adopts a double-chain coupled hidden Markov.
3. The method for evaluating the performance degradation of rolling bearings according to claim 2, characterized in that, The process of extracting cyclic stationary fast spectrum correlation features in step 2) is as follows: Perform a short-time Fourier transform on the segmented signal according to set parameters, including a window function. and its window length Maximum cycle frequency and sampling frequency ; The calculation formula used is as follows: Fast spectral correlation is performed based on the short-time Fourier transform results. ； ； ； in This is the result of the short-time Fourier transform. For discrete frequencies, The cycle frequency, This is a block shift of the short-time Fourier transform. , express conjugate, The center time index of the window function.
4. The method for evaluating the performance degradation of rolling bearings according to claim 3, characterized in that, Step 2) also includes summing the energy of the obtained fast spectrum as a step to extract the relevant features of the three-dimensional cyclic stationary fast spectrum.
5. The method for evaluating the performance degradation of rolling bearings according to claim 4, characterized in that, The method also includes a step of normalizing the extracted three-dimensional cyclic stationary fast spectrum correlation features.
6. The method for evaluating the performance degradation of rolling bearings according to claim 3, characterized in that, In step 4), the maximum likelihood probability output by the trained coupled hidden Markov knowledge base model is... As an indicator for evaluating the performance degradation of rolling bearings.
7. The method for evaluating the performance degradation of rolling bearings according to claim 6, characterized in that, The method also includes underflow prevention processing for the obtained performance degradation evaluation index, using the following formula: ； in To prevent performance degradation evaluation metrics after underflow treatment, For data length, This represents the number of channels.
8. The method for evaluating the performance degradation of rolling bearings according to claim 7, characterized in that, The method also includes processing the performance degradation evaluation index after the anti-overflow treatment to obtain the final performance index, which is: ； in This represents the final performance index of the bearing at time t, where β is a weighting coefficient with a value between 0 and 1. This is an evaluation index for the performance degradation of the bearing after anti-overflow treatment at time t.
9. The method for evaluating the performance degradation of rolling bearings according to claim 8, characterized in that, The weighting coefficient β is 0.5.