CN115422687A - Service life prediction method of rolling bearing - Google Patents

Service life prediction method of rolling bearing Download PDF

Info

Publication number
CN115422687A
CN115422687A CN202211268695.2A CN202211268695A CN115422687A CN 115422687 A CN115422687 A CN 115422687A CN 202211268695 A CN202211268695 A CN 202211268695A CN 115422687 A CN115422687 A CN 115422687A
Authority
CN
China
Prior art keywords
rolling bearing
parameters
lstm
target
curve
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202211268695.2A
Other languages
Chinese (zh)
Inventor
梁天添
王润泽
郑祥
王英东
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dalian Jiaotong University
Original Assignee
Dalian Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Dalian Jiaotong University filed Critical Dalian Jiaotong University
Priority to CN202211268695.2A priority Critical patent/CN115422687A/en
Publication of CN115422687A publication Critical patent/CN115422687A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/04—Ageing analysis or optimisation against ageing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Geometry (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention belongs to the technical field of bearing reliability assessment and service life prediction, and discloses a service life prediction method of a rolling bearing, which comprises the following steps: acquiring vibration signal data of a target rolling bearing; extracting characteristic parameters of the vibration signal data in a time domain, a frequency domain and a time-frequency domain; performing dimensionality reduction processing on the characteristic parameters by adopting a kernel principal component analysis method to obtain kernel principal component variables; constructing a Weibull proportion fault rate model by taking the kernel principal component variable as a covariate, and obtaining a fault curve and a reliability curve about the target rolling bearing through the Weibull proportion fault rate model; optimizing parameters of the LSTM neural network by using an improved whale optimization algorithm to obtain a WOA-LSTM optimization prediction model with optimal parameters; establishing an evaluation index for screening target parameters from the characteristic parameters; and training a WOA-LSTM optimization prediction model by using the fault curve, the reliability curve and the target parameters, and predicting the residual service life of the target rolling bearing by using the trained WOA-LSTM optimization prediction model.

Description

Service life prediction method of rolling bearing
Technical Field
The invention belongs to the technical field of bearing reliability assessment and service life prediction, and particularly relates to a service life prediction method of a rolling bearing.
Background
With the advance of the increasing sophistication of modern mechanical equipment, emerging reliability concepts show a significant exponential growth compared to state maintenance (CBM) based approaches and show a strong demand for big data based health management (PHM) approaches as well as data analysis, statistical model based, data driven prediction approaches. Following the development of the era, the need for maintenance has driven progress from major corrective maintenance to routine maintenance practices. However, as system complexity increases, the need for cost-effective maintainability methods also increases; therefore, predictive maintenance is required. By predicting when a system will fail, necessary precautions can be taken to extend service life, make replacements or set up alternatives to avoid downtime, accidents, expensive maintenance, etc.
The rolling bearing is the most commonly used core basic part in mechanical equipment, and during the operation of the mechanical equipment, the conditions of long-time operation, insufficient lubrication degree, chemical corrosion and the like can occur, so that the bearing can generate faults such as abrasion, scratch, fatigue peeling, electric corrosion, inner and outer ring fracture and the like, once the fault condition occurs, on one hand, huge economic loss can be caused, and on the other hand, the personal safety of the public can be damaged. Therefore, under the research background, the reliability evaluation and the accurate prediction of the residual life of the rolling bearing can provide guidance for predictive maintenance of mechanical equipment, so that the maintenance is carried out in advance, unnecessary maintenance times are reduced, and the reliability evaluation and the residual life prediction of the rolling bearing have very important significance for equipment maintenance.
At present, a reliability evaluation and service life prediction method of a rolling bearing is developed mainly by extracting fault features based on vibration signals, and for the above conditions, research is mainly performed by extracting relevant indexes such as amplitude values, entropy values and the like of a time domain, a frequency domain or a time-frequency domain as parameters. Under actual conditions, vibration characteristic signals are difficult to extract effective signals due to the fact that the composite fault type is diverse and complex, tiny faults are easy to operate and noise is annihilated, fault signals are mutually coupled and interfered and the like, and therefore, a thorough research needs to be urgently carried out on how to extract an effective characteristic vector set so as to guarantee the accuracy of prediction.
The concept of fault Prediction and Health Management (PHM) has recently received attention from many scholars to determine the operational status of a device by evaluating the reliability of the device. In the current reliability research process of equipment, the position parameters of three-parameter Weibull distribution are often simplified, so that the parameters are reduced, the calculation is convenient, and the model is simplified into a two-parameter Weibull distribution model. A Weibull proportional fault rate model (WPHM) is constructed through two parameters of Weibull distribution, in most of previous researches, time domain statistical analysis is directly carried out on running state information, and then one or more time domain features are selected for modeling. However, the problem of insufficient evaluation capability of a single feature or a single-domain feature exists, the whole period process of the performance degradation of the bearing cannot be accurately represented, and the accuracy of reliability evaluation and service life prediction is seriously influenced. Although the performance degradation process of the full-period bearing can be comprehensively characterized by multi-domain characteristics such as time domain, frequency domain, time frequency domain and the like, redundancy exists when the characteristics are too many, and the more the WPHM covariates are, the more difficult the parameter estimation of the model is, so that it is not practical to directly substitute the multi-domain characteristics as the WPHM covariates for reliability evaluation. Therefore, the characteristics which do not contribute much to reflecting the fault characteristics or even are insensitive are effectively removed, the correlation among the characteristics is reduced, the information redundancy is reduced, and the characteristics which accurately represent the bearing performance degradation process are selected as covariates, so that the reliability evaluation of the bearing is important.
In recent years, a bearing residual life prediction method based on data driving becomes a hot spot of domestic and foreign research, and meanwhile, a Recurrent Neural Network (RNN) is widely applied to the residual life prediction of a rolling bearing due to the superiority of the RNN in time series processing. Because RNN presents long-term dependence, gradient explosion and extinction problems during training, long-short term memory networks (LSTM) have been proposed for this purpose. The LSTM effectively utilizes long-distance time sequence information by designing structures of a forgetting gate, an input gate and an output gate, and improves the prediction effect. At the same time, the predicted effect of the LSTM neural network depends largely on the settings of the parameters. The model may have different performance due to different parameters, and the prediction effect may be unstable due to the artificial setting of the parameters. Therefore, the size of the hyperparameter in the neural network is determined according to the structural planning and the prediction requirements of the neural network by introducing corresponding algorithms and improving different network structures, and the method has important significance for improving the prediction precision of the residual life of the bearing.
Disclosure of Invention
In order to realize accurate prediction of the service life of the bearing, the invention aims to provide a service life prediction method of a rolling bearing, and the prediction method specifically comprises the following steps:
acquiring vibration signal data of a target rolling bearing;
extracting characteristic parameters of the vibration signal data in a time domain, a frequency domain and a time-frequency domain;
performing dimensionality reduction processing on the characteristic parameters by adopting a kernel principal component analysis method to obtain kernel principal component variables;
constructing a Weibull proportion fault rate model by taking the kernel principal component variable as a covariate, and obtaining a fault curve and a reliability curve about the target rolling bearing through the Weibull proportion fault rate model;
optimizing parameters of the LSTM neural network by using an improved whale optimization algorithm to obtain a WOA-LSTM optimization prediction model with optimal parameters;
establishing an evaluation index and screening a target parameter from the characteristic parameters according to the evaluation index;
and training the WOA-LSTM optimized prediction model by using the fault curve, the reliability curve and the target parameters, and predicting the residual service life of the target rolling bearing by using the trained WOA-LSTM optimized prediction model.
Preferably, the characteristic parameters of the vibration signal data in the time domain at least comprise RMS value, peak value and kurtosis; the characteristic parameters of the vibration signal data in the frequency domain at least comprise root mean square values, variances and mean values; the characteristic parameters of the vibration signal data in the time-frequency domain at least comprise subdata obtained through wavelet decomposition, subdata obtained through empirical mode decomposition and subdata obtained through variational mode decomposition.
Preferably, the decomposition function of the wavelet decomposition is:
Figure BDA0003894195190000031
where x (t) denotes the vibration signal data to be analyzed, ψ (t) denotes a wavelet basis function, α denotes a scale function, and τ denotes a distance of shift.
Preferably, the decomposition function of the empirical mode decomposition is:
Figure BDA0003894195190000032
where x (t) is the original signal in the vibration signal data, c i (t) is the sub-data obtained by decomposition, r n (t) is a residual term.
Preferably, the feature parameters are subjected to dimensionality reduction by using a kernel principal component analysis method taking a Gaussian function as a kernel function, so as to obtain nonlinear kernel principal component variables.
Preferably, the improved whale optimization algorithm is as follows:
Figure BDA0003894195190000041
Figure BDA0003894195190000042
Figure BDA0003894195190000043
wherein t is the current iteration number; max _ iter is the maximum number of iterations;
Figure BDA0003894195190000044
omega is the adaptive parameter value, and lambda, mu, delta and gamma are the control parameters.
Preferably, the value range of the adaptive parameter value is 0-1.
Preferably, the state updating calculation process of the LSTM neural network base unit is as follows:
f t ＝σ(W f ·[h t-1 ,x t ]+b f )
i t ＝σ(W i ·[h t-1 ,x t ]+b i )；
o t ＝σ(W o ·[h t-1 ,x t ]+b o )
wherein i t 、f t 、o t Respectively representing the state calculation results of the input gate, the forgetting gate and the output gate; wi, W f 、W o And b i 、b f 、b o Respectively representing a weight matrix and a bias term of a corresponding gate, wherein sigma represents a sigmoid activation function;
preferably, the output result of the memory module of the LSTM neural network at time t is:
Figure BDA0003894195190000045
wherein: ct represents the cell state input at time t; tan h is a hyperbolic tangent activation function; w c 、b c Respectively representing the state weight matrix and the bias item of the input layer; * Indicating that the elements are multiplied by position.
Preferably, the evaluation index includes time correlation, monotonicity, and robustness with respect to the characteristic parameter.
Preferably, the training of the WOA-LSTM optimized prediction model using the fault curve, the reliability curve and the target parameter includes:
determining the degradation starting time of the target rolling bearing through the fault curve and the reliability curve;
normalizing the degradation starting time and the target parameters to obtain a training set;
training the WOA-LSTM optimized prediction model through the training set.
Compared with the prior art, the invention has the following beneficial effects:
(1) According to the prediction method, time domain, frequency domain and time-frequency domain analysis are carried out on the original vibration signal of the rolling bearing, signal characteristics are fully and comprehensively mined, the extracted characteristic parameters are used as data support for reliability evaluation prediction and service life prediction of the rolling bearing, the problem that evaluation capability of single characteristics or single-domain characteristics is insufficient is effectively solved, the performance degradation full-period process of the bearing can be accurately represented, and the accuracy of reliability evaluation and service life prediction is improved.
(2) According to the prediction method, the improved whale optimization algorithm and the long-short term memory LSTM neural network are fused to form a WOA-LSTM optimized prediction model for automatically optimizing model parameters according to input data, so that the problem that accuracy of a prediction result is unstable due to manual parameter selection is solved.
(3) According to the prediction method, the kernel principal component analysis method with the kernel function as the Gaussian function is adopted to perform dimensionality reduction processing on the characteristic parameters, and the obtained kernel principal component variables are used as covariates to construct a Weibull proportion fault rate model, so that the accuracy of the model in bearing reliability evaluation is effectively improved.
(4) The prediction method improves the traditional whale optimization algorithm, and replaces a fixed probability threshold value with a self-adaptive threshold value within 0-1, so that the algorithm can effectively select a proper optimization strategy at different periods, the global search capability and the local development capability of the algorithm are coordinated, and the convergence rate of the algorithm is effectively improved; meanwhile, the self-adaptive threshold value is used as a weight coefficient, and the reliability of the algorithm is increased along with the increase of the iteration times, so that the optimization accuracy of the algorithm is improved.
Drawings
FIG. 1 is a flowchart of a life prediction method of a rolling bearing according to the present invention;
fig. 2 is a failure graph (a) and a reliability graph (b) with respect to the target rolling bearing;
FIG. 3 is a graph comparing the predicted reliability and the actual reliability with respect to the target rolling bearing;
fig. 4 is a graph of the screening results of screening the characteristic parameter (a) with respect to the target rolling bearing by the evaluation index (b);
FIG. 5 is a comparison of the actual life of the target rolling bearing, the predicted life of the LSTM neural network, the predicted life of the WOA-LSTM network model, and the predicted life of the WOA-LSTM optimization prediction model.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Examples
Referring to fig. 1, the method for predicting the life of a rolling bearing provided by the present invention specifically includes the following steps:
s1, acquiring vibration signal data of a target rolling bearing;
in this embodiment, a plurality of sets of rolling bearing full-life vibration acceleration data sets (which are public data) are acquired from an Intelligent Maintenance Systems (IMS) center test bed of the cincinnati university in the united states.
The data set describes a number of experiments from testing to failure. Each data set consists of individual files that are snapshots of the vibration signal recorded as 1 second at specific time intervals every ten minutes. Each file consists of 20,480 dots, with a sample rate set at 20kHz. When the measured value of the vibration acceleration reaches the preset threshold value, it is considered that the remaining service life of the rolling bearing is over, in this example, the first bearing in the experiment is selected as an experimental object (target rolling bearing), and one or more data sets corresponding to the experimental object are the vibration signal data of the target rolling bearing required by this embodiment.
S2, extracting characteristic parameters of the vibration signal data in a time domain, a frequency domain and a time-frequency domain;
the characteristic parameters in the time domain include at least RMS value, peak value, kurtosis;
the characteristic parameters in the frequency domain at least comprise a root mean square value, a variance and a mean value;
the characteristic parameters in the time-frequency domain at least comprise subdata obtained through wavelet decomposition, subdata obtained through empirical mode decomposition and subdata obtained through variational mode decomposition.
Wherein:
with respect to wavelet decomposition, the decomposition function is:
Figure BDA0003894195190000071
where x (t) denotes the vibration signal data to be analyzed, ψ (t) denotes a wavelet basis function, α denotes a scale function, and τ denotes a distance of shift.
Specifically, the analyzed vibration signal data is transformed into a two-dimensional space including a scale domain and a time domain by the decomposition function. The wavelet decomposition function has two regulating factors, the time position of the window is moved by regulating the size of tau, and the shape and the frequency position of the window are regulated by regulating the scale function alpha, so that the decomposition effect is optimal.
Regarding empirical mode decomposition, the original signal of the vibration signal data may be decomposed into sub-data with different frequencies, and the decomposition function is specifically:
Figure BDA0003894195190000072
where x (t) is the original signal in the vibration signal data, c i (t) is the sub-data obtained by decomposition, r n (t) is a residual term.
S3, performing dimensionality reduction on the characteristic parameters by adopting a kernel principal component analysis method to obtain kernel principal component variables;
the kernel principal component analysis method is a method for performing nonlinear analysis, and projects the original data of the vibration signal data to a feature space, and principal component analysis is performed in a high-dimensional feature space, so that the original data has better separability. The gaussian kernel function has the characteristics of few parameters to be estimated, simple calculation process, wide application and the like, and the gaussian kernel function is used as the kernel function of the nonlinear mapping in the kernel principal component analysis method, so that the nonlinear kernel principal component variable is obtained.
And S4, constructing a Weibull proportion fault rate model by taking the core principal component variable as a covariate, and obtaining a fault curve (a) and a reliability curve (b) about the target rolling bearing shown in the figure 2 through the Weibull proportion fault rate model.
In addition, the true reliability of the target rolling bearing is compared with the predicted reliability obtained based on the weibull proportional failure rate model described above, and a comparison result as shown in fig. 3 is obtained. As can be seen from FIG. 3, the Weibull proportional fault rate model constructed by the invention has a high-accuracy bearing reliability evaluation effect.
S5, optimizing parameters of the LSTM neural network by using an improved whale optimization algorithm to obtain a WOA-LSTM optimization prediction model with optimal parameters;
regarding a traditional whale optimization algorithm, the steps are specifically expressed as follows:
(1) starting from a set of random solutions, the search agents update their locations according to a randomly selected search agent or the best solution currently obtained, which can be represented as:
Figure BDA0003894195190000081
wherein: t represents the current iteration;
Figure BDA0003894195190000082
is the position vector of the best solution and,
Figure BDA0003894195190000083
for the purpose of the current position vector,
Figure BDA0003894195190000084
and
Figure BDA0003894195190000085
in the form of a vector of coefficients,
Figure BDA0003894195190000086
Figure BDA0003894195190000087
linearly decreasing from 2 to 0 in an iterative process;
Figure BDA0003894195190000088
is [0,1 ]]The random vector of (1).
②
Figure BDA0003894195190000089
Also through the fluctuation range of
Figure BDA00038941951900000810
Decrease when
Figure BDA00038941951900000811
A random search agent is selected when the search is to be performed,
Figure BDA00038941951900000812
and then the optimal solution is selected to update the position of the search agent.
(3) Whale optimization algorithms are divided into two types of optimization:
Figure BDA00038941951900000813
depending on the value of p, the whale can switch between spiral or circular motion.
(4) The whale optimization algorithm is terminated by satisfying a termination criterion (typically reaching a maximum number of iterations).
Specifically, because the probability threshold of the traditional whale optimization algorithm is fixed, the actual optimization effect of the algorithm is limited, and based on the fact that the adaptive threshold with the value range of 0-1 is preferably used to replace the traditional fixed probability threshold in the invention, the whale optimization algorithm is further improved, and the improved whale optimization algorithm is as follows:
Figure BDA0003894195190000091
Figure BDA0003894195190000092
Figure BDA0003894195190000093
wherein t is the current iteration number; max _ iter is the maximum number of iterations;
Figure BDA0003894195190000094
omega is an adaptive parameter value (an adaptive threshold value with the value range of 0-1), and lambda, mu, delta and gamma are control parameters.
With respect to LSTM neural networks
The state updating calculation process of the basic unit is as follows:
f t ＝σ(W f ·[h t-1 ,x t ]+b f )
i t ＝σ(W i ·[h t-1 ,x t ]+b i )；
o t ＝σ(W o ·[h t-1 ,x t ]+b o )
wherein i t 、f t 、o t Respectively representing the state calculation results of the input gate, the forgetting gate and the output gate; wi, W f 、W o And b i 、b f 、b o Respectively representing a weight matrix and a bias term of a corresponding gate, wherein sigma represents a sigmoid activation function;
the output result of the memory module at the time t is as follows:
Figure BDA0003894195190000101
wherein: ct represents the cell state input at time t; tan h is a hyperbolic tangent activation function; w c 、b c Respectively representing the state weight matrix and the bias item of the input layer; * Indicating that the elements are multiplied by position.
In conclusion, the improved whale optimization algorithm is combined with the LSTM neural network to obtain a WOA-LSTM optimized prediction model with optimal parameters.
S6, establishing an evaluation index (b) shown in the figure 4 and screening a target parameter (a) shown in the figure 4 from the characteristic parameters through the evaluation index (b);
specifically, the evaluation index includes time correlation, monotonicity and robustness of the characteristic parameters;
wherein:
the correlation is used for measuring the degree of correlation between the characteristic parameter sequence and the time sequence;
monotonicity is used for describing the trend change of the characteristic parameter sequence which continuously increases or decreases;
the robustness is used for reflecting the anti-noise or fault signal interference capability of the characteristic parameter sequence;
in conclusion, the screened target parameters can be more accurately reflected as the characteristic quantity of the degradation process of the target rolling bearing in the whole life cycle.
S7, training the WOA-LSTM optimization prediction model by using the fault curve, the reliability curve and the target parameters, and predicting the remaining service life of the target rolling bearing through the trained WOA-LSTM optimization prediction model;
the specific operation of training the WOA-LSTM optimized prediction model by using the fault curve, the reliability curve and the target parameter is as follows:
determining the degradation starting time of the target rolling bearing according to the fault curve and the reliability curve;
normalizing the degradation starting time and the target parameters to obtain a training set;
training the WOA-LSTM optimized prediction model through the training set.
In summary, the invention respectively uses the LSTM neural network, the WOA-LSTM network model, and the WOA-LSTM optimized prediction model with the optimal parameters of the invention to predict the remaining service life of the target rolling bearing, and compares the obtained prediction result with the actual service life of the target rolling bearing to obtain the comparison graph shown in fig. 5. As can be seen from FIG. 5, the prediction result of the WOA-LSTM optimized prediction model with the optimal parameters on the basis of the invention on the residual service life of the target rolling bearing is closest to the real service life of the target rolling bearing, so that the prediction method provided by the invention has a more accurate prediction effect.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (10)

1. A method for predicting a lifetime of a rolling bearing, comprising:
acquiring vibration signal data of a target rolling bearing;
extracting characteristic parameters of the vibration signal data in a time domain, a frequency domain and a time-frequency domain;
performing dimensionality reduction processing on the characteristic parameters by adopting a kernel principal component analysis method to obtain kernel principal component variables;
constructing a Weibull proportional fault rate model by taking the kernel principal component variable as a covariate, and obtaining a fault curve and a reliability curve about a target rolling bearing through the Weibull proportional fault rate model;
optimizing parameters of the LSTM neural network by using an improved whale optimization algorithm to obtain a WOA-LSTM optimization prediction model with optimal parameters;
establishing an evaluation index for screening target parameters from the characteristic parameters;
and training the WOA-LSTM optimized prediction model by using the fault curve, the reliability curve and the target parameters, and predicting the residual service life of the target rolling bearing by using the trained WOA-LSTM optimized prediction model.
2. A life prediction method of a rolling bearing according to claim 1, characterized in that:
the characteristic parameters of the vibration signal data in the time domain at least comprise RMS (root mean square) values, peak values and kurtosis;
the characteristic parameters of the vibration signal data in the frequency domain at least comprise a root mean square value, a variance and a mean value;
the characteristic parameters of the vibration signal data in the time-frequency domain at least comprise subdata obtained through wavelet decomposition, subdata obtained through empirical mode decomposition and subdata obtained through variational mode decomposition.
3. Method for predicting the lifetime of a rolling bearing according to claim 2, characterized in that said wavelet decomposition has a decomposition function of:
Figure FDA0003894195180000011
where x (t) represents the vibration signal data to be analyzed, ψ (t) represents a wavelet basis function, α represents a scale function, and τ represents the distance of the shift.
4. Method for predicting the lifetime of a rolling bearing according to claim 3, characterized in that said empirical mode decomposition has a decomposition function of:
Figure FDA0003894195180000021
wherein x (t) is the original signal in the vibration signal data, c i (t) is the decomposed subdata, r n (t) is a residual term.
5. The life prediction method of a rolling bearing according to claim 4, characterized in that: and performing dimensionality reduction on the characteristic parameters by adopting a kernel principal component analysis method taking a Gaussian function as a kernel function to obtain nonlinear kernel principal component variables.
6. Method for predicting the life of a rolling bearing according to claim 5, characterized in that said modified whale optimization algorithm is:
Figure FDA0003894195180000022
Figure FDA0003894195180000023
Figure FDA0003894195180000024
wherein t is the current iteration number; max _ iter is the maximum number of iterations;
Figure FDA0003894195180000026
omega is the adaptive parameter value, and lambda, mu, delta and gamma are the control parameters.
7. The life prediction method of a rolling bearing according to claim 6, characterized in that: the value range of the self-adaptive parameter value is 0-1.
8. The life prediction method of a rolling bearing according to claim 7, characterized in that:
the state updating calculation process of the LSTM neural network basic unit comprises the following steps:
Figure FDA0003894195180000025
wherein i t 、f t 、o t Respectively representing the state calculation results of the input gate, the forgetting gate and the output gate; w is a group of i 、W f 、W o And b i 、b f 、b o Respectively representing a weight matrix and a bias term of a corresponding gate, wherein sigma represents a sigmoid activation function;
the output result of the memory module of the LSTM neural network at the moment t is as follows:
Figure FDA0003894195180000031
wherein: c t A cell state input representing time t; tan h is a hyperbolic tangent activation function; w is a group of c 、b c Respectively representing the state weight matrix and the bias item of the input layer; * Indicating that the elements are multiplied by position.
9. The life prediction method of a rolling bearing according to claim 8, characterized in that: the evaluation index includes time correlation, monotonicity and robustness of the characteristic parameter.
10. The method for predicting the life of a rolling bearing according to claim 9, wherein training the WOA-LSTM optimized prediction model using the fault curve, the reliability curve, and the target parameters comprises:
determining the degradation starting time of the target rolling bearing according to the fault curve and the reliability curve;
normalizing the degradation starting time and the target parameters to obtain a training set;
training the WOA-LSTM optimized prediction model through the training set.
CN202211268695.2A 2022-10-17 2022-10-17 Service life prediction method of rolling bearing Pending CN115422687A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211268695.2A CN115422687A (en) 2022-10-17 2022-10-17 Service life prediction method of rolling bearing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211268695.2A CN115422687A (en) 2022-10-17 2022-10-17 Service life prediction method of rolling bearing

Publications (1)

Publication Number Publication Date
CN115422687A true CN115422687A (en) 2022-12-02

Family

ID=84207720

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211268695.2A Pending CN115422687A (en) 2022-10-17 2022-10-17 Service life prediction method of rolling bearing

Country Status (1)

Country Link
CN (1) CN115422687A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116502049A (en) * 2023-06-25 2023-07-28 山东科技大学 Rolling bearing residual service life prediction method, system, equipment and storage medium
CN117647392A (en) * 2024-01-30 2024-03-05 成都三一能源环保技术有限公司 Downhole drilling instrument scrapping monitoring and early warning system based on data analysis

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112434390A (en) * 2020-12-01 2021-03-02 江苏科技大学 PCA-LSTM bearing residual life prediction method based on multi-layer grid search

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112434390A (en) * 2020-12-01 2021-03-02 江苏科技大学 PCA-LSTM bearing residual life prediction method based on multi-layer grid search

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
刘亮等: "一种求解函数优化问题的改进鲸鱼优化算法", 《计算机应用研究》, vol. 37, no. 4, pages 2 *
杨咸启: "《接触力学理论与滚动轴承设计分析》", pages: 326 *
王奉涛等: "基于KPCA和WPHM的滚动轴承可靠性评估与寿命预测", 《振动、测试与诊断》, vol. 37, no. 3, pages 1 - 2 *
王奉涛等: "基于长短期记忆网络的滚动轴承寿命预测方法", 《振动、测试与诊断》, vol. 40, no. 2, pages 1 - 3 *
郑直等: "基于改进鲸鱼算法优化LSTM的滚动轴承故障诊断", 《振动与冲击》, vol. 40, no. 7, pages 5 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116502049A (en) * 2023-06-25 2023-07-28 山东科技大学 Rolling bearing residual service life prediction method, system, equipment and storage medium
CN116502049B (en) * 2023-06-25 2023-09-08 山东科技大学 Rolling bearing residual service life prediction method, system, equipment and storage medium
CN117647392A (en) * 2024-01-30 2024-03-05 成都三一能源环保技术有限公司 Downhole drilling instrument scrapping monitoring and early warning system based on data analysis
CN117647392B (en) * 2024-01-30 2024-04-09 成都三一能源环保技术有限公司 Downhole drilling instrument scrapping monitoring and early warning system based on data analysis

Similar Documents

Publication Publication Date Title
CN110263866B (en) Power consumer load interval prediction method based on deep learning
CN116757534B (en) Intelligent refrigerator reliability analysis method based on neural training network
CN115422687A (en) Service life prediction method of rolling bearing
CN112364560B (en) Intelligent prediction method for working hours of mine rock drilling equipment
CN113723007A (en) Mechanical equipment residual life prediction method based on DRSN and sparrow search optimization BilSTM
CN116502049B (en) Rolling bearing residual service life prediction method, system, equipment and storage medium
CN116702612A (en) CVT health state prediction method and system based on multidimensional index fusion
CN114896861A (en) Rolling bearing residual life prediction method based on square root volume Kalman filtering
CN115982141A (en) Characteristic optimization method for time series data prediction
CN116379360A (en) Knowledge migration-based hydrogen-doped natural gas pipeline damage prediction method and system
CN117313936B (en) Clean flue gas SO in flue gas desulfurization process of coal-fired power plant 2 Concentration prediction method
CN117367800A (en) Multi-feature index bearing residual life prediction method based on random correlation
CN113887571A (en) Electronic equipment fault prediction method for improving SVR algorithm
CN113255963A (en) Road surface use performance prediction method based on road element splitting and deep learning model LSTM
CN117390836A (en) BEMD-based GRU interval system inertia evaluation and prediction method
CN117155806A (en) Communication base station flow prediction method and device
CN116822345A (en) Landslide displacement prediction method based on SSA-VMD-GRU
CN116579233A (en) Method for predicting residual life of mechanical equipment
CN116432856A (en) Pipeline dynamic early warning method and device based on CNN-GLSTM model
CN112257327A (en) Submarine pipeline residual life prediction method based on wavelet transform denoising and stacking self-coding feature selection
CN115389743A (en) Method, medium and system for predicting content interval of dissolved gas in transformer oil
Luo et al. A novel method for remaining useful life prediction of roller bearings involving the discrepancy and similarity of degradation trajectories
CN115292820A (en) Method for predicting residual service life of urban rail train bearing
CN113468720A (en) Service life prediction method for digital-analog linked random degradation equipment
Wang et al. Health indicator forecasting for improving remaining useful life estimation

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination