A method and system for predicting the remaining service life of a rolling bearing
By extracting the vibration and sound data characteristics of rolling bearings and combining with deep learning algorithms, the accuracy and complexity of traditional prediction methods are solved, and the accurate prediction and robustness of bearing life are achieved.
Patent Information
- Application Number
- CN202411053070.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-12-10
- Estimated Expiration
- 2044-08-02
AI Technical Summary
Traditional bearing life prediction methods are difficult to accurately consider small differences in working environment and load conditions, and rely on expert experience, it is difficult to meet the needs of large-scale equipment monitoring and maintenance, and vibration and sound signal processing is high in complexity, making it difficult to extract effective features.
The timing vibration and sound data of rolling bearings are collected, and the frequency and time-frequency characteristics are extracted through fast Fourier transform and wavelet transform. The prediction model is constructed based on the deep learning algorithm, and the real-time temperature data is corrected to achieve accurate prediction of the remaining service life of the bearing.
It improves the accuracy and availability of bearing life prediction, reduces manual participation, enhances the robustness of the model, can automatically learn complex relationships, and is suitable for prediction under complex operating conditions.
Smart Images
Figure CN118583498B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of bearing service life prediction, and in particular to a method and system for predicting the remaining service life of a rolling bearing. Background Art
[0002] Rolling bearings are key components in many mechanical equipment, and their normal operation is crucial to ensuring the performance and safety of mechanical equipment. Therefore, how to accurately predict the remaining service life of rolling bearings has become a core issue in the field of equipment maintenance. In the past, traditional bearing life prediction methods mainly relied on physical models and empirical formulas, such as the Weber model and the Pareto distribution. However, these methods often find it difficult to accurately consider and describe factors such as the working environment, load conditions, and slight differences in the manufacturing process of the bearing. Moreover, these models often seem to be incapable of early fault diagnosis and predictive performance under complex working conditions. In addition, traditional fault diagnosis methods usually require the experience and knowledge of experts, and the degree of manual participation is high, which makes it difficult to meet the needs of large-scale equipment monitoring and maintenance tasks.
[0003] At the same time, since bearings generate abundant vibration and sound signals during operation, these signals often contain important information about the health status of the bearings. For example, the wear and defects of the inner ring, outer ring or rolling element of the bearing will generate vibration signals of specific frequencies. However, how to effectively extract features useful for bearing life prediction from large-scale vibration and sound data remains an important and challenging problem. In addition, the working state of bearings is complex and changeable, and is affected by multiple factors at the same time, which makes the vibration and sound signals of bearings nonlinear and non-stationary, and puts higher requirements on signal processing and feature extraction. Therefore, how to use modern signal processing and deep learning methods to extract effective features from the vibration and sound signals of bearings and establish an accurate life prediction model is an important direction of current research.
[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention
[0005] The purpose of the present invention is to provide a method and system for predicting the remaining service life of a rolling bearing to solve the problems raised in the above-mentioned background technology.
[0006] To achieve the above object, the present invention provides the following technical solutions:
[0007] A method and system for predicting the remaining service life of a rolling bearing, the specific steps comprising:
[0008] Step 1: Collect the time-series vibration data and time-series sound data of a batch of rolling bearings in normal working condition, conduct long-term tracking observation on the batch of bearings, record the timestamp when the bearings turn into fault state, and obtain the remaining service life of each rolling bearing at the moment of collecting its vibration data and sound data;
[0009] Step 2: Apply fast Fourier transform to extract frequency features from time-series vibration data, and use wavelet transform to analyze time-series sound data to extract time-frequency features;
[0010] Step 3: Construct a rolling bearing remaining service life prediction model based on a deep learning algorithm, use the frequency characteristics and time-frequency characteristics of the rolling bearing as training sets, use the corresponding rolling bearing predicted remaining service life as labels, train the model, and obtain a trained rolling bearing remaining service life prediction model;
[0011] Step 4: Collect the vibration data and sound data of the rolling bearing to be evaluated in real time, perform fast Fourier transform on the vibration data to extract the frequency characteristics, perform wavelet transform on the sound data to extract the time-frequency characteristics, and input the processed data into the model to output the predicted remaining service life of the rolling bearing to be evaluated;
[0012] Step 5: Collect the temperature data of the rolling bearing to be evaluated in real time, and use the collected temperature data to correct the predicted remaining service life of the rolling bearing to be evaluated to obtain a comprehensive prediction index of the rolling bearing to be evaluated.
[0013] Furthermore, the process of collecting vibration data, sound data and temperature data of the rolling bearing in normal working state specifically includes:
[0014] Vibration sensors, sound sensors and temperature sensors are installed on the bearing device to collect data in real time. The vibration data, sound data and temperature data of the bearing are collected by the sensors. The data are all time series data and are divided into continuous time steps. Each time step represents a specific moment in the data, that is, each time step corresponds to a numerical value.
[0015] Furthermore, the process of using the frequency characteristics and time-frequency characteristics of the rolling bearing as a training set specifically includes:
[0016] Apply fast Fourier transform to extract frequency features from vibration data: perform fast Fourier transform on each time-series vibration signal segment to obtain frequency features corresponding to each time-series vibration signal segment, wherein the frequency features include time, fundamental frequency and harmonics; construct a data chain with the frequency features corresponding to the time-series vibration signal segment, i.e., a data chain composed of time, fundamental frequency and harmonics, wherein the pointer between time and fundamental frequency is a timestamp index, which points to the fundamental frequency value at a specific moment, and the pointer between fundamental frequency and harmonics is a fundamental frequency index, which points to the harmonic value at a specific fundamental frequency;
[0017] The method uses wavelet transform to analyze sound data and extract time-frequency features: performing wavelet transform on each time-series sound signal segment to obtain the time-frequency features corresponding to each time-series sound signal segment, wherein the frequency features include time, frequency and signal intensity; forming a data chain with the spectrum features corresponding to the time-series sound signal segment, i.e., a data chain composed of time, frequency and intensity, wherein the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency;
[0018] Based on the remaining service life of each rolling bearing at the time when its vibration data and sound data are collected, the data chain at the corresponding time is labeled to construct a training set.
[0019] Furthermore, the process of performing wavelet transform on the sound data specifically includes:
[0020] Select Molet wavelet for sound signal The continuous wavelet transform is based on the formula:
[0021] ;
[0022] in, are the wavelet coefficients, is time, that is, a variable in the time domain, used to represent the signal The values at different time points, is a scaled and translated version of the wavelet basis function, defined as:
[0023] ;
[0024] in, is a scale parameter, which is used to adjust the frequency of the wavelet basis function and controls the expansion and contraction of the wavelet basis function, that is, the frequency resolution. , the smaller For high frequency analysis, the larger Corresponding to low frequency analysis, The selection of appropriate parameters is determined through experiments and experience; is the translation parameter, which controls the translation of the wavelet basis function on the time axis, that is, the time resolution. is a real number, Continuously transforming within the time range of the signal, The selection of appropriate parameters is determined through experiments and experience. is the complex conjugate of the wavelet basis function.
[0025] Extracting the modulus maximum: In the modulus of wavelet transform Find the local maximum point in
[0026] Compute the energy distribution of wavelet coefficients at different scales and positions:
[0027] ;
[0028] in, Represents the energy of the wavelet coefficients at a specific scale a and position b, are the wavelet coefficients.
[0029] Furthermore, the process of constructing a rolling bearing remaining service life prediction model based on a deep learning algorithm specifically includes:
[0030] The collected historical vibration data and sound data of rolling bearings are subjected to fast Fourier transform and wavelet transform respectively, and the frequency features and time-frequency features are extracted. The data are used as training sets, and a rolling bearing remaining service life prediction model is constructed based on the deep learning algorithm. The corresponding rolling bearing predicted remaining service life is used as a label to train the model and obtain a trained rolling bearing remaining service life prediction model.
[0031] Furthermore, the remaining service life prediction model of the rolling bearing is used to obtain the predicted remaining service life of the rolling bearing to be evaluated. The process includes:
[0032] The vibration data and sound data of the rolling bearing to be evaluated are collected in real time, the frequency characteristics are extracted by fast Fourier transform of the vibration data, and the time-frequency characteristics are extracted by wavelet transform of the sound data. The processed data are input into the model, and the predicted remaining service life of the rolling bearing to be evaluated is output. ; The model includes an input layer, a fully connected layer, an activation function layer and an output layer;
[0033] Among them, the input layer is used to receive the processed vibration data and sound data; the fully connected layer is used to integrate the features of the previous layer and make the final prediction. When predicting the remaining service life, the fully connected layer maps the features to the prediction results; the activation function layer is used to introduce nonlinearity; and the output layer generates the final prediction results.
[0034] Furthermore, the predicted remaining service life of the rolling bearing to be evaluated is corrected using the collected temperature data to obtain a comprehensive prediction index of the rolling bearing to be evaluated. The process is based on the following formula:
[0035] ;
[0036] in, is the comprehensive prediction index of the rolling bearing to be evaluated, is the estimated remaining service life of the rolling bearing to be evaluated, is its preset scaling factor, and greater than zero, is the real-time temperature of the rolling bearing to be evaluated, The preset scale factor is is greater than zero, and , is the constant correction exponent.
[0037] A rolling bearing remaining service life prediction system, comprising:
[0038] The data acquisition module is used to collect the time-series vibration data and time-series sound data of a batch of rolling bearings in normal working condition, and conduct long-term tracking observation of the batch of bearings, record the timestamp when the bearings turn into fault state, and obtain the remaining service life of each rolling bearing at the moment of collecting its vibration data and sound data;
[0039] The feature extraction module applies fast Fourier transform to extract frequency features from time-series vibration data and uses wavelet transform to analyze time-series sound data to extract time-frequency features;
[0040] A model building module, which builds a rolling bearing remaining service life prediction model based on a deep learning algorithm, uses the frequency characteristics and time-frequency characteristics of the rolling bearing as a training set, uses the corresponding rolling bearing predicted remaining service life as a label, trains the model, and obtains a trained rolling bearing remaining service life prediction model;
[0041] The prediction module is used to collect the vibration data and sound data of the rolling bearing to be evaluated in real time, perform fast Fourier transform on the vibration data to extract the frequency characteristics, perform wavelet transform on the sound data to extract the time-frequency characteristics, and input the processed data into the model to output the predicted remaining service life of the rolling bearing to be evaluated;
[0042] The correction module is used to collect the temperature data of the rolling bearing to be evaluated in real time, and use the collected temperature data to correct the predicted remaining service life of the rolling bearing to be evaluated to obtain a comprehensive prediction index of the rolling bearing to be evaluated.
[0043] Compared with the prior art, the present invention has the following beneficial effects:
[0044] The present invention collects vibration data and sound data of rolling bearings, and uses signal processing methods to extract their frequency and time-frequency characteristics, which not only utilizes the characteristics of bearing faults in the frequency domain, but also takes into account the non-stationarity of sound signals. The prediction model constructed based on the deep learning algorithm can automatically learn the complex relationship between these characteristics and the remaining service life of the bearing, thereby improving the accuracy of the prediction. The bearing temperature data collected in real time is used to correct the prediction results, so that the model has better robustness in practical applications. In general, the present invention improves the technical problems of bearing remaining service life prediction and improves the accuracy and availability of the prediction by comprehensively utilizing vibration, sound and temperature data, as well as the powerful learning ability of deep learning. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] Figure 1 It is a schematic diagram of the overall method flow of the present invention;
[0046] Figure 2 It is a schematic diagram of the overall system module of the present invention. DETAILED DESCRIPTION
[0047] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.
[0048] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0049] Example:
[0050] See also Figure 1 , the present invention provides a technical solution:
[0051] A method for predicting the remaining service life of a rolling bearing, the specific steps comprising:
[0052] Step 1: Collect the time-series vibration data and time-series sound data of a batch of rolling bearings in normal working condition, conduct long-term tracking observation on the batch of bearings, record the timestamp when the bearings turn into fault state, and obtain the remaining service life of each rolling bearing at the moment of collecting its vibration data and sound data;
[0053] In this embodiment, the process of collecting vibration data, sound data and temperature data of a rolling bearing in a normal working state specifically includes: installing a vibration sensor, a sound sensor and a temperature sensor on the bearing device for real-time data collection, collecting the vibration data, sound data and temperature data of the bearing through the sensor, wherein the data are all time series data and are divided into continuous time steps, each time step represents a specific moment in the data, that is, each time step corresponds to a numerical value;
[0054] Dividing the data into continuous time steps, with each time step corresponding to a numerical value, helps with time series analysis, can reveal the changing trend of bearing performance over time, and provide a basis for predictive maintenance; directly collecting data through sensors can reduce human errors and improve data quality and accuracy.
[0055] Step 2: Apply fast Fourier transform to extract frequency features from vibration data, and use wavelet transform to analyze sound data and extract time-frequency features;
[0056] In this embodiment, the process of using the frequency characteristics and time-frequency characteristics of the rolling bearing as a training set specifically includes:
[0057] Apply fast Fourier transform to extract frequency features from vibration data: perform fast Fourier transform on each time-series vibration signal segment to obtain frequency features corresponding to each time-series vibration signal segment, wherein the frequency features include time, fundamental frequency and harmonics; construct a data chain with the frequency features corresponding to the time-series vibration signal segment, i.e., a data chain composed of time, fundamental frequency and harmonics, wherein the pointer between time and fundamental frequency is a timestamp index, which points to the fundamental frequency value at a specific moment, and the pointer between fundamental frequency and harmonics is a fundamental frequency index, which points to the harmonic value at a specific fundamental frequency;
[0058] The method uses wavelet transform to analyze sound data and extract time-frequency features: performing wavelet transform on each time-series sound signal segment to obtain the time-frequency features corresponding to each time-series sound signal segment, wherein the frequency features include time, frequency and signal intensity; forming a data chain with the spectrum features corresponding to the time-series sound signal segment, i.e., a data chain composed of time, frequency and intensity, wherein the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency;
[0059] Based on the remaining service life of each rolling bearing at the time when its vibration data and sound data are collected, the data chain at the corresponding time is annotated to construct a training set;
[0060] Using FFT to extract frequency features from vibration data can effectively identify periodic components in bearing vibration. These components are often associated with specific failure modes of the bearing, which helps to accurately diagnose the type of bearing failure.
[0061] Wavelet transform can provide time-frequency analysis of sound data, which is very useful for capturing non-stationary signals, such as transient changes in bearing sound signals. Time-frequency analysis can reveal the specific time and frequency characteristics of fault occurrence and improve the accuracy of fault detection. The process of performing wavelet transform on sound data specifically includes:
[0062] Select Molet wavelet for sound signal The continuous wavelet transform is based on the formula:
[0063] ;
[0064] in, are the wavelet coefficients, is time, that is, a variable in the time domain, used to represent the signal The values at different time points, is a scaled and translated version of the wavelet basis function, defined as:
[0065] ;
[0066] in, is a scale parameter, which is used to adjust the frequency of the wavelet basis function and controls the expansion and contraction of the wavelet basis function, that is, the frequency resolution. , the smaller For high frequency analysis, the larger Corresponding to low frequency analysis, The selection of appropriate parameters is determined through experiments and experience; is the translation parameter, which controls the translation of the wavelet basis function on the time axis, that is, the time resolution. is a real number, Continuously transforming within the time range of the signal, The selection of appropriate parameters is determined through experiments and experience. is the complex conjugate of the wavelet basis function.
[0067] Extracting the modulus maximum: In the modulus of wavelet transform Find the local maximum point in
[0068] Compute the energy distribution of wavelet coefficients at different scales and positions:
[0069] ;
[0070] in, Represents the energy of the wavelet coefficients at a specific scale a and position b, are the wavelet coefficients.
[0071] Step 3: Construct a rolling bearing remaining service life prediction model based on a deep learning algorithm, use the frequency characteristics and time-frequency characteristics of the rolling bearing as training sets, use the corresponding rolling bearing predicted remaining service life as labels, train the model, and obtain a trained rolling bearing remaining service life prediction model;
[0072] In this embodiment, the process of constructing a rolling bearing remaining service life prediction model based on a deep learning algorithm specifically includes: performing fast Fourier transform and wavelet transform on the collected historical vibration data and sound data of the rolling bearing, respectively, extracting frequency features and time-frequency features, and using them as training sets, constructing a rolling bearing remaining service life prediction model based on a deep learning algorithm, using the corresponding rolling bearing predicted remaining service life as a label, training the model, and obtaining a trained rolling bearing remaining service life prediction model;
[0073] Deep learning algorithms can process and identify patterns in large amounts of complex data, providing more accurate prediction results than traditional machine learning methods. Especially when processing data with time series characteristics, the model can automatically learn and extract useful features from the data, avoiding the complexity and limitations of manual feature engineering and improving the generalization ability of the model. Since there is a nonlinear relationship between bearing failure and its remaining service life, deep learning can effectively capture these nonlinear relationships and provide more accurate predictions.
[0074] Step 4: Collect the vibration data and sound data of the rolling bearing to be evaluated in real time, perform fast Fourier transform on the vibration data to extract the frequency characteristics, perform wavelet transform on the sound data to extract the time-frequency characteristics, and input the processed data into the model to output the predicted remaining service life of the rolling bearing to be evaluated;
[0075] In this embodiment, the remaining service life prediction model of the rolling bearing is used to obtain the predicted remaining service life of the rolling bearing to be evaluated. The process specifically includes: collecting the vibration data and sound data of the rolling bearing to be evaluated in real time, performing fast Fourier transform on the vibration data to extract the frequency characteristics, performing wavelet transform on the sound data to extract the time-frequency characteristics, and inputting the processed data into the model to output the predicted remaining service life of the rolling bearing to be evaluated. ; The model includes an input layer, a fully connected layer, an activation function layer and an output layer;
[0076] Among them, the input layer is used to receive the processed vibration data and sound data; the fully connected layer is used to integrate the features of the previous layer and make the final prediction. When predicting the remaining service life, the fully connected layer maps the features to the prediction results; the activation function layer is used to introduce nonlinearity; and the output layer generates the final prediction results.
[0077] Step 5: Collect the temperature data of the rolling bearing to be evaluated in real time, and use the collected temperature data to correct the predicted remaining service life of the rolling bearing to be evaluated to obtain a comprehensive prediction index of the rolling bearing to be evaluated.
[0078] In this embodiment, the predicted remaining service life of the rolling bearing to be evaluated is corrected using the collected temperature data to obtain a comprehensive prediction index of the rolling bearing to be evaluated. The process is based on the following formula:
[0079] ;
[0080] in, is the comprehensive prediction index of the rolling bearing to be evaluated, is the estimated remaining service life of the rolling bearing to be evaluated, is its preset scaling factor, and greater than zero, is the real-time temperature of the rolling bearing to be evaluated, The preset scale factor is is greater than zero, and ,The real-time temperature is the most critical factor that directly affects the remaining service life of the bearing and has a large weight in the model. is a constant correction index, and its value range is [0.5, 2.0]. This range can meet the needs of most situations, and can make necessary corrections without excessively distorting the original prediction results of the model. When the comprehensive prediction index increases will increase accordingly; when the real-time temperature of the rolling bearing to be evaluated When the temperature increases, high temperature will cause the mechanical properties of the bearing material to deteriorate, including the reduction of hardness and strength, and the lubricant of the bearing will become thinner or oxidized at high temperature, thus shortening the service life of the bearing. This means that the predicted remaining service life of the bearing is and the comprehensive forecast index Positively correlated, real-time temperature and the comprehensive forecast index There is a negative correlation.
[0081] Real-time temperature data reflects the current working conditions and status of the rolling bearing. Using this data to correct the predicted remaining service life of the rolling bearing to be evaluated can improve the accuracy and reliability of the prediction model.
[0082] The present invention further provides a rolling bearing remaining service life prediction system, the rolling bearing remaining service life prediction system is used to execute the above-mentioned rolling bearing remaining service life prediction system method, comprising:
[0083] The data acquisition module is used to collect the time-series vibration data and time-series sound data of a batch of rolling bearings in normal working condition, and conduct long-term tracking observation of the batch of bearings, record the timestamp when the bearings turn into fault state, and obtain the remaining service life of each rolling bearing at the moment of collecting its vibration data and sound data;
[0084] The feature extraction module applies fast Fourier transform to extract frequency features from time-series vibration data and uses wavelet transform to analyze time-series sound data to extract time-frequency features;
[0085] A model building module, which builds a rolling bearing remaining service life prediction model based on a deep learning algorithm, uses the frequency characteristics and time-frequency characteristics of the rolling bearing as a training set, uses the corresponding rolling bearing predicted remaining service life as a label, trains the model, and obtains a trained rolling bearing remaining service life prediction model;
[0086] The prediction module is used to collect the vibration data and sound data of the rolling bearing to be evaluated in real time, perform fast Fourier transform on the vibration data to extract the frequency characteristics, perform wavelet transform on the sound data to extract the time-frequency characteristics, and input the processed data into the model to output the predicted remaining service life of the rolling bearing to be evaluated;
[0087] The correction module is used to collect the temperature data of the rolling bearing to be evaluated in real time, and use the collected temperature data to correct the predicted remaining service life of the rolling bearing to be evaluated to obtain a comprehensive prediction index of the rolling bearing to be evaluated.
[0088] In the formula and The specific value of is generally determined by those skilled in the art according to actual conditions. The technicians in this field collect multiple groups of sample data, set corresponding preset proportional coefficients for each group of sample data, substitute the preset proportional coefficients and the collected sample data into the formula, observe the accuracy of the model output and the rationality of the results through repeated experiments and parameter adjustments, gradually adjust these factor coefficients, and compare the performance and effect of the model under different parameter settings to find the optimal coefficient combination, screen the calculated factor coefficients and take the average, and obtain and The value of .
[0089] In addition, the size of the preset factor coefficient is to quantify each parameter to obtain a specific value. In order to facilitate subsequent comparison, the size of the coefficient depends on the amount of sample data and the preset proportional coefficient initially set by technical personnel in this field for each set of sample data. It is not unique as long as it does not affect the proportional relationship between the parameter and the quantized value.
[0090] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.
[0091] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.
[0092] 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, and 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 solution of this embodiment.
[0093] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.
Claims
1. A method for predicting the remaining service life of a rolling bearing, characterized in that: The specific steps include: Step 1: Collect the time-series vibration data and time-series sound data of a batch of rolling bearings in normal working condition, conduct long-term tracking observation on the batch of bearings, record the timestamp when the bearings turn into fault state, and obtain the remaining service life of each rolling bearing at the moment of collecting its vibration data and sound data; Step 2: Apply fast Fourier transform to extract frequency features from time-series vibration data, and use wavelet transform to analyze time-series sound data to extract time-frequency features; Step 3: Construct a rolling bearing remaining service life prediction model based on a deep learning algorithm, use the frequency characteristics and time-frequency characteristics of the rolling bearing as training sets, use the corresponding rolling bearing predicted remaining service life as labels, train the model, and obtain a trained rolling bearing remaining service life prediction model; The process of using the frequency characteristics and time-frequency characteristics of the rolling bearing as a training set specifically includes: Apply fast Fourier transform to extract frequency features from vibration data: perform fast Fourier transform on each time-series vibration signal segment to obtain frequency features corresponding to each time-series vibration signal segment, wherein the frequency features include time, fundamental frequency and harmonics; construct a data chain with the frequency features corresponding to the time-series vibration signal segment, i.e., a data chain composed of time, fundamental frequency and harmonics, wherein the pointer between time and fundamental frequency is a timestamp index, which points to the fundamental frequency value at a specific moment, and the pointer between fundamental frequency and harmonics is a fundamental frequency index, which points to the harmonic value at a specific fundamental frequency; The method uses wavelet transform to analyze sound data and extract time-frequency features: performing wavelet transform on each time-series sound signal segment to obtain the time-frequency features corresponding to each time-series sound signal segment, wherein the frequency features include time, frequency and signal intensity; forming a data chain with the spectrum features corresponding to the time-series sound signal segment, i.e., a data chain composed of time, frequency and intensity, wherein the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency; Based on the remaining service life of each rolling bearing at the time when its vibration data and sound data are collected, the data chain at the corresponding time is annotated to construct a training set; The process of performing wavelet transform on the sound data specifically includes: Select Molet wavelet and perform continuous wavelet transform on the sound signal s(t). The formula is as follows: Among them, W s (a, b) are wavelet coefficients, t is time, that is, a variable in the time domain, used to represent the value of signal s at different time points. is a scaled and translated version of the wavelet basis function, defined as: Among them, a is the scale parameter, which is used to adjust the frequency of the wavelet basis function and control the expansion and contraction of the wavelet basis function, that is, the frequency resolution. a>0, a smaller a corresponds to high-frequency analysis, and a larger a corresponds to low-frequency analysis. The selection of a is determined by experiments and experience to determine the appropriate parameter range; b is the translation parameter, which controls the translation of the wavelet basis function on the time axis, that is, the time resolution. b is a real number, and b changes continuously within the time range of the signal. The selection of b is determined by experiments and experience to determine the appropriate parameter range. is the complex conjugate of the wavelet basis function; Extract the modulus maximum: In the wavelet transform modulus |W s Find the local maximum point in (a,b)|; Compute the energy distribution of wavelet coefficients at different scales and positions: E(a,b)=|W s (a,b)| 2 Where E(a,b) represents the energy of the wavelet coefficients at a specific scale a and position b, |W s (a,b)| is the wavelet coefficient; Step 4: Collect the vibration data and sound data of the rolling bearing to be evaluated in real time, perform fast Fourier transform on the vibration data to extract the frequency characteristics, perform wavelet transform on the sound data to extract the time-frequency characteristics, and input the processed data into the model to output the predicted remaining service life of the rolling bearing to be evaluated; Specifically, it includes: collecting vibration data and sound data of the rolling bearing to be evaluated in real time, performing fast Fourier transform on the vibration data to extract frequency features, performing wavelet transform on the sound data to extract time-frequency features, and inputting the processed data into the model, and outputting the predicted remaining service life M of the rolling bearing to be evaluated; the model includes an input layer, a fully connected layer, an activation function layer and an output layer; Among them, the input layer is used to receive the processed vibration data and sound data; the fully connected layer is used to integrate the features of the previous layers and make the final prediction. When predicting the remaining service life, the fully connected layer maps the features to the prediction results; the activation function layer is used to introduce nonlinearity; the output layer generates the final prediction results; Step 5: collecting temperature data of the rolling bearing to be evaluated in real time, and using the collected temperature data to correct the predicted remaining service life of the rolling bearing to be evaluated, to obtain a comprehensive prediction index of the rolling bearing to be evaluated; The process of using the collected temperature data to correct the predicted remaining service life of the rolling bearing to be evaluated to obtain the comprehensive prediction index β of the rolling bearing to be evaluated is based on the following formula: Among them, QS is the comprehensive prediction index of the rolling bearing to be evaluated, M is the predicted remaining service life of the rolling bearing to be evaluated, α is its preset proportional coefficient, and α is greater than zero, T is the real-time temperature of the rolling bearing to be evaluated, β is its preset proportional coefficient, β is greater than zero, and α<β, and C1 is the constant correction index.
2. A method for predicting the remaining service life of a rolling bearing according to claim 1, characterized in that: The process of collecting vibration data and sound data of the rolling bearing in normal working state specifically includes: Vibration sensors and sound sensors are installed on the bearing device to collect data in real time. The vibration data and sound data of the bearing are collected by the sensors. The data are all time series data and are divided into continuous time steps. Each time step represents a specific moment in the data, that is, each time step corresponds to a numerical value.
3. A rolling bearing remaining service life prediction system, characterized in that: The rolling bearing remaining service life prediction system is used to execute the rolling bearing remaining service life prediction method according to any one of claims 1 to 2, comprising: The data acquisition module is used to collect the time-series vibration data and time-series sound data of a batch of rolling bearings in normal working condition, and conduct long-term tracking observation of the batch of bearings, record the timestamp when the bearings turn into fault state, and obtain the remaining service life of each rolling bearing at the moment of collecting its vibration data and sound data; The feature extraction module applies fast Fourier transform to extract frequency features from vibration data and uses wavelet transform to analyze sound data and extract time-frequency features; A model building module, which builds a rolling bearing remaining service life prediction model based on a deep learning algorithm, uses the frequency characteristics and time-frequency characteristics of the rolling bearing as a training set, uses the corresponding rolling bearing predicted remaining service life as a label, trains the model, and obtains a trained rolling bearing remaining service life prediction model; The prediction module is used to collect the vibration data and sound data of the rolling bearing to be evaluated in real time, perform fast Fourier transform on the vibration data to extract the frequency characteristics, perform wavelet transform on the sound data to extract the time-frequency characteristics, and input the processed data into the model to output the predicted remaining service life of the rolling bearing to be evaluated; The correction module is used to collect the temperature data of the rolling bearing to be evaluated in real time, and use the collected temperature data to correct the predicted remaining service life of the rolling bearing to be evaluated to obtain a comprehensive prediction index of the rolling bearing to be evaluated.