CN114564877A - Rolling bearing service life prediction method, system, equipment and readable storage medium - Google Patents
Rolling bearing service life prediction method, system, equipment and readable storage medium Download PDFInfo
- Publication number
- CN114564877A CN114564877A CN202111648638.2A CN202111648638A CN114564877A CN 114564877 A CN114564877 A CN 114564877A CN 202111648638 A CN202111648638 A CN 202111648638A CN 114564877 A CN114564877 A CN 114564877A
- Authority
- CN
- China
- Prior art keywords
- bearing
- life
- data
- fusion
- representing
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000005096 rolling process Methods 0.000 title claims abstract description 38
- 238000003860 storage Methods 0.000 title claims abstract description 16
- 238000006731 degradation reaction Methods 0.000 claims abstract description 52
- 230000015556 catabolic process Effects 0.000 claims abstract description 50
- 230000004927 fusion Effects 0.000 claims abstract description 38
- 230000015654 memory Effects 0.000 claims abstract description 22
- 230000005540 biological transmission Effects 0.000 claims abstract description 20
- 238000000605 extraction Methods 0.000 claims description 14
- 230000006403 short-term memory Effects 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 13
- 230000007787 long-term memory Effects 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 239000013598 vector Substances 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims 2
- 150000001875 compounds Chemical class 0.000 claims 1
- 238000010276 construction Methods 0.000 claims 1
- 238000007500 overflow downdraw method Methods 0.000 claims 1
- 230000007704 transition Effects 0.000 claims 1
- 239000000284 extract Substances 0.000 abstract description 6
- 238000013461 design Methods 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000000306 component Substances 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- 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
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
-
- 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
-
- 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
- G06F2119/10—Noise analysis or noise optimisation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Geometry (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a method, a system, equipment and a readable storage medium for predicting the service life of a rolling bearing, wherein convolution and a long-short term network are combined, and the constructed convolution long-short term memory network can adaptively extract typical characteristics of a time scale and a space scale in an original signal, so that the participation of expert knowledge is greatly reduced; the convolution long-short term memory network with the multi-branch structure extracts multi-dimensional degradation characteristics from multi-channel vibration signals, improves the comprehensiveness of the extracted characteristics, designs an information transmission layer among all the branch networks, and realizes the selective transmission and fusion of the characteristics among different channels through a gate structure, so that the hidden degradation information in each channel characteristic can be effectively utilized, and the accuracy and the reliability of the residual life prediction of the bearing are improved.
Description
Technical Field
The invention belongs to the technical field of residual life prediction of rolling bearings, and relates to a method, a system, equipment and a readable storage medium for predicting the life of a rolling bearing, in particular to a method and a system for predicting the life of a rolling bearing based on a convolution long-term and short-term memory fusion network.
Background
The rolling bearing is used as a core component of rotary mechanical equipment and widely applied to large-scale equipment such as numerical control machines, aircraft engines, wind driven generators and the like. According to incomplete statistics, about 30% of failures in rotating machines are due to rolling bearings. Once a rolling bearing fails, not only the operation of the whole mechanical equipment is affected and economic losses are caused, but also great life and property safety is possibly caused. Therefore, in actual production, the running state of the bearing of the equipment and the residual time before complete failure are predicted by the residual life prediction method, and the method has great significance for improving the running management efficiency and the maintenance efficiency of the mechanical equipment.
The existing method for predicting the residual life of the rolling bearing mainly comprises two methods based on a physical model and data driving. With the application of technologies such as sensing and monitoring in the industrial field, it becomes easier to acquire massive monitoring data, and the data driving method becomes a hot spot of current research by virtue of its strong modeling capability on the rolling bearing degradation process under complex working conditions. The conventional data-driven life prediction method generally comprises the steps of degradation feature extraction, feature dimension reduction and regression prediction, a series of features reflecting the degradation process of a single-channel vibration signal are extracted from the single-channel vibration signal, typical degradation features are reserved in feature selection and feature dimension reduction of the features, and then a regression model is used for predicting the residual life of the bearing from the features. Most of the existing methods only use single-channel vibration signals, and feature selection and dimension reduction of the existing methods often require a large amount of expert knowledge, so that typical features reflecting bearing degradation information are difficult to obtain, and meanwhile, single-channel data are difficult to ensure that comprehensive degradation information can be obtained, and the influence on subsequent prediction is large.
Disclosure of Invention
The invention aims to provide a method, a system, equipment and a readable storage medium for predicting the service life of a rolling bearing, which effectively combine a multi-source information fusion theory with a CLSTM model, solve the problems of complicated degradation characteristic selection process and incomplete degradation information characterization capability of a single-channel signal in the residual service life prediction of the rolling bearing, realize the effective fusion of multi-channel degradation characteristics, and improve the accuracy and the universality of a residual service life prediction model of the rolling bearing under complicated working conditions.
A life prediction method for a rolling bearing includes the steps of:
1) acquiring multi-directional original vibration signal time sequence data of the bearing by using a vibration sensor, dividing the data into a health stage and a degradation stage by using a 3 sigma criterion, and obtaining frequency domain data by using fast Fourier transform on the data of the degradation stage so as to construct a data set;
2) constructing a multi-branch feature extraction layer comprising a plurality of convolution long and short term memory networks (CLSTMs), and respectively inputting the frequency domain data sets in all directions obtained in the step 1) into all branch networks for extracting degradation feature;
3) constructing an information transmission layer, transmitting the degradation characteristics of the vibration signals in all directions extracted by each branch CLSTM in the step 2) in the information transmission layer, and performing selective screening and adaptive weighting fusion to realize preliminary fusion of the degradation characteristics;
4) constructing a multi-sensing information fusion layer, and finally fusing the hidden features finally output by each CLSTM;
5) constructing a regression prediction layer, and performing regression prediction on the residual life of the bearing by using the fusion characteristics obtained in the step 4);
6) and optimizing a loss function of the network by using an adaptive distance estimation (Adam) algorithm to obtain a trained rolling bearing life prediction model.
7) And predicting the residual life of the bearing outside the training set by using the trained rolling bearing life prediction model, and outputting the percentage of the residual life of the bearing in the total life of the degradation stage.
A rolling bearing life prediction system comprising:
the preprocessing module is used for preprocessing the multi-directional original vibration signal time sequence data of the bearing and determining a degradation starting point of the bearing; constructing a multi-branch feature extraction layer comprising a plurality of convolution long and short term memory networks, and respectively inputting the multi-directional original vibration signal time sequence data of the acquired multi-directional original vibration signal time sequence data into the networks for feature extraction to obtain vibration signal features;
the training module is used for transmitting the extracted vibration signal characteristics in each branch CLSTM to realize information preliminary fusion, finally fusing the hidden characteristics finally output by each CLSTM, performing regression prediction on the residual life of the bearing according to the obtained fusion characteristics, and optimizing a loss function of a network by using an adaptive distance estimation algorithm to obtain a trained rolling bearing life prediction model;
and the prediction module is used for predicting the residual life of the bearing to be tested and outputting the percentage of the residual life of the bearing to be tested to the total life of the degradation stage.
A terminal device comprising a memory, a processor and a computer program stored in said memory and executable on said processor, said processor implementing the steps of the rolling bearing life prediction method described above when executing said computer program.
A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned rolling bearing life prediction method.
Compared with the prior art, the invention has the following beneficial technical effects:
according to the rolling bearing service life prediction method, convolution and a long-short term network are combined, the constructed convolution long-short term memory network can adaptively extract typical characteristics of time scales and space scales in original signals, and the participation of expert knowledge is greatly reduced; the convolution long-short term memory network with the multi-branch structure extracts multi-dimensional degradation characteristics from multi-channel vibration signals, improves the comprehensiveness of the extracted characteristics, designs an information transmission layer among all the branch networks, and realizes the selective transmission and fusion of the characteristics among different channels through a gate structure, so that the hidden degradation information in each channel characteristic can be effectively utilized, and the accuracy and the reliability of the residual life prediction of the bearing are improved.
According to the invention, a multi-source information fusion theory and a CLSTM model are effectively combined, the problems that the selection process of degradation characteristics is complex and the capability of single-channel signals for representing degradation information is incomplete in the residual service life prediction of the rolling bearing are solved, the effective fusion of the multi-channel degradation characteristics is realized, and the accuracy and the universality of the residual service life prediction model of the rolling bearing under the complex working condition are improved.
The invention can accurately predict the residual service life of the bearing under complex working conditions, thereby maintaining and repairing the parts in time, avoiding deep damage of mechanical equipment and prolonging the service life of the bearing.
Drawings
FIG. 1 is a flowchart of a method for predicting the remaining life of a rolling bearing according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of an information transmission network according to an embodiment of the present invention.
FIG. 3 is a diagram of a convolutional long short term memory convergence network according to an embodiment of the present invention.
FIG. 4 is a comparison result of the residual life prediction of the bearing in the embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, 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.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The effectiveness of the diagnostic method of the present invention is verified by taking as an example the bearing degradation data set of PHM2012 challenge race provided by the french FEMTO research institute.
The data set used in this example sets the bearing to three different operating conditions, as shown in table 1. The bearing in the data set is respectively provided with a vibration acceleration sensor in the horizontal direction and the vertical direction to synchronously acquire vibration signals in two directions, the sampling frequency is 25.6kHz, the sampling is carried out once every 10 seconds, and the sampling time is 0.1 second every time.
TABLE 1 FEMTO bearing data set behavior
The invention relates to a method for predicting the residual life of a rolling bearing based on a convolution long-term and short-term memory network, which specifically comprises the following steps:
and step S1, acquiring multi-directional original vibration signal time series data of the bearing acquired by the bearing vibration signal sensor, preprocessing the data, determining a degradation starting point, performing Fast Fourier Transform (FFT) on the data in the degradation stage to obtain frequency domain signal data of the data in the degradation stage, and dividing the frequency domain signal data into a training set and a test set.
S11, collecting original data collected by a multi-channel bearing vibration signal sensor, preprocessing the original data, and removing data abnormal points and filling data missing points;
acquiring multi-channel rolling bearing original degradation time series vibration data u1，u2，…，uiWherein 1, 2, …, i represents different channels;
s12, dividing the operation process of the rolling bearing into a normal operation stage and a degradation stage according to the kurtosis characteristic and the 3 sigma criterion: dividing the degradation starting point of the data by using a 3 sigma criterion, and removing the data in the health stage and taking the data in the degradation stage as training data or test data according to a calculation formula.
|kurt+i-μk|＞3σk(i＝0,1,2)
In the formula, kurt+iRepresents the kurtosis value at the time t + i; mu.sk、σkAnd the mean value and the standard deviation of the settlement kurtosis characteristic of the normal operation of the bearing are shown.
When the kurtosis value at a certain moment is larger than the 3 sigma interval of the kurtosis value in the normal operation stage, the bearing is considered to be in the degradation stage, and data of the degradation stage are obtained according to the kurtosis value
S13, processing the data in the degradation stage by using Fast Fourier Transform (FFT), where 2560 data points are given as the input data, obtaining a frequency domain signal with a dimension of 1 × 1280 after FFT, and taking a frequency domain signal x obtained after normalization as an input of the life prediction model:
specifically, the method comprises the following steps: and transforming the data in the degradation stage into frequency domain data by using fast Fourier transform, wherein the data of the bearing 1-1, the bearing 2-2 and the bearing 3-3 are used as test data, and the data of the rest bearings are all used as training data.
and S21, inputting the frequency domain data obtained in the step S13 into a convolution network of a branch network for dimension transformation and feature extraction, wherein the convolution network comprises three convolution layers, convolution kernels used by each convolution layer are the same in size and are all 5 multiplied by 1, and the features of the bearing after the dimension transformation are obtained preliminarily.
S22, after the features of the bearing after dimension transformation are obtained, a long-term and short-term memory network is constructed, the features obtained in the S21 are input to carry out extraction and screening of degradation features, and hidden features h containing bearing degradation information are further obtainedtThe calculation method is as follows:
It＝σ(Wix*xt+Wih*ht-1+bi)
Ft＝σ(Wfx*xt+Wfh*ht-1+bf)
Ot＝σ(Wox*xt+Woh*ht-1+bo)
in the formula It，Ft，OtRespectively representing the values of an input gate, a forgetting gate and an output gate at the current moment; c. Ct，ct-1Respectively representing the cell states at the current moment and the previous moment; h ist，ht-1Outputs representing the current time and the previous time, respectively; x is the number oftAn input representing a current time; wix、Wih，Wfx、Wfh，Wox、Woh，Wcx、WchRespectively representing input gatesA forgetting gate, an output gate and a convolution kernel of a cell state; bi，bf，bo，bcBias vectors representing the input gate, the forgetting gate, the output gate, and the cell state, respectively; σ () represents a sigmoid activation function; tanh () represents a hyperbolic tangent activation function; denotes convolution operation;
representing a Hadamard product;
step S3, as shown in fig. 2, an information transport layer ITL is constructed, the hidden output of each layer of CLSTM in the branch network and the hidden output of each layer of CLSTM in the other branch network are transferred between the branch networks through the information transport layer, and are subjected to preliminary fusion, the fusion process is the fusion of adaptive weight distribution, and the fused features are used as the input of the next layer of CLSTM to perform the next degradation feature extraction, and the fusion process is shown as follows:
in the formula, Tt 1，Tt 2Respectively representing the values of transmission gates of the 1 st and 2 nd branch networks at the time t;
respectively representing hidden state information output by the 1 st and 2 nd branch networks at the time t-1;
W1，
W2respectively representing the transmission gate conversion weights of the 1 st branch network and the 2 nd branch network;
respectively representing the output of the 1 st and 2 nd branch networks after the characteristic preliminary fusion at the time t;
the values of the 1 st and 2 nd branch network output gates and memory cells at time t are shown, respectively.
Step S4, after the hidden feature output of the last layer of CLSTM is obtained in step S3, a final fusion network is constructed, the hidden output features of the two branch networks are fused by convolution operation, and the calculation method is shown as the following formula:
in the formula: h isfuseThe state information after final fusion is obtained; conv () represents a convolution operation.
It should be noted that, since the verification data used only includes data of 2 channels, only the fusion of two channels is shown in fig. 2, but when there is more data of the number of channels used for lifetime prediction, the structure shown in fig. 2 can be extended accordingly. Therefore, assuming that N sensors acquire N channels of data, the proposed fusion model is expressed in the following general form:
in the formula, Tt i，
The value and hidden characteristics of the transmission gate for the ith channel at time t.
Further, at the end of the branch networks, the hidden features from each branch network are fused at the fusion layer, and the fusion formula can be correspondingly expanded as follows:
step S5, constructing a regression prediction layer, taking the fusion characteristics obtained in step S4 as input, and obtaining output as the residual life percentage of the bearing, wherein the expression is as follows:
in the formula, Wr，brWeight matrices and bias vectors of the regression prediction layer are respectively represented.
And defining the mean square error of the bearing life prediction result, and taking the error as a loss function of network training. Wherein the mean square error is defined as:
in the formula, L represents an error value; k represents the number of training samples; y isiRepresenting a true remaining life value;
representing the predicted remaining life value.
In one embodiment of the present invention, a terminal device is provided that includes a processor and a memory, the memory storing a computer program comprising program instructions, the processor executing the program instructions stored by the computer storage medium. The processor adopts a Central Processing Unit (CPU), or other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), ready-made programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., which are a computing core and a control core of the terminal, and are adapted to implement one or more instructions, and in particular, to load and execute one or more instructions to implement a corresponding method flow or a corresponding function; the processor provided by the embodiment of the invention can be used for the operation of the life prediction method of the rolling bearing.
A rolling bearing life prediction system comprising:
the preprocessing module is used for preprocessing the multi-directional original vibration signal time sequence data of the bearing and determining a degradation starting point of the bearing; constructing a multi-branch feature extraction layer comprising a plurality of convolution long and short term memory networks, and respectively inputting the multi-directional original vibration signal time sequence data of the acquired multi-directional original vibration signal time sequence data into the networks for feature extraction to obtain vibration signal features;
the training module is used for transmitting the extracted vibration signal characteristics in each branch CLSTM to realize information preliminary fusion, finally fusing the hidden characteristics finally output by each CLSTM, performing regression prediction on the residual life of the bearing according to the obtained fusion characteristics, and optimizing a loss function of a network by using an adaptive distance estimation algorithm to obtain a trained rolling bearing life prediction model;
and the prediction module is used for predicting the residual life of the bearing to be tested and outputting the percentage of the residual life of the bearing to be tested to the total life of the degradation stage.
In still another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a Memory device in the terminal device and is used for storing programs and data. The computer-readable storage medium includes a built-in storage medium in the terminal device, provides a storage space, stores an operating system of the terminal, and may also include an extended storage medium supported by the terminal device. Also, one or more instructions, which may be one or more computer programs (including program code), are stored in the memory space and are adapted to be loaded and executed by the processor. It should be noted that the computer-readable storage medium may be a high-speed RAM memory, or may be a Non-volatile memory (Non-volatile memory), such as at least one disk memory. One or more instructions stored in the computer-readable storage medium may be loaded and executed by a processor to implement the corresponding steps of the method for predicting the life of a rolling bearing in the above embodiments.
To fully illustrate the effectiveness of the method proposed by the present invention, the present embodiment compares with a long short term memory network (LSTM) and a convolution long short term memory network (CLSTM), and quantitatively describes the prediction effect of each model by using the Root Mean Square Error (RMSE) and the Mean Absolute Percentage Error (MAPE) as evaluation criteria. The predicted results of the test set are shown in table 2 and fig. 4, respectively.
TABLE 2 comparison of predicted results for three methods
As can be seen from table 2, the method proposed by the present invention achieves the best results on all three tested bearings, with the prediction error being significantly lower than the other two methods. The method provided by the embodiment mainly utilizes vibration signals of the bearing in two directions, and constructs an information transmission layer to screen and weight and fuse the characteristics of the vibration signals in the two directions, so that the influence of noise and interference is inhibited, and the potential degradation characteristics of the bearing are fully extracted. The effectiveness of the proposed method is further verified by the prediction regression curve of fig. 4, and it can be seen from the regression curve that the proposed method can approach the actual life curve with less oscillation.
In conclusion, the convolution and the long-short term network are combined, the constructed convolution long-short term memory network can adaptively extract the typical characteristics of the time scale and the space scale in the original signal, and the participation of expert knowledge is greatly reduced; the convolution long-term and short-term memory network with the multi-branch structure extracts multi-dimensional degradation characteristics from multi-channel vibration signals, the comprehensiveness of the extracted characteristics is improved, an information transmission layer is designed among the branch networks, and selective transmission and fusion of the characteristics among different channels are achieved through a gate structure, so that the hidden degradation information in each channel characteristic can be effectively utilized, and the accuracy and the reliability of residual life prediction of a bearing are improved.
According to the method, the potential degradation characteristics are extracted from the original monitoring signals of the multi-channel bearing through the convolution long-term and short-term memory fusion network, screening and weighting fusion are carried out, the prediction precision of the residual life of the bearing and the generalization capability and robustness of a model can be improved, and a new method idea is provided for the field of residual life prediction of the bearing.
Claims (10)
1. A method for predicting the life of a rolling bearing, comprising the steps of:
s1, collecting multi-directional original vibration signal time sequence data of the bearing, preprocessing the data, and determining a bearing degradation starting point;
s2, constructing a multi-branch feature extraction layer comprising a plurality of convolution long and short term memory networks, and respectively inputting multi-directional original vibration signal time sequence data of the multi-directional original vibration signal time sequence data acquired in the S1 into the networks for feature extraction to obtain vibration signal features;
s3, constructing an information transmission layer: transmitting the vibration signal characteristics extracted in the S2 in each branch CLSTM to realize the preliminary information fusion;
s4, constructing a multi-sensing information fusion layer: finally fusing the hidden features finally output by each CLSTM;
s5, constructing a regression prediction layer: performing regression prediction on the residual life of the bearing by using the fusion characteristics obtained in the S4;
s6, optimizing a loss function of the network by using an adaptive distance estimation algorithm to obtain a trained rolling bearing life prediction model;
and S7, predicting the residual life of the bearing to be tested by using the trained rolling bearing life prediction model, and outputting the percentage of the residual life of the bearing to be tested to the total life of the degradation stage.
2. The method as claimed in claim 1, wherein the data set is constructed by collecting time series data of original vibration signals of the bearing in multiple directions by using a vibration sensor, dividing the data into a healthy stage and a degradation stage by using a 3 σ criterion, and obtaining frequency domain data from the data of the degradation stage by using fast fourier transform.
3. The method as claimed in claim 1, wherein the step of constructing the multi-branch feature extraction network comprising a plurality of convolution long and short term memory networks comprises the following steps:
s21, constructing a convolution-based long-short term memory cell structure, inputting data at the time t into a CLSTM, and obtaining hidden feature output h at the current timetThe specific calculation method is as follows:
It＝σ(Wix*xt+Wih*ht-1+bi)
Ft＝σ(Wfx*xt+Wfh*ht-1+bf)
Ot＝σ(Wox*xt+Woh*ht-1+bo)
in the formula It，Ft，OtRespectively representing the values of an input gate, a forgetting gate and an output gate at the current moment; c. Ct，ct-1Respectively representing the cell states at the current moment and the previous moment; h ist，ht-1Outputs representing a current time and a previous time, respectively; x is the number oftAn input representing a current time; wix、Wih，Wfx、Wfh，Wox、Woh，Wcx、WchConvolution kernels respectively representing an input gate, a forgetting gate, an output gate and a cell state; bi，bf，bo，bcBias vectors representing the input gate, the forgetting gate, the output gate, and the cell state, respectively; σ () represents a sigmoid activation function; tanh () represents a hyperbolic tangent activation function; denotes a convolution operation;
representing a Hadamard product;
s22, based on the convolution long-short term memory cell structure constructed in S21, N CLSTM networks are constructed according to the number N of the channels of the collected signals to extract the characteristics of the signals of each channel, and the final hidden characteristics h of each channel are obtained1，h2。
4. The method as claimed in claim 1, wherein the information transmission layer in S3 interacts data in each branch network through weight learning, and shares part of feature information to implement preliminary fusion of features, and the specific calculation steps are as follows:
s31, calculating the value of the transmission gate according to the information of each CLSTM, as shown in the formula:
in the formula, Tt iA value representing the transmission gate of the ith branch network at time t;
indicating hidden state information, W, output by the ith branch network at time t-1iTransmission gate transition weights representing ith branch network
S32, inputting the values of the transmission gates obtained in S31 into the information transmission layer to perform preliminary fusion with the values of the transmission gates of other branch networks, where the fusion method is as follows:
in the formula (I), the compound is shown in the specification,
the output after the characteristic of the ith branch network at the moment t and the characteristics of other branch networks are preliminarily fused is represented;
respectively representing the values of the ith branch network output gate and the memory cell at the t moment; wiRepresenting a transformation matrix; n is the number of channels.
5. The method for predicting the life of a rolling bearing according to claim 1, wherein the specific construction of the multiple sensing information fusion layers in S4 is as follows:
in the formula: h isfuseThe state information after final fusion is obtained; conv () represents a convolution operation.
6. The method as claimed in claim 5, wherein the regression prediction layer in S5 is represented by H in S4fuseAs an input, a predicted remaining life percentage is obtained
The expression is as follows:
in the formula, Wr，brThe weight matrix and the bias vector of the regression prediction layer are respectively represented.
7. The method as claimed in claim 1, wherein the loss function in S6 is a mean square error loss function defined as:
in the formula, L represents an error value; k represents the number of training samples; y isiRepresenting a true remaining life value;
representing the predicted remaining life value.
8. A rolling bearing life prediction system, comprising:
the preprocessing module is used for preprocessing the multi-directional original vibration signal time sequence data of the bearing and determining a degradation starting point of the bearing; constructing a multi-branch feature extraction layer comprising a plurality of convolution long and short term memory networks, and respectively inputting the multi-directional original vibration signal time sequence data of the acquired multi-directional original vibration signal time sequence data into the networks for feature extraction to obtain vibration signal features;
the training module is used for transmitting the extracted vibration signal characteristics in each branch CLSTM to realize information preliminary fusion, finally fusing the hidden characteristics finally output by each CLSTM, performing regression prediction on the residual life of the bearing according to the obtained fusion characteristics, and optimizing a loss function of a network by using an adaptive distance estimation algorithm to obtain a trained rolling bearing life prediction model;
and the prediction module is used for predicting the residual life of the bearing to be tested and outputting the percentage of the residual life of the bearing to be tested to the total life of the degradation stage.
9. A terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 7 are implemented when the computer program is executed by the processor.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111648638.2A CN114564877B (en) | 2021-12-29 | 2021-12-29 | Rolling bearing life prediction method, system, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111648638.2A CN114564877B (en) | 2021-12-29 | 2021-12-29 | Rolling bearing life prediction method, system, equipment and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114564877A true CN114564877A (en) | 2022-05-31 |
| CN114564877B CN114564877B (en) | 2024-05-10 |
Family
ID=81711783
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111648638.2A Active CN114564877B (en) | 2021-12-29 | 2021-12-29 | Rolling bearing life prediction method, system, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114564877B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115982896A (en) * | 2023-03-21 | 2023-04-18 | 山东金帝精密机械科技股份有限公司 | Method and device for detecting service life of bearing retainer |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109726524A (en) * | 2019-03-01 | 2019-05-07 | 哈尔滨理工大学 | A kind of rolling bearing remaining life prediction technique based on CNN and LSTM |
| EP3620983A1 (en) * | 2018-09-05 | 2020-03-11 | Sartorius Stedim Data Analytics AB | Computer-implemented method, computer program product and system for data analysis |
| CN111832216A (en) * | 2020-04-14 | 2020-10-27 | 新疆大学 | Rolling bearing residual service life prediction method based on EEMD-MCNN-GRU |
| WO2021232655A1 (en) * | 2020-05-20 | 2021-11-25 | 国网河北省电力有限公司电力科学研究院 | Method for evaluating mechanical state of high-voltage shunt reactor on the basis of vibration feature |
-
2021
- 2021-12-29 CN CN202111648638.2A patent/CN114564877B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3620983A1 (en) * | 2018-09-05 | 2020-03-11 | Sartorius Stedim Data Analytics AB | Computer-implemented method, computer program product and system for data analysis |
| CN109726524A (en) * | 2019-03-01 | 2019-05-07 | 哈尔滨理工大学 | A kind of rolling bearing remaining life prediction technique based on CNN and LSTM |
| CN111832216A (en) * | 2020-04-14 | 2020-10-27 | 新疆大学 | Rolling bearing residual service life prediction method based on EEMD-MCNN-GRU |
| WO2021232655A1 (en) * | 2020-05-20 | 2021-11-25 | 国网河北省电力有限公司电力科学研究院 | Method for evaluating mechanical state of high-voltage shunt reactor on the basis of vibration feature |
Non-Patent Citations (2)
| Title |
|---|
| 姜洪开;邵海东;李兴球;: "基于深度学习的飞行器智能故障诊断方法", 机械工程学报, no. 07 * |
| 胡城豪;胡昌华;司小胜;杜党波;高旭东;: "基于MSCNN-LSTM的滚动轴承剩余寿命预测方法", 中国测试, no. 09 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115982896A (en) * | 2023-03-21 | 2023-04-18 | 山东金帝精密机械科技股份有限公司 | Method and device for detecting service life of bearing retainer |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114564877B (en) | 2024-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112131673B (en) | Engine surge fault prediction system and method based on fusion neural network model | |
| CN110232203B (en) | Knowledge distillation optimization RNN short-term power failure prediction method, storage medium and equipment | |
| CN116757534B (en) | Intelligent refrigerator reliability analysis method based on neural training network | |
| CN113298297B (en) | Wind power output power prediction method based on isolated forest and WGAN network | |
| Xia et al. | Multi-stage fault diagnosis framework for rolling bearing based on OHF Elman AdaBoost-Bagging algorithm | |
| CN112149554B (en) | Training and fault detection methods of fault classification model and related devices | |
| CN106909756A (en) | A kind of rolling bearing method for predicting residual useful life | |
| CN111965476B (en) | Low-voltage diagnosis method based on graph convolution neural network | |
| CN108879732B (en) | Transient stability evaluation method and device for power system | |
| CN113722985B (en) | Method and system for evaluating health state and predicting residual life of aero-engine | |
| CN102270302A (en) | Grey support vector machine-based multi-stress accelerated life testing forecasting method | |
| CN114490065A (en) | Load prediction method, device and equipment | |
| CN105137354A (en) | Motor fault detection method based on nerve network | |
| CN114841199A (en) | Power distribution network fault diagnosis method, device, equipment and readable storage medium | |
| CN115456306A (en) | Bus load prediction method, system, equipment and storage medium | |
| CN114564877B (en) | Rolling bearing life prediction method, system, equipment and readable storage medium | |
| Zeng et al. | Hybrid CEEMDAN-DBN-ELM for online DGA serials and transformer status forecasting | |
| CN116706907B (en) | Photovoltaic power generation prediction method based on fuzzy reasoning and related equipment | |
| CN113807951A (en) | Transaction data trend prediction method and system based on deep learning | |
| Chen et al. | Data-driven distribution network topology identification considering correlated generation power of distributed energy resource | |
| CN112633362A (en) | Self-adaptive fault diagnosis method, device, equipment and medium for rotary machine | |
| CN112203311B (en) | Network element abnormity diagnosis method, device, equipment and computer storage medium | |
| ul Hassan et al. | Online static security assessment for cascading failure using stacked De-noising Auto-encoder | |
| CN116008733A (en) | Single-phase grounding fault diagnosis method based on integrated deep neural network | |
| CN116384223A (en) | Nuclear equipment reliability assessment method and system based on intelligent degradation state identification |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |