CN109992872A - A kind of mechanical equipment method for predicting residual useful life based on stacking separation convolution module - Google Patents
A kind of mechanical equipment method for predicting residual useful life based on stacking separation convolution module Download PDFInfo
- Publication number
- CN109992872A CN109992872A CN201910235692.0A CN201910235692A CN109992872A CN 109992872 A CN109992872 A CN 109992872A CN 201910235692 A CN201910235692 A CN 201910235692A CN 109992872 A CN109992872 A CN 109992872A
- Authority
- CN
- China
- Prior art keywords
- mechanical equipment
- convolution
- feature
- input
- data
- 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
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
- 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
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Physics & Mathematics (AREA)
- Geometry (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computational Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
A kind of mechanical equipment method for predicting residual useful life based on stacking separation convolution module, first obtains the original vibration signal of mechanical equipment under different operating conditions, is pre-processed, and establishes the mechanical equipment predicting residual useful life model based on stacking separation convolution module;Using stacking separation convolution module extract original vibration signal in and the maximally related high-level characteristic feature of mechanical equipment health status;Then high-level characteristic feature is input to fully-connected network, obtains the predicting residual useful life value of mechanical equipment;The mean square error objective function for constructing predicting residual useful life model obtains optimal predicting residual useful life model by Adam optimization algorithm iteration update prediction model to training parameter;Pretreated mechanical equipment vibration signal is recently entered, mechanical equipment is completed and obtains predicting residual useful life;The present invention has the advantages that precision of prediction is higher, stability is more preferable, robustness is stronger.
Description
Technical field
The invention belongs to mechanical equipment predicting residual useful life technical fields, and in particular to one kind is based on stacking separation convolution mould
The mechanical equipment method for predicting residual useful life of block.
Background technique
With the raising of the level of IT application and industrial automation level, modern machinery and equipment is just towards high-performance, high-precision
Develop with high efficiency direction.Due to usually working under more complicated operating condition, mechanical equipment fault incidence is higher, causes to set
It is standby to be difficult to safe and reliable operation.Therefore, it is necessary to carry out predicting residual useful life to mechanical equipment, tieed up to formulate effective early stage
Shield scheme guarantees that equipment works normally.Since equipment monitoring point quantity is more, sample frequency is high and data collection lasts the originals such as long
Cause, mechanical health monitoring enter " big data " epoch, this brings stern challenge to the predicting residual useful life of mechanical equipment.
Therefore, it is necessary to invent a kind of mechanical equipment method for predicting residual useful life of new data-driven, ensure that mechanical equipment is on active service safely.
Theoretical by introducing deep learning, the mechanical equipment method for predicting residual useful life based on data-driven can be effectively
The degradation information of excavation machinery from monitoring data, achieves good prediction result.However, current is driven based on data
Dynamic mechanical equipment method for predicting residual useful life needs manual extraction feature, this process from vibration signal not only to need profession
Signal processing knowledge and expertise abundant, it is also desirable to pay a large amount of human cost.Meanwhile the monitoring number of mechanical equipment
According to being obtained by multiple sensors, different degrees of mechanical equipment degradation information is contained in different sensors data, is reflected
Interaction mechanism between different unit failures.But current method has ignored the phase between different sensors data
Guan Xing, thus cannot efficiently extract with the maximally related information of mechanical equipment health status, seriously affect under mechanical big data background
The accuracy of the predicting residual useful life result of mechanical equipment.
Summary of the invention
It is the shortcomings that in order to overcome the above prior art, a kind of based on stacking separation convolution mould it is an object of the invention to propose
The mechanical equipment method for predicting residual useful life of block excavates most related to its health status directly from the vibration signal of mechanical equipment
Information, automatically extract high-level characteristic feature, and the residue of prediction mechanical equipment is gone using these high-level characteristic features
Service life;This method has the advantages that precision of prediction is higher, stability is more preferable, robustness is stronger.
In order to achieve the above object, the technical scheme adopted by the invention is as follows:
A kind of mechanical equipment method for predicting residual useful life based on stacking separation convolution module, comprising the following steps:
1) original vibration signal of mechanical equipment under different operating conditions is obtainedWherein,M is sample of signal
Number, N be each vibration signal sample include data points, C is the number of vibrating sensor;Meanwhile building has deeply
The stacking of layer structure separates convolution module prediction model;
2) to original vibration signalIt is pre-processed: first to vibration signal xiZ-score normalized is carried out,
Then, by time window embedding strategy xiEffective time information data before is embedded into xiIn, the specific steps are as follows:
2.1) by Z-score to vibration signal xiIt is normalized, obtains monitoring data
Z-score normalizing operation expression formula is as follows:
In formula, x is initial data;For the mean value of initial data;σ is the standard deviation of initial data;x*For Z-score mark
Data after standardization;
2.2) setting time window size is S, passes through time window embedding strategy integral dataWith S-1 before it
Vibration signal sample, obtains dataI.e.
3) T separation convolution module is laminated, from pretreated original vibration signalMiddle extraction and mechanical equipment health
The maximally related high-level characteristic feature of situation, the specific steps are as follows:
3.1) input sampleIt is utilized respectively channel convolution kernel kcwWith a convolution kernel kpwWith input sampleIt is rolled up
Product obtains feature
3.2) to featureAverage pond is carried out, the average value of non-overlapping pond region interior element is obtained, obtains Chi Huahou
Feature be
3.3) by featureIt is input in T separation convolution modules with residual error connection of stacking, extraction is set with machinery
The specific implementation of the standby maximally related high-level characteristic feature of health status, each separation convolution module is as follows:
(a) it firstly, being handled using preactivate strategy the input of separation convolution module, passes sequentially through batch and normalizes
Layer and line rectification function (Rectified linear unit, ReLU) layer;
Batch normalization layer operation expression formula is as follows:
In formula, xl-1For the input for separating convolution module;yl-1Output after being normalized for batch;μB、Respectively input
xl-1Expectation and variance;γ, β are the reconstruction parameter that normalizes layer and can learn;
(b) data in step (a) by preactivate operation are input to separation convolutional layer, extract the deep layer expression of data
Feature
Separation convolution includes the convolution on the convolution sum point direction in channel direction, and calculation expression is as follows:
In formula,For the data after preactivate operation;For the output of convolution in channel direction;It is scrolled up for a side
Long-pending output;kcw、Indicate the convolution kernel and biasing in channel direction;kpw、Indicate the convolution kernel and biasing on point direction;c
Indicate the input channel of c-th of convolution operation;N indicates the output channel of n-th of convolution operation;
(c) step (a) and (b) is executed again, further extract the deep layer expression characteristic of dataLearn not simultaneous interpretation
Correlation between sensor data;
(d) to the deep layer expression characteristic of extractionAverage pond is carried out, the flat of non-overlapping pond region interior element is obtained
Mean value, the feature for obtaining Chi Huahou are
(e) by the feature of Chi HuahouIt is input to feature alignment layer, is carried out by squeeze operation and self-calibration operation
The recalibration of characteristic response, to obtain and the maximally related feature of mechanical equipment health status;
Squeeze operation refers to global average pond, and calculation expression is as follows:
In formula, H indicates the global length of compression excitation layer input;Indicate the input of compression excitation layer;
Self-calibration operation refers to the information content that each channel is estimated using adaptive door machine system, generates the power of corresponding channel
Weight, calculation expression are as follows:
In formula, σ (), δ () are respectively Sigmoid and ReLU activation primitive; Its
Middle r is the ratio of dimensionality reduction, and C is port number;
(f) finally, separation convolution module is connected using residual error, the calculation expression of output are as follows:
xl=xl-1+F(xl-1,Wl)
In formula, xl-1It is l layers of input, xlIt is l layers of output;F () is residual error function, and expression formula is as follows:
F(xl-1) :=H (xl-1)-xl-1
In formula, H (xl-1) it is expectation mapping;
4) the high-level characteristic feature of extraction is input in fully-connected network, obtains the life prediction value of mechanical equipment,
Specific step is as follows;
4.1) global average pond is carried out to the characteristic feature of extraction, the characteristic pattern of output is made all to contain only an element,
Obtain the feature of Chi Huahou
4.2) by the feature of Chi HuahouIt is laid into one-dimensional vectorUtilize fully-connected network calculating machine equipment
Remaining life preRULi;
5) it is based on Adam optimization algorithm, step 3) is repeated with the number of iterations N, iteration 4), is arranged and updates stacking separation convolution
The parameter of module and fully-connected network minimizes mean square error objective function to obtain optimal predicting residual useful life model:
In formula, yiFor the real surplus life-span value of mechanical equipment;
6) by pretreated mechanical equipment vibration signalIt is input in optimal predicting residual useful life model, predicts
The remaining life of mechanical equipment.
The invention has the benefit that
The present invention directly can extract high-level characteristic feature from original vibration signal, and consider different sensors number
According to the correlation in characteristic feature learns, and then it can accurately excavate and believe with the maximally related degeneration of mechanical equipment health status
Breath.Conventional method is overcome to be too dependent on expertise and domain knowledge and cannot efficiently extract and mechanical equipment health
The shortcomings that situation maximally related information, realize the Accurate Prediction of mechanical equipment remaining life under mechanical big data.
Detailed description of the invention
Fig. 1 is the flow chart of the method for the present invention.
Fig. 2 is the basic composite structural diagram for separating convolution module.
Fig. 3 is the predicting residual useful life result figure of rolling bearing under three kinds of operating conditions of embodiment.
Fig. 4 is embodiment rolling bearing predicting residual useful life performance comparison figure.
Specific embodiment
The present invention is elaborated further with reference to the accompanying drawing.
Referring to Fig.1, a kind of mechanical equipment method for predicting residual useful life based on stacking separation convolution module, including following step
It is rapid:
1) original vibration signal of mechanical equipment under different operating conditions is obtainedWherein,M is sample of signal
Number, N be each vibration signal sample include data points, C is the number of vibrating sensor;Meanwhile building has deeply
The stacking of layer structure separates convolution module prediction model;
2) to original vibration signalIt is pre-processed: first to vibration signal xiZ-score normalized is carried out,
Then, by time window embedding strategy xiEffective time information data before is embedded into xiIn, the specific steps are as follows:
2.1) by Z-score to vibration signal xiIt is normalized, obtains monitoring data
Z-score normalizing operation expression formula is as follows:
In formula, x is initial data;For the mean value of initial data;σ is the standard deviation of initial data;x*For Z-score mark
Data after standardization;
2.2) setting time window size is S, passes through time window embedding strategy integral dataWith S-1 before it
Vibration signal sample, obtains dataI.e.
3) T separation convolution module as shown in Figure 2 is laminated, from pretreated original vibration signalIt is middle extraction with
The maximally related high-level characteristic feature of mechanical equipment health status, the specific steps are as follows:
3.1) input sampleIt is utilized respectively channel convolution kernel kcwWith a convolution kernel kpwWith input sampleIt is rolled up
Product obtains feature
Above-mentioned convolution operation includes the convolution in the convolution sum point direction in channel direction, featureIt can be by following convolution table
It is calculated up to formula:
yc=kcw*xc+bc
In formula, xcFor pretreated vibration signal data;ycFor the output of convolution in channel direction;znFor on direction
The output of convolution;kcw、bcIndicate the convolution kernel and biasing in channel direction;kpw、bnIndicate the convolution kernel and biasing on point direction;
C indicates the input channel of c-th of convolution operation;N indicates the output channel of n-th of convolution operation;
3.2) to featureAverage pond is carried out, the average value of non-overlapping pond region interior element is obtained, obtains Chi Huahou
Feature be
The feature of Chi Huahou can be calculated by following expression:
xAP=pool (xcp,p,s)
In formula, xcpFeature after indicating convolution;xAPIndicate the feature of Chi Huahou;Pool () indicates down-sampling function;p,
The size and sliding step in s difference pond;
3.3) by featureIt is input in T separation convolution modules with residual error connection of stacking, extraction is set with machinery
The specific implementation of the standby maximally related high-level characteristic feature of health status, each separation convolution module is as follows:
(a) it firstly, being handled using preactivate strategy the input of separation convolution module, passes sequentially through batch and normalizes
Layer and line rectification function (Rectified linear unit, ReLU) layer;
Batch normalization layer operation expression formula is as follows:
In formula, xl-1For the input for separating convolution module;yl-1Output after being normalized for batch;μB、Respectively input
xl-1Expectation and variance;γ, β are the reconstruction parameter that normalizes layer and can learn;
(b) data in step (a) by preactivate operation are input to separation convolutional layer, extract the deep layer expression of data
Feature
Separation convolution includes the convolution on the convolution sum point direction in channel direction, and calculation expression is as follows:
In formula,For the data after preactivate operation;For the output of convolution in channel direction;It is scrolled up for a side
Long-pending output;kcw、Indicate the convolution kernel and biasing in channel direction;kpw、Indicate the convolution kernel and biasing on point direction;c
Indicate the input channel of c-th of convolution operation;N indicates the output channel of n-th of convolution operation;
(c) step (a) and (b) is executed again, further extract the deep layer expression characteristic of dataLearn not simultaneous interpretation
Correlation between sensor data;
(d) to the deep layer expression characteristic of extractionAverage pond is carried out, the flat of non-overlapping pond region interior element is obtained
Mean value, the feature for obtaining Chi Huahou are
(e) by the feature of Chi HuahouIt is input to feature alignment layer, is carried out by squeeze operation and self-calibration operation
The recalibration of characteristic response, to obtain and the maximally related feature of mechanical equipment health status;
Squeeze operation refers to global average pond, and calculation expression is as follows:
In formula, H indicates the global length of compression excitation layer input;Indicate the input of compression excitation layer;
Self-calibration operation refers to the information content that each channel is estimated using adaptive door machine system, generates the power of corresponding channel
Weight, calculation expression are as follows:
In formula, σ (), δ () are respectively Sigmoid and ReLU activation primitive; Its
Middle r is the ratio of dimensionality reduction, and C is port number;
(f) finally, separation convolution module is connected using residual error, the calculation expression of output are as follows:
xl=xl-1+F(xl-1,Wl)
In formula, xl-1It is l layers of input, xlIt is l layers of output;F () is residual error function, and expression formula is as follows:
F(xl-1) :=H (xl-1)-xl-1
In formula, H (xl-1) it is expectation mapping;
4) the high-level characteristic feature of extraction is input in fully-connected network, obtains the life prediction value of mechanical equipment,
Specific step is as follows;
4.1) global average pond is carried out to the characteristic feature of extraction, the characteristic pattern of output is made all to contain only an element,
Obtain the feature of Chi Huahou
4.2) by the feature of Chi HuahouIt is laid into one-dimensional vectorUtilize fully-connected network calculating machine equipment
Remaining life preRULi;
5) it is based on Adam optimization algorithm, step 3) is repeated with the number of iterations N, iteration 4), is arranged and updates stacking separation convolution
The parameter of module and fully-connected network minimizes mean square error objective function to obtain optimal predicting residual useful life model:
In formula, yiFor the real surplus life-span value of mechanical equipment;
6) by pretreated mechanical equipment vibration signalIt is input in optimal predicting residual useful life model, predicts
The remaining life of mechanical equipment.
Embodiment: using the rolling bearing in mechanical equipment as case, rolling bearing accelerated life test data are based on, to this
The validity of inventive method is verified.Rolling bearing accelerated life test data set used by the present embodiment includes 3 altogether
Subset respectively corresponds under three kinds of different working conditions, i.e. 12kN/2100rpm, 11kN/2250rpm and 10kN/2400rpm.
Wherein, every kind of operating condition includes the life cycle management vibration signal of 5 rolling bearings.As shown in table 1, the method for the present invention pair is being used
When roller bearing remaining life is predicted, using preceding 4 rolling bearing data under each operating condition as training dataset, finally
1 rolling bearing data is as test data set.
The parameter setting of stacking separation convolution module prediction model is as follows: the number T for separating convolution module is 3;When
Between window S be 5;Convolution kernel size in channel direction is 8, quantity 16;Pond area size and step-length are all arranged 4;Dimensionality reduction ratio
It is set as 16;It is laminated using 3 layers of separation convolution module;Small lot training number is 128;The number of iterations is selected as 100 times.Use this
Inventive method carries out predicting residual useful life to rolling bearing test data set under 3 kinds of operating conditions, and prediction result is as shown in figure 3, from Fig. 3
As can be seen that although true lifetime and bimetry deviation of the rolling bearing in early stage are larger, with pushing away for time in
It moves, the bimetry of rolling bearing gradually levels off to true lifetime, this illustrates that the method for the present invention can be effectively to rolling bearing
Carry out predicting residual useful life.Further to verify superiority of the invention, by the method for the present invention and based on depth confidence network
Method for predicting residual useful life (DBN), the method for predicting residual useful life (MCNN) based on multiple dimensioned convolutional neural networks compare,
Three kinds of methods are evaluated using score function and root-mean-square error estimated performance index, as a result as shown in Figure 4.From Fig. 4
As can be seen that two kinds of estimated performance index values of the method for the present invention are both less than in the predicting residual useful life of three test bearings
Other two kinds of prediction techniques illustrate that the predicting residual useful life precision of the method for the present invention is higher, stability is more preferable, robustness is stronger.
Table 1
It can be sent out by the predicting residual useful life result of the above rolling bearing and with the estimated performance comparison of two methods
Existing, the method for the present invention directly can extract high-level characteristic feature from original vibration signal using stacking separation convolution module, and
And correlation of the different sensors data in characteristic feature study is considered, and then can accurately excavate strong with mechanical equipment
The maximally related degradation information of health situation, effectively increases the precision of mechanical equipment predicting residual useful life, obtains more superior
Estimated performance.
Claims (1)
1. a kind of mechanical equipment method for predicting residual useful life based on stacking separation convolution module, which is characterized in that including following
Step:
1) original vibration signal of mechanical equipment under different operating conditions is obtainedWherein,M is of sample of signal
Number, N are the data points that each vibration signal sample includes, and C is the number of vibrating sensor;Meanwhile building has deep layer knot
The stacking of structure separates convolution module prediction model;
2) to original vibration signalIt is pre-processed: first to vibration signal xiZ-score normalized is carried out, so
Afterwards, by time window embedding strategy xiEffective time information data before is embedded into xiIn, the specific steps are as follows:
2.1) by Z-score to vibration signal xiIt is normalized, obtains monitoring data
Z-score normalizing operation expression formula is as follows:
In formula, x is initial data;For the mean value of initial data;σ is the standard deviation of initial data;x*For Z-score standardization
Data afterwards;
2.2) setting time window size is S, passes through time window embedding strategy integral dataWith the S-1 vibration letter before it
Number sample, obtains dataI.e.
3) T separation convolution module is laminated, from pretreated original vibration signalMiddle extraction and mechanical equipment health status
Maximally related high-level characteristic feature, the specific steps are as follows:
3.1) input sampleIt is utilized respectively channel convolution kernel kcwWith a convolution kernel kpwWith input sampleConvolution is carried out, is obtained
Obtain feature
3.2) to featureAverage pond is carried out, the average value of non-overlapping pond region interior element is obtained, obtains the spy of Chi Huahou
Sign is
3.3) by featureIt is input in T separation convolution modules with residual error connection of stacking, extracts and mechanical equipment health
The specific implementation of the maximally related high-level characteristic feature of situation, each separation convolution module is as follows:
(a) firstly, using preactivate strategy to separation convolution module input handle, pass sequentially through batch normalization layer and
Line rectification function (Rectified linear unit, ReLU) layer;
Batch normalization layer operation expression formula is as follows:
In formula, xl-1For the input for separating convolution module;yl-1Output after being normalized for batch;μB、Respectively input xl-1's
It is expected that and variance;γ, β are the reconstruction parameter that normalizes layer and can learn;
(b) data in step (a) by preactivate operation are input to separation convolutional layer, extract the deep layer expression characteristic of data
Separation convolution includes the convolution on the convolution sum point direction in channel direction, and calculation expression is as follows:
In formula,For the data after preactivate operation;For the output of convolution in channel direction;For convolution on direction
Output;kcw、Indicate the convolution kernel and biasing in channel direction;kpw、Indicate the convolution kernel and biasing on point direction;C is indicated
The input channel of c-th of convolution operation;N indicates the output channel of n-th of convolution operation;
(c) step (a) and (b) is executed again, further extract the deep layer expression characteristic of dataLearn different sensors number
Correlation between;
(d) to the deep layer expression characteristic of extractionAverage pond is carried out, the average value of non-overlapping pond region interior element is obtained,
The feature for obtaining Chi Huahou is
(e) by the feature of Chi HuahouIt is input to feature alignment layer, feature is carried out by squeeze operation and self-calibration operation
The recalibration of response, to obtain and the maximally related feature of mechanical equipment health status;
Squeeze operation refers to global average pond, and calculation expression is as follows:
In formula, H indicates the global length of compression excitation layer input;Indicate the input of compression excitation layer;
Self-calibration operation refers to the information content that each channel is estimated using adaptive door machine system, generates the weight of corresponding channel,
Calculation expression is as follows:
In formula, σ (), δ () are respectively Sigmoid and ReLU activation primitive; Wherein r is
The ratio of dimensionality reduction, C are port number;
(f) finally, separation convolution module is connected using residual error, the calculation expression of output are as follows:
xl=xl-1+F(xl-1,Wl)
In formula, xl-1It is l layers of input, xlIt is l layers of output;F () is residual error function, and expression formula is as follows:
F(xl-1) :=H (xl-1)-xl-1
In formula, H (xl-1) it is expectation mapping;
4) the high-level characteristic feature of extraction is input in fully-connected network, obtains the life prediction value of mechanical equipment, specifically
Steps are as follows;
4.1) global average pond is carried out to the characteristic feature of extraction, so that the characteristic pattern of output is all contained only an element, obtains
The feature of Chi Huahou
4.2) by the feature of Chi HuahouIt is laid into one-dimensional vectorUtilize the residue of fully-connected network calculating machine equipment
Service life preRULi;
5) it is based on Adam optimization algorithm, step 3) is repeated with the number of iterations N, iteration 4), is arranged and updates stacking separation convolution module
Mean square error objective function is minimized with the parameter of fully-connected network to obtain optimal predicting residual useful life model:
In formula, yiFor the real surplus life-span value of mechanical equipment;
6) by pretreated mechanical equipment vibration signalIt is input in optimal predicting residual useful life model, prediction machinery is set
Standby remaining life.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910235692.0A CN109992872B (en) | 2019-03-27 | 2019-03-27 | Mechanical equipment residual life prediction method based on stacked separation convolution module |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910235692.0A CN109992872B (en) | 2019-03-27 | 2019-03-27 | Mechanical equipment residual life prediction method based on stacked separation convolution module |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109992872A true CN109992872A (en) | 2019-07-09 |
CN109992872B CN109992872B (en) | 2020-07-28 |
Family
ID=67131710
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910235692.0A Active CN109992872B (en) | 2019-03-27 | 2019-03-27 | Mechanical equipment residual life prediction method based on stacked separation convolution module |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109992872B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111783362A (en) * | 2020-07-09 | 2020-10-16 | 哈尔滨工程大学 | Method and system for determining residual service life of electric gate valve |
CN112131760A (en) * | 2019-11-04 | 2020-12-25 | 中国人民解放军国防科技大学 | CBAM model-based prediction method for residual life of aircraft engine |
CN112881518A (en) * | 2021-01-08 | 2021-06-01 | 东冶及策河北能源技术有限公司 | Method for predicting residual life of dynamic filter compensator |
CN113686577A (en) * | 2021-08-17 | 2021-11-23 | 山东科技大学 | Bearing fault diagnosis method based on rapid nonlinear sparse spectrum |
CN115048873A (en) * | 2022-08-12 | 2022-09-13 | 太原科技大学 | Residual service life prediction system for aircraft engine |
CN116579505A (en) * | 2023-07-12 | 2023-08-11 | 中国科学院空间应用工程与技术中心 | Electromechanical equipment cross-domain residual life prediction method and system without full life cycle sample |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060064291A1 (en) * | 2004-04-21 | 2006-03-23 | Pattipatti Krishna R | Intelligent model-based diagnostics for system monitoring, diagnosis and maintenance |
CN103793752A (en) * | 2013-09-13 | 2014-05-14 | 中国人民解放军第二炮兵工程大学 | Degradation modeling-based equipment failure number prediction method |
CN106769048A (en) * | 2017-01-17 | 2017-05-31 | 苏州大学 | Self adaptation depth confidence network Method for Bearing Fault Diagnosis based on Nesterov momentum methods |
CN107451760A (en) * | 2017-09-04 | 2017-12-08 | 西安交通大学 | Based on when the limited Boltzmann machine of window sliding Fault Diagnosis of Roller Bearings |
CN108106830A (en) * | 2017-12-13 | 2018-06-01 | 武汉科技大学 | A kind of Variable Speed Rotating Machinery method for diagnosing faults based on time-frequency spectrum segmentation |
CN109376401A (en) * | 2018-09-29 | 2019-02-22 | 西安交通大学 | A kind of adaptive multi-source information preferably with the mechanical method for predicting residual useful life that merges |
CN109460618A (en) * | 2018-11-13 | 2019-03-12 | 华中科技大学 | A kind of rolling bearing remaining life on-line prediction method and system |
US20190086911A1 (en) * | 2017-09-15 | 2019-03-21 | General Electric Company | Machine health monitoring, failure detection and prediction using non-parametric data |
-
2019
- 2019-03-27 CN CN201910235692.0A patent/CN109992872B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060064291A1 (en) * | 2004-04-21 | 2006-03-23 | Pattipatti Krishna R | Intelligent model-based diagnostics for system monitoring, diagnosis and maintenance |
CN103793752A (en) * | 2013-09-13 | 2014-05-14 | 中国人民解放军第二炮兵工程大学 | Degradation modeling-based equipment failure number prediction method |
CN106769048A (en) * | 2017-01-17 | 2017-05-31 | 苏州大学 | Self adaptation depth confidence network Method for Bearing Fault Diagnosis based on Nesterov momentum methods |
CN107451760A (en) * | 2017-09-04 | 2017-12-08 | 西安交通大学 | Based on when the limited Boltzmann machine of window sliding Fault Diagnosis of Roller Bearings |
US20190086911A1 (en) * | 2017-09-15 | 2019-03-21 | General Electric Company | Machine health monitoring, failure detection and prediction using non-parametric data |
CN108106830A (en) * | 2017-12-13 | 2018-06-01 | 武汉科技大学 | A kind of Variable Speed Rotating Machinery method for diagnosing faults based on time-frequency spectrum segmentation |
CN109376401A (en) * | 2018-09-29 | 2019-02-22 | 西安交通大学 | A kind of adaptive multi-source information preferably with the mechanical method for predicting residual useful life that merges |
CN109460618A (en) * | 2018-11-13 | 2019-03-12 | 华中科技大学 | A kind of rolling bearing remaining life on-line prediction method and system |
Non-Patent Citations (3)
Title |
---|
T H LOUTAS 等: "Remaining useful life estimation in rolling bearings utilizing data-driven probabilistic E-support vectors regression", 《IEEE TRANSACTIONS ON RELIABILITY》 * |
刘小勇: "基于深度学习的机械设备退化状态建模及剩余寿命预测研究", 《中国优秀硕士学位论文全文数据库(工程科技Ⅱ辑)》 * |
郑建飞: "考虑不完全维护影响的随机退化设备剩余寿命预测", 《电子学报 》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112131760A (en) * | 2019-11-04 | 2020-12-25 | 中国人民解放军国防科技大学 | CBAM model-based prediction method for residual life of aircraft engine |
CN111783362A (en) * | 2020-07-09 | 2020-10-16 | 哈尔滨工程大学 | Method and system for determining residual service life of electric gate valve |
CN112881518A (en) * | 2021-01-08 | 2021-06-01 | 东冶及策河北能源技术有限公司 | Method for predicting residual life of dynamic filter compensator |
CN113686577A (en) * | 2021-08-17 | 2021-11-23 | 山东科技大学 | Bearing fault diagnosis method based on rapid nonlinear sparse spectrum |
CN115048873A (en) * | 2022-08-12 | 2022-09-13 | 太原科技大学 | Residual service life prediction system for aircraft engine |
CN116579505A (en) * | 2023-07-12 | 2023-08-11 | 中国科学院空间应用工程与技术中心 | Electromechanical equipment cross-domain residual life prediction method and system without full life cycle sample |
CN116579505B (en) * | 2023-07-12 | 2023-10-13 | 中国科学院空间应用工程与技术中心 | Electromechanical equipment cross-domain residual life prediction method and system without full life cycle sample |
Also Published As
Publication number | Publication date |
---|---|
CN109992872B (en) | 2020-07-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109992872A (en) | A kind of mechanical equipment method for predicting residual useful life based on stacking separation convolution module | |
CN112149316B (en) | Aero-engine residual life prediction method based on improved CNN model | |
CN110738360B (en) | Method and system for predicting residual life of equipment | |
CN107544904B (en) | Software reliability prediction method based on deep CG-LSTM neural network | |
CN109297689B (en) | Large-scale hydraulic machinery intelligent diagnosis method introducing weight factors | |
CN108469507B (en) | Effluent BOD soft measurement method based on self-organizing RBF neural network | |
CN109102032A (en) | A kind of pumping plant unit diagnostic method based on depth forest and oneself coding | |
CN111339712A (en) | Method for predicting residual life of proton exchange membrane fuel cell | |
CN108344564A (en) | A kind of state recognition of main shaft features Testbed and prediction technique based on deep learning | |
CN110210621A (en) | A kind of object detection method based on residual error network improvement | |
CN111639430A (en) | Digital twin driven natural gas pipeline leakage identification system | |
CN111768000A (en) | Industrial process data modeling method for online adaptive fine-tuning deep learning | |
CN111783362A (en) | Method and system for determining residual service life of electric gate valve | |
CN109993281A (en) | A kind of causality method for digging based on deep learning | |
CN108805195A (en) | A kind of motor group method for diagnosing faults based on two-value deep-neural-network | |
CN112558185A (en) | Bidirectional GRU typhoon track intelligent prediction and forecast system based on attention mechanism, computer equipment and storage medium | |
CN114492642A (en) | Mechanical fault online diagnosis method for multi-scale element depth residual shrinkage network | |
CN115797297A (en) | Post-earthquake building structure health diagnosis multitask learning method | |
CN110486009B (en) | Automatic parameter reverse solving method and system for infinite stratum | |
CN106709829B (en) | Learning situation diagnosis method and system based on online question bank | |
CN112598186B (en) | Improved LSTM-MLP-based small generator fault prediction method | |
CN108960332A (en) | A kind of on-line monitoring method based on multidirectional the analysis of main elements | |
CN114841063A (en) | Aero-engine residual life prediction method based on deep learning | |
CN114021620A (en) | Electrical submersible pump fault diagnosis method based on BP neural network feature extraction | |
CN115391443B (en) | Method, device and system for providing artificial intelligent data of Internet of things equipment and terminal equipment |
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 |