High-precision bearing life prediction method, system, equipment and medium

Abstract

The present invention discloses a high-precision bearing life prediction method, system, device and medium, the method comprising the following steps: denoising the original vibration signal data of the bearing, and calculating the RMS feature of the denoised signal; performing a bearing state division operation based on the RMS feature and the PELT algorithm; extracting a time-frequency diagram corresponding to the bearing degradation feature based on the original vibration signal data of the bearing; constructing an input data set based on the bearing state division result and the time-frequency diagram; using the input data set to train a bearing life prediction model, and using the trained bearing life prediction model to predict the bearing life; the present invention can use a bearing state division operation with a relatively low amount of calculation, and at the same time use simplified steps to realize the construction of model input data, and accurately predict the bearing life through a prediction model based on the Informer deep learning framework.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology. Specifically, the present invention is applied to the field of workpiece life prediction, and in particular to a high-precision bearing life prediction method, system, equipment and medium. Background Art

[0002] Bearings are common parts of mechanical equipment, and the quality of bearings is also a key factor affecting the industrial production and manufacturing process. Once a bearing fails, it may cause the equipment to stop running and affect production efficiency, especially in special industries (such as aerospace and rail transportation), which will bring other safety hazards.

[0003] In order to solve the above problems, bearings are often maintained regularly. In order to achieve the best maintenance effect, it is necessary to avoid excessive or insufficient maintenance. Therefore, the existing technology usually adopts a bearing life prediction scheme to perform preventive maintenance before bearing failure occurs. The specific scheme is as follows:

[0004] Existing solution 1: In the rolling bearing life analysis based on K-means degradation identification and random forest, a new method based on PCA-Kmeans-RF is proposed. This solution constructs a more accurate degradation label for the training of the bearing remaining life prediction model through steps such as noise reduction, feature extraction and fusion, and clustering to determine the degradation points;

[0005] Existing Solution 2: In the prediction of the remaining life of rolling bearings based on the locust optimized LSTM network, the VMD method is used to decompose and reconstruct the bearing vibration signal, and the extracted features are used as the input of the model. The locust optimization algorithm (GOA) is introduced into the traditional LSTM bearing life prediction model for parameter optimization;

[0006] Existing solution three: In the prediction of the remaining life of bearings based on one-dimensional noise enhanced convolutional neural network, in order to avoid the previous complex bearing data preprocessing process, an unconditional noise injection method is used to improve the training speed and prediction accuracy of the network model, and the obtained one-dimensional noise enhanced convolutional neural network model (1DNECNN) is used for the prediction of the remaining life of bearings;

[0007] Existing Solution 3: Chinese patent CN118332897A discloses a bearing life prediction method of a convolutional neural network that integrates multi-scale and attention mechanisms. In order to enhance the feature extraction and learning capabilities of the model, this solution introduces an attention mechanism into the convolutional neural network (CNN);

[0008] Existing Solution 4: Chinese patent CN118228397A discloses a bearing life assessment method, electronic equipment and storage medium. This solution uses the maximum likelihood estimation method to fit the life test data of new bearing parts and service bearing parts respectively, and finally obtains the corresponding formula and corresponding curve, and applies it to the prediction of the remaining service life of the bearing.

[0009] After analyzing the above existing solutions, it is found that the current bearing life prediction solutions have the following problems:

[0010] (1) The current scheme lacks a convenient and effective method to extract bearing degradation characteristics.

[0011] (2) The current solution requires a large amount of data support, has high computational complexity, and consumes a lot of computing resources.

[0012] (III) The current scheme does not consider the differences in bearing degradation characteristics at different stages when predicting bearing life, resulting in low prediction accuracy. Summary of the invention

[0013] The object of the present invention is to provide a high-precision bearing life prediction method, system, device and medium, thereby solving all or one of the above-mentioned problems existing in the prior art.

[0014] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0015] In one aspect, the present invention provides a high-precision bearing life prediction method, comprising the following steps:

[0016] Bearing state division step: configuring the original vibration signal data of the bearing, denoising the original vibration signal data of the bearing, and calculating the RMS characteristics of the denoised signal; performing the bearing state division operation based on the RMS characteristics and the PELT algorithm to obtain the bearing state division result;

[0017] Degradation feature extraction step: extracting a time-frequency diagram corresponding to the bearing degradation feature based on the original vibration signal data of the bearing;

[0018] Prediction model building steps: construct an input data set based on the bearing state division results and the time-frequency diagram; construct an informer-based bearing life prediction model; use the input data set to train the bearing life prediction model, and use the trained bearing life prediction model to predict the bearing life.

[0019] As an improved solution, the denoising process includes:

[0020] Performing de-averaging processing on the original vibration signal data of the bearing;

[0021] Performing Hilbert transform on the signal after the mean value removal process to obtain a signal envelope value.

[0022] As an improved solution, the calculating of the RMS characteristic of the denoised signal includes:

[0023] Substituting the signal envelope value into the RMS characteristic calculation formula to obtain the RMS characteristic;

[0024] The RMS characteristic calculation formula is: ;in, is the root mean square value, is the signal envelope value.

[0025] As an improved solution, the bearing state division operation includes:

[0026] Establish the relationship inequality between the time point and the optimal cost function;

[0027] The pruning strategy of the PELT algorithm is called to optimize the relational inequality;

[0028] The candidate change point positions in the RMS feature are removed according to the optimized relationship inequality to obtain the bearing state division result based on RMS.

[0029] As an improved solution, the extracting of a time-frequency diagram corresponding to the bearing degradation characteristics based on the original bearing vibration signal data includes:

[0030] Performing continuous wavelet transform on the original vibration signal data of the bearing to obtain the time-frequency diagram;

[0031] The formula of the continuous wavelet transform is: ;in, is the input signal; is the mother wavelet function; a is the scale parameter; b is the translation parameter; The superscript indicates the complex conjugate.

[0032] As an improved solution, the input data set is constructed based on the bearing state division result and the time-frequency diagram, including:

[0033] Labeling each of the time-frequency graphs according to the bearing state classification result and the corresponding value of the remaining service life of the bearing;

[0034] Using the annotated time-frequency diagram as the input data set of the bearing life prediction model;

[0035] The marked label value is used as the basis for the segmented fitting of the bearing life prediction model and the prediction target value.

[0036] As an improved solution, the construction of the informer-based bearing life prediction model includes:

[0037] The informer-based bearing life prediction model is constructed based on multi-head attention mechanism, probabilistic sparse self-attention mechanism and Distilling mechanism.

[0038] On the other hand, the present invention also provides a high-precision bearing life prediction system, comprising:

[0039] Bearing condition classification module for:

[0040] Configuring the original vibration signal data of the bearing, performing denoising on the original vibration signal data of the bearing, and calculating the RMS characteristics of the denoised signal; performing a bearing state division operation based on the RMS characteristics and the PELT algorithm to obtain a bearing state division result;

[0041] Degradation feature extraction module, used for:

[0042] Extracting a time-frequency diagram corresponding to the bearing degradation characteristics based on the original bearing vibration signal data;

[0043] Prediction model building module for:

[0044] Based on the bearing state division result and the time-frequency diagram, an input data set is constructed; an informer-based bearing life prediction model is constructed; the input data set is used to train the bearing life prediction model, and the trained bearing life prediction model is used to predict the bearing life.

[0045] On the other hand, the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the high-precision bearing life prediction method are implemented.

[0046] On the other hand, the present invention further provides a computer device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; wherein:

[0047] The memory is used to store computer programs;

[0048] The processor is used to execute the steps of the high-precision bearing life prediction method by running the program stored in the memory.

[0049] The beneficial effects of the technical solution of the present invention are:

[0050] 1. The high-precision bearing life prediction method described in the present invention can realize the bearing state division operation with low computational complexity, and at the same time adopts simplified steps to realize the construction of model input data. Finally, the bearing life is accurately predicted through the prediction model based on the Informer deep learning framework, forming an efficient, accurate, lightweight and highly versatile bearing life prediction strategy, which has strong reliability and strong comprehensive performance, makes up for the defects of the existing technology, and has high application value.

[0051] 2. The high-precision bearing life prediction system described in the present invention can realize the high-precision bearing life prediction method described in the present invention through the mutual cooperation of system modules.

[0052] 3. The computer-readable storage medium described in the present invention can realize the cooperation of the guidance system module, thereby realizing the high-precision bearing life prediction method described in the present invention, and the computer-readable storage medium described in the present invention also effectively improves the operability of the high-precision bearing life prediction method.

[0053] 4. The computer device described in the present invention can store and execute the computer-readable storage medium, thereby realizing the high-precision bearing life prediction method described in the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 is a flow chart of the high-precision bearing life prediction method described in Example 1 of the present invention;

[0056] Figure 2 It is a detailed flow chart of the high-precision bearing life prediction method described in Example 1 of the present invention;

[0057] Figure 3 Schematic diagram of the RMS effect of the high-precision bearing life prediction method described in Example 1 of the present invention;

[0058] Figure 4 It is a schematic diagram of the bearing state classification result based on RMS in the high-precision bearing life prediction method described in Example 1 of the present invention;

[0059] Figure 5is a schematic diagram of a time-frequency diagram of bearing degradation characteristics in the high-precision bearing life prediction method described in Example 1 of the present invention;

[0060] Figure 6 It is a schematic diagram of the structure of the bearing life prediction model in the high-precision bearing life prediction method described in Example 1 of the present invention;

[0061] Figure 7 is a schematic diagram of the architecture of a high-precision bearing life prediction system according to Embodiment 2 of the present invention;

[0062] Figure 8 is a schematic diagram of the structure of the computer device described in Example 4 of the present invention;

[0063] The symbols in the accompanying drawings are explained as follows:

[0064] 1501, processor; 1502, communication interface; 1503, memory; 1504, communication bus. DETAILED DESCRIPTION

[0065] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0066] In the description of the present invention, it should be noted that the embodiments described in the present invention are only part of the embodiments of the present invention, rather than all of the embodiments; based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.

[0067] The terms "first", "second", etc. in the specification and claims of this article and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of this article described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0068] In the description of the present invention, it should be noted that the PELT algorithm is a change point detection method; the algorithm uses dynamic programming and pruning technology to effectively identify structural changes in sample data while considering computational efficiency, detect change points in time series sample data, implement segmented operations on samples, and can process larger data sets.

[0069] In the description of the present invention, it should be noted that Informer is a time series prediction model that optimizes the Transformer architecture, which is specially designed for processing long sequence data and is suitable for processing bearing life prediction tasks.

[0070] Embodiment 1: This embodiment provides a high-precision bearing life prediction method, such as Figure 1 to Figure 6 As shown, the following steps are included:

[0071] S1, configuring the original vibration signal data of the bearing; performing denoising on the original vibration signal data of the bearing; calculating the RMS characteristics of the denoised signal; performing a bearing state division operation based on the RMS characteristics and the PELT algorithm to obtain a bearing state division result, the specific steps are as follows:

[0072] S101. Configure data set:

[0073] In this step, a bearing vibration signal data set is configured for the entire process from the start of the bearing operation to the occurrence of a fault, and the data set can be collected under different working conditions to improve the comprehensiveness of the data set;

[0074] Optionally, the IEEE PHM Challenge 2012 public bearing data set is used. The data set contains bearing vibration signals collected in a laboratory environment, recording the entire process from the start of bearing operation to failure. Three working conditions are considered, and each working condition provides two bearing vibration signals over the entire life cycle. The data set is used for subsequent life prediction model training. The data set is shown in Table 1:

[0075] Table 1:

[0076] Working conditions Radial force / N Speed / (r·min-1) Training set Test Set Condition 1 4000 1800 Bearing1_1, Bearing1_2 Bearing1_3, Bearing1_4, Bearing2_5, Bearing2_6, Bearing2_7 Condition 2 4200 1650 Bearing2_1, Bearing2_2 Bearing2_3, Bearing2_4, Bearing2_5, Bearing2_6, Bearing2_7 Condition 3 5000 1500 Bearing3_1, Bearing3_2 Bearing3_3

[0077] S102, Classification of bearing status:

[0078] In this step, since the noise and non-stationary components in the original vibration signal will affect the judgment of the bearing fault characteristics, it is necessary to perform Hilbert transform on the original vibration signal to obtain the envelope value of the signal to improve the data validity; then the root mean square value feature is calculated based on the envelope value of the signal. This feature has good monotonicity and robustness and can be used to evaluate the energy and health status of the vibration signal; finally, based on the root mean square value feature, an accurate linear time algorithm is used to accurately divide the bearing status;

[0079] Specifically, the original vibration signal is subjected to Hilbert transform, including:

[0080] (1) De-mean the data set to eliminate the DC component in the signal, so that the signal is more focused on its dynamically changing part:

[0081] ;

[0082] in: is the signal after removing the mean; is the original signal;

[0083] (2) Perform Hilbert transform based on the signal after de-averaging:

[0084] ;

[0085] ;

[0086] ;

[0087] in: is an analytical signal, including the real and imaginary parts; is the Hilbert transform; PV is the Cauchy principal value integral; is the envelope of the signal.

[0088] Specifically, the calculation of the root mean square (RMS) feature includes:

[0089] ;

[0090] Among them: draw the RMS image and do the normalization effect, such as Figure 3 As shown in the figure, the bearing shows different degradation trends in different degradation stages, which can be divided into normal operation stage, fatigue decline stage and rapid failure stage;

[0091] Specifically, the bearing state division operation based on the PELT (Exact Linear Time) algorithm includes:

[0092] (1) Assume that there are m change points in one observation segment, then the set of change points is , divide the data sample into m+1 segments through m change points, and define the [i:j]th segment as ; Based on the above settings, the solution of the change point problem is converted into the calculation of the minimum value of the following formula:

[0093] ;

[0094] in: is the cost function of the [i: j] segment; β is the penalty parameter used to control the number of change points;

[0095] (2) The statistical characteristics of the data sample correspond to the selected cost function. Therefore, this method uses the mean as the statistical characteristic, so the loss function is as follows:

[0096] ;

[0097] in: is the mean of segment [ i : j ];

[0098] (3) Dynamic programming is used to recursively calculate the optimal change point position:

[0099] (3.1) Let F(t) be the optimal cost function before time point t (i.e., the minimized loss), let F(k) be the optimal cost function before time point k, and let the algorithm goal be to determine the set of change point positions that minimizes the following formula, then:

[0100] ;

[0101] (3.2) The pruning strategy of the PELT algorithm (breakpoint detection algorithm based on dynamic programming) is used to reduce the candidate change point locations as much as possible and reduce the amount of calculation. The strategy is as follows:

[0102] ;

[0103] Among them: if the above inequality does not hold for a candidate change point position s and time point t, then the candidate change point position s is removed and it is judged that it cannot be part of the optimal solution, thus implementing the pruning operation;

[0104] (3.3) Based on the above recursive calculation, the result of the bearing state stage division based on RMS is obtained, as follows: Figure 4 As shown in the figure, the bearing degradation stages are automatically divided, corresponding to the three stages of the actual bearing degradation process.

[0105] S2. Extracting a time-frequency diagram corresponding to the bearing degradation characteristics based on the original bearing vibration signal data, the specific steps are as follows:

[0106] S201. Extract bearing degradation features:

[0107] In this step, the original horizontal vibration signal of the bearing is transformed by wavelet to obtain the time-frequency diagram (i.e. Figure 5 ), the time-frequency diagram is used as the input of the bearing life prediction model; in the wavelet transform, a specific mother wavelet is selected for translation and scaling processing to generate the corresponding basis function, and the basis function is used to approximate the signal to achieve local time-frequency analysis of the signal, thereby capturing the instantaneous changes in the signal;

[0108] Specifically, the continuous wavelet transform (CWT) is as follows:

[0109] ;

[0110] in: is the input signal; is the mother wavelet function; a is the scale parameter related to frequency; b is the translation parameter related to time; * represents complex conjugate;

[0111] Specifically, this method selects "amor" in the Morlet wavelet as the mother wavelet, and its expression is:

[0112] ;

[0113] Specifically, see Figure 5 Taking the time-frequency diagram of Bearing2_2 as an example, it can be analyzed that most of the energy of the bearing in the normal operation stage is concentrated in the medium and low frequencies. With the increase of running time, its energy begins to diffuse to the high frequency and gradually forms enrichment in the high frequency. Then, it gradually evolves into a double-peak enrichment in the fatigue decay stage, and finally the energy in the rapid failure stage gradually explodes on a large scale.

[0114] S3, based on the bearing state classification result and the time-frequency diagram, construct an input data set; construct an informer-based bearing life prediction model; use the input data set to train the bearing life prediction model; use the trained bearing life prediction model to predict the bearing life, as follows:

[0115] S301. Use a bearing life prediction model to predict bearing life:

[0116] In this step, based on the aforementioned bearing state classification results and bearing degradation characteristics, a model input data set is constructed; an informer-based bearing life prediction model is constructed; the data set is used to train the model, and the trained model is used to predict the bearing life;

[0117] Specifically, the construction of the input dataset includes:

[0118] (1) Annotate each spectrum graph generated after wavelet transformation according to the results of the bearing degradation state classification and the corresponding values ​​of the remaining service life of the bearing;

[0119] (2) The annotated spectrum graph is used as the input of the bearing life prediction model, and the two annotated column label values ​​are used as the basis for the model segment fitting and the predicted target value;

[0120] Specifically, the construction of the bearing life prediction model includes:

[0121] (1) If Figure 6As shown in the figure, the Informer-based bearing life prediction model uses a multi-head attention mechanism to process attention in different dimensions in parallel, so as to better capture the long-term dependencies in complex time series;

[0122] (2) If Figure 6 As shown in the figure, the Informer-based bearing life prediction model uses the ProbSparse self-attention mechanism (probabilistic sparse self-attention mechanism) to sparse the attention score, significantly reducing the computational complexity and improving the efficiency of the model when processing long sequences;

[0123] (2.1) Among them, the ProbSparse self-attention mechanism introduces KL divergence to calculate the sparsity of the query component based on the traditional self-attention mechanism. The sparsity evaluation formula of the i-th query component is as follows:

[0124] ;

[0125] Among them: the first term is the calculation of Log-Sum-Exp (LSE) for all key vectors; the second term is the arithmetic mean; this formula is obtained by randomly selecting u ( ) dot product operations , reducing the complexity to 𝒪( ); then select the largest u query vectors and get the improved ProbSparse self-attention formula as follows:

[0126] ;

[0127] in: ; is the dimension of the input; is the filtered query vector, which contains only the first u key query vectors; select the query vector with the largest score ( ), calculate the attention scores of the first u key query vectors corresponding to each key value; for other query vectors, assign the value to the input mean of the self-attention layer; select the points with high attention scores to reduce the computational complexity.

[0128] (3) If Figure 6 As shown in the figure, the bearing life prediction model based on Informer uses the Distilling mechanism to extract key information, reduce redundant inputs, ensure model accuracy and efficiency, and improve the overall performance of the model;

[0129] (4) In an optional implementation, the bearing life prediction model is encoded based on Python 3.9. The model uses a Windows 64-bit operating system, 16G random access memory, a 12th Gen Intel(R) Core(TM) i7-12700H processor and an NVIDA GeForce RTX 3060 graphics processor. The key parameter settings of the model are shown in Table 2:

[0130] Table 2:

[0131] parameter Parameter Explanation value batch_size The number of batches of images to process each time 32 learning_rate Learning Rate 0.003 seq_len Input sequence length (image size 128*128) 128 out_len Output sequence length (predicted life span value) 1 factor Sparse attention sampling factor 5 epochs Iterations 50

[0132] Specifically, model training includes:

[0133] The above-mentioned spectrum graph is input into the Informer-based bearing life prediction model. The bearing life prediction model processes the long sequence of spectrum data based on the above-mentioned architecture, extracts the key features, and models the complex relationship between different time steps; finally, the bearing life prediction model predicts the remaining service life of the bearing based on the processed information.

[0134] Specifically, life expectancy prediction includes:

[0135] The test set data is input into the trained Informer-based bearing life prediction model to obtain the bearing life prediction results for the test set. The quality of the prediction results is judged based on MAE, RMSE, and MAPE, as follows:

[0136] ;

[0137] ;

[0138] ;

[0139] Where: n is the number of samples of test data; is the actual life value of the test data; It is the predicted value of life of the deep learning bearing life prediction method based on state division and time-frequency diagram features; the three indicators are inversely proportional to the model fitting effect, and the smaller the indicator, the better the model fitting effect.

[0140] In addition, this method was experimentally verified and its prediction effect was compared with the existing CNN-BiLSTM-attention network prediction method. This method has better performance in all three indicators. The effect comparison is shown in Table 3:

[0141] Table 3:

[0142] RUL prediction method MAE MSE RMSE informer 0.0631 0.0042 0.0466 Cnn-bilstm-atten 0.0798 0.0084 0.0582 Informer-State Division 0.0431 0.0027 0.0238 Cnn-bilstm-attention-state division 0.0587 0.0045 0.0356

[0143] In summary, this method adopts the PELT (Pruned Exact Linear Time) algorithm to divide the bearing state, directly uses the spectrum diagram obtained from the original vibration signal through wavelet transform as the model input, adopts the Informer deep learning framework to build a prediction model, and processes long sequence data through the sparse self-attention mechanism. Based on the above technical features, it reduces the amount of calculation, improves the calculation efficiency, simplifies the data preprocessing process, improves the data validity and enables the model to more accurately capture the bearing fault characteristics, thereby improving the comprehensive prediction performance of the model.

[0144] It should be noted that all the above examples are only for explaining the present invention and cannot limit the protection scope of the present invention.

[0145] Embodiment 2, this embodiment is based on the same inventive concept as the high-precision bearing life prediction method described in Embodiment 1, and provides a high-precision bearing life prediction system, such as Figure 7 As shown, including:

[0146] The bearing state division module is used to: configure the original vibration signal data of the bearing, perform denoising on the original vibration signal data of the bearing, and calculate the RMS characteristics of the denoised signal; perform the bearing state division operation based on the RMS characteristics and the PELT algorithm to obtain the bearing state division result;

[0147] A degradation feature extraction module is used to extract a time-frequency diagram corresponding to the bearing degradation feature based on the original vibration signal data of the bearing;

[0148] A prediction model building module is used to: construct an input data set based on the bearing state division result and the time-frequency diagram; construct an informer-based bearing life prediction model; use the input data set to train the bearing life prediction model, and use the trained bearing life prediction model to predict the bearing life.

[0149] Embodiment 3: This embodiment provides a computer-readable storage medium, including:

[0150] The storage medium is used to store computer software instructions used to implement the high-precision bearing life prediction method described in the above-mentioned Example 1, which includes a program for executing the above-mentioned program set for the high-precision bearing life prediction method; specifically, the executable program can be built into the high-precision bearing life prediction system described in Example 2, so that the high-precision bearing life prediction system can implement the high-precision bearing life prediction method described in Example 1 by executing the built-in executable program.

[0151] In addition, the computer-readable storage medium of this embodiment may adopt any combination of one or more computer-readable storage media, wherein the computer-readable storage medium includes electrical, optical, electromagnetic, infrared or semiconductor systems, devices or components, or any combination thereof.

[0152] Embodiment 4, this embodiment provides a computer device, such as Figure 8 As shown, the computer device may include: a processor 1501 , a communication interface 1502 , a memory 1503 and a communication bus 1504 , wherein the processor 1501 , the communication interface 1502 , and the memory 1503 communicate with each other via the communication bus 1504 .

[0153] Memory 1503, used for storing computer programs;

[0154] The processor 1501 is used to implement the steps of the high-precision bearing life prediction method described in the above-mentioned embodiment 1 when executing the computer program stored in the memory 1503.

[0155] As an embodiment of the present invention, the communication bus mentioned in the above terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0156] As an implementation mode of the present invention, the communication interface is used for communication between the above-mentioned terminal and other devices.

[0157] As an embodiment of the present invention, the memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0158] As an embodiment of the present invention, the above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0159] Different from the prior art, the present application adopts a high-precision bearing life prediction method, system, equipment and medium, which can perform bearing state division operations with lower computational complexity, and at the same time use simplified steps to realize the construction of model input data. Finally, the bearing life is accurately predicted through the prediction model based on the Informer deep learning framework, forming an efficient, accurate, lightweight and highly versatile bearing life prediction strategy, which has strong reliability and strong comprehensive performance, makes up for the defects of the prior art, and has high application value.

[0160] It should be understood that in the various embodiments of this document, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this document.

[0161] It should also be understood that in the embodiments of this article, the term "and / or" is only a description of the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0162] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this article.

[0163] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0164] In the several embodiments provided herein, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or can be electrical, mechanical or other forms of connection.

[0165] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of this article.

[0166] In addition, each functional unit in each embodiment of this invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional unit.

[0167] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this article is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of this article. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0168] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A high-precision bearing life prediction method, characterized in that: The following steps are involved: Bearing status classification steps: Configure the original vibration signal data of the bearing; Performing denoising processing on the original vibration signal data of the bearing; Calculating the RMS characteristic of the denoised signal; Perform a bearing state division operation based on the RMS feature and the PELT algorithm to obtain a bearing state division result; Degradation feature extraction steps: Extracting a time-frequency diagram corresponding to the bearing degradation characteristics based on the original bearing vibration signal data; Steps to build a prediction model: Based on the bearing state classification result and the time-frequency diagram, construct an input data set; Construct an informer-based bearing life prediction model; Using the input data set to train the bearing life prediction model; Using the trained bearing life prediction model to predict the bearing life; The denoising process includes: performing a de-meaning process on the original vibration signal data of the bearing; performing a Hilbert transform on the de-meaning signal to obtain a signal envelope value; The calculating of the RMS feature of the denoised signal includes: substituting the signal envelope value into an RMS feature calculation formula to obtain the RMS feature; the RMS feature calculation formula is: ;in, is the root mean square value, is the signal envelope value; The bearing state division operation includes: establishing a relational inequality between a time point and an optimal cost function; invoking a pruning strategy of a PELT algorithm to optimize the relational inequality; removing candidate change point positions in the RMS feature according to the optimized relational inequality to obtain a bearing state division result based on the RMS; The input data set is constructed based on the bearing state classification result and the time-frequency diagram, including: labeling each of the time-frequency diagrams based on the bearing state classification result and the corresponding value of the remaining service life of the bearing; using the labeled time-frequency diagram as the input data set of the bearing life prediction model; and using the labeled label value as the basis for the segmented fitting of the bearing life prediction model and the prediction target value.

2. The high-precision bearing life prediction method according to claim 1, characterized in that: The extracting of a time-frequency diagram corresponding to the bearing degradation characteristics based on the original bearing vibration signal data includes: Performing continuous wavelet transform on the original vibration signal data of the bearing to obtain the time-frequency diagram; The formula of the continuous wavelet transform is: ;in, is the input signal; is the mother wavelet function; a is the scale parameter; b is the translation parameter; The superscript indicates the complex conjugate.

3. The high-precision bearing life prediction method according to claim 1, characterized in that: The construction of the informer-based bearing life prediction model includes: The informer-based bearing life prediction model is constructed based on multi-head attention mechanism, probabilistic sparse self-attention mechanism and Distilling mechanism.

4. A high-precision bearing life prediction system, characterized in that: include: Bearing condition classification module for: Configuring the original vibration signal data of the bearing, performing denoising on the original vibration signal data of the bearing, and calculating the RMS characteristics of the denoised signal; performing a bearing state division operation based on the RMS characteristics and the PELT algorithm to obtain a bearing state division result; The bearing state classification module is further used to: perform de-averaging processing on the original vibration signal data of the bearing; Performing Hilbert transform on the signal after the mean value removal processing to obtain a signal envelope value; calculating the RMS feature of the signal after the denoising processing includes: substituting the signal envelope value into an RMS feature calculation formula to obtain the RMS feature; The RMS characteristic calculation formula is: ;in, is the root mean square value, is the signal envelope value; The bearing state classification module is further used to: establish a relational inequality between a time point and an optimal cost function; call the pruning strategy of the PELT algorithm to optimize the relational inequality; remove the candidate change point positions in the RMS feature according to the optimized relational inequality to obtain a bearing state classification result based on RMS; Degradation feature extraction module, used for: Extracting a time-frequency diagram corresponding to the bearing degradation characteristics based on the original bearing vibration signal data; Prediction model building module for: Based on the bearing state division result and the time-frequency diagram, construct an input data set; construct an informer-based bearing life prediction model; use the input data set to train the bearing life prediction model, and use the trained bearing life prediction model to predict the bearing life; The prediction model building module is also used to: mark each of the time-frequency diagrams based on the bearing state classification results and the corresponding values ​​of the remaining bearing service life; use the marked time-frequency diagrams as the input data set of the bearing life prediction model; and use the marked label values ​​as the basis for the segmented fitting of the bearing life prediction model and the prediction target value.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the high-precision bearing life prediction method described in any one of claims 1 to 3 are implemented.

6. A computer device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; wherein: The memory is used to store computer programs; The processor is used to execute the steps of the high-precision bearing life prediction method described in any one of claims 1 to 3 by running the program stored in the memory.

Images