CN108645615B - A kind of Adaptive Fuzzy Neural-network gear method for predicting residual useful life - Google Patents
A kind of Adaptive Fuzzy Neural-network gear method for predicting residual useful life Download PDFInfo
- Publication number
- CN108645615B CN108645615B CN201810304816.1A CN201810304816A CN108645615B CN 108645615 B CN108645615 B CN 108645615B CN 201810304816 A CN201810304816 A CN 201810304816A CN 108645615 B CN108645615 B CN 108645615B
- Authority
- CN
- China
- Prior art keywords
- layer
- fuzzy
- gear
- output
- network
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/02—Gearings; Transmission mechanisms
- G01M13/021—Gearings
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/02—Gearings; Transmission mechanisms
- G01M13/028—Acoustic or vibration analysis
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
A kind of modified Adaptive Fuzzy Neural-network gear method for predicting residual useful life, belongs to Mechanical Reliability technical field, is characterized in that implementation steps are as follows: 1, using vibrating sensor to gear degeneration real-time monitoring;2, feature extraction is carried out to gear fatigue state, slump evaluations is carried out to gear wear degraded performance;3, fuzzy system and neural network are combined, with the deficiency of Neural Network Self-learning mechanism Compensation Fuzzy control system, establishes a kind of fuzzy message fuzzy neural network;4, memory unit is added in all nodes of Fuzzy Processing layer, it by last moment imformation memory and is applied in output this moment, information is made to continue to save, reinforce information forward-backward correlation, predicted value and actual value deviation are reduced, modified Adaptive Fuzzy Neural-network forecasting system is established;5, gear remaining life is predicted according to training modified Adaptive Fuzzy Neural-network;Advantage is can effectively to predict gear degenerate state and real-time remaining life, provides foundation for gear preventative maintenance.
Description
Technical field
The invention belongs to Mechanical Reliability technical fields, and in particular to a kind of gear method for predicting residual useful life,
Background technique
Gear is the component of machine of passing movement and power, drives the progressively engaged change campaign of the gear teeth by motor
Size and Orientation in turn hands on the power between roller bearing, is mostly applied in mechanical equipment with closing forms such as gear-boxes, phase
For the transmission mechanism of other forms, gear drive transmitting has the range of peripheral speed and power wide, high-efficient, can guarantee perseverance
Fixed transmission ratio, it is safe and reliable the advantages that, nowadays properties of product are continuously improved, and the structure of mechanical equipment system is also exquisite multiple therewith
It is miscellaneous, gear it is easy to appear vibration frequencies when long-term load is operated failures such as high, abrasion or broken teeth, crackle, the study found that more
Several gearbox faults is all as caused by gear, and the quality of running state of gear box directly affects the normal fortune of machinery equipment
Make, once equipment part is unable to normal operation, it is possible to and it damages whole equipment or even influences entire production process, cause shutdown etc.
Economic loss even results in catastrophic casualties, therefore, carries out predicting residual useful life to gear, is to ensure mechanical equipment
The important measures of safe and efficient running and raising product quality.
Summary of the invention
It is an object of the present invention to provide a kind of modified Adaptive Fuzzy Neural-network gear method for predicting residual useful life, this calculations
Method has merged the real-time monitoring information of multi-measuring point, increases memory unit in obscuring on node layer, simultaneously by last moment imformation memory
It is applied in output this moment, effectively improves the precision of prediction of network model, improved prediction model is with the number of iterations
Increase, error reduces a lot compared to traditional adaptive neural network fuzzy system.
The invention is realized in this way including following implementation steps:
Step 1 installs acceleration transducer in gear-box, obtains the Real-time Monitoring Data of characterization gear condition;
Acceleration transducer is mounted on the bearing block position of main examination case, the mounting temperature sensor in gear-box, in main examination
Noise transducer is installed in the surface of case,
Step 2 carries out feature extraction to the fatigue state of gear, carries out slump evaluations to gear wear degraded performance;
Slump evaluations are carried out to gear wear degraded performance using square amplitude, in each sampling time Δ t/length,
The square amplitude x of the time series of discrete random signalrms(Δ t) may be expressed as:
Δ t is the sampling time in formula;N=Fs× Δ t, FsFor sample frequency, n is sampling number, and ∑ indicates summation, s ∈
(1,2,3....n), xsIt (t) is sampled value;
Step 3 combines fuzzy system and neural network, not with the self-study mechanism Compensation Fuzzy system of neural network
Foot establishes a kind of fuzzy neural network that can handle fuzzy message;
Step 4 establishes a kind of fuzzy neural network that can handle fuzzy message, and note is added in all nodes of Fuzzy Processing layer
Recall unit, by last moment imformation memory and is applied in the output at current time, makes information continue to save, reinforce information
Forward-backward correlation establishes a kind of modified Adaptive Fuzzy Neural-network, self-adaptive processing fuzzy message, neural network input marking
For { x0,x1,…,xt, the output token of hidden layer is { s0,s1,…,st, output token is { y0,y1,…,yt, x0,x1,…,
xtIndicate the input layer variable that neural network is walked from 0 to t, s0,s1,…,stIndicate the hidden layer variable walked from 0 to t, y0,
y1,…,ytIndicate that from input layer to hidden layer, finally to output layer, memory is added in the output layer variable walked from 0 to t, information flow
After unit, meeting guidance information is returned from output unit hides layer unit, and the input of hidden layer not only has this layer input, also comprising upper
The state of one hidden layer, the i.e. node interconnection of hidden layer can also connect certainly, stState, s are walked for the t of hidden layert=f (Uxt+
Wst-1), wherein f is activation primitive, and U is weight of the input layer to hidden layer, and W is weight of the hidden layer to hidden layer, is being calculated
s1That is first hiding layer state, needs to use s0, but be not present, 0 is set in the implementation,
1st layer: choosing 4 variable { xt-3r,xt-2r,xt-r,xt, wherein xt-3r,xt-2r,xt-r,xtFor 4 input state variables
Amount, Aj,BjFor each input distribute 2 subordinating degree function values, i.e. j=1,2, produce 16 if-then fuzzy rules, obscure
Rule is the linear combination of input variable, if x1It is Aj, x2It is Bj, then yl=c1,lx2+c2,lx1+c3,l, l=1 in formula, 2,3,
4, c1,l,c2,l,c3,lFor consequent parameter, ylFor by the output of the l articles fuzzy rule as a result,It is by the 1st layer of fuzzy rule
Output is as a result, input variable is blurred are as follows:
X in formulaiInput variable is represented,It is arbitrary parametrization membership function, is taken as sigmoid function,
After memory unit is added,
The membership function value for being added and inputting after memory unit is represented, indicates input variable xiIt is under the jurisdiction of AjDegree,
Exp indicates that the truth of a matter is the exponential function using natural constant e the bottom of as, bji,mjiIt is premise parameter, its value variation will affect
The shape of sigmoid function, xi (1)It is the 1st layer of input variable, xi (2)It is the 2nd layer of input variable, θji (2)It is to be fed back in the 2nd layer
Weight, initial value 0, is continued to optimize in an iterative process,It is a delay cell, it can be by last moment data
Included in device status information remain into subsequent time, the 2nd layer: fuzzy operator calculates the relevance grade of each rule,
It is denoted as by the output result of the 2nd layer of fuzzy ruleAlso simplification is denoted as ωl, Π is product signs,
3rd layer: the relevance grade of each rule is normalized,
It is denoted as by the output result of the 3rd layer of fuzzy ruleAlso simplification is denoted as∑ is summation number,
4th layer: the output of each rule is calculated,
Wherein it is denoted as by the output result of the 4th layer of fuzzy rulec1,l,c2,l,c3,l,c4,l,c5,lReferred to as conclusion is joined
Number, fl=c1,lx4+c2,lx3+c3,lx2+c4,lx1+c5,lFor the linear combination of input variable,
5th layer: the output of computing system,
Wherein it is denoted as by the output result of the 5th layer of fuzzy rule
Then modified Adaptive Fuzzy Neural-network prediction model output result is denoted as Y:
Training data input network is corrected into each parameter using hybrid algorithm training network, first gives bji, mjiInitial value is assigned,
C is estimated by least square method1,l,c2,l,c3,l,c4,l,c5,l, finally using gradient descent method backpropagation systematic error to correct
bji,mji, parameter θ is added during blurring in modified Adaptive Fuzzy Neural-network prediction modelji (2), system is first
Each parameter, θ are corrected when secondary operation in the manner described aboveji (2)It is 0;After iteration starts, the value that can be blurred last moment is rolled up
Enter into blurring output this moment, value 0.9 learns to repair from training sample automatically using Neural Network Self-learning ability
Positive weight variable, adjusts subordinating degree function, generates fuzzy rule, practical defeated by constantly learning to make the response of model constantly to approach
Out;
Step 5 predicts gear using trained modified Adaptive Fuzzy Neural-network prediction model input test data
State, by the degenerate state value predicted and known degenerate state fault threshold can solve for the first time reach fault threshold when
Between.
Invention advantage:
The present invention proposes a kind of modified Adaptive Fuzzy Neural-network gear method for predicting residual useful life, in fuzzy neural
Memory unit is added in the fuzzy node layer of network, the facility information for including of last moment can be remained into subsequent time and answered
It uses in output, improves the precision of prediction of entire model, apply the invention in the gear predicting residual useful life of gear housing,
Establish prediction model using sample data, be applied in test data show it is good as a result, gear remaining life and each shadow
Inner link and rule between the factor of sound, can make gear remaining life and be effectively predicted, and the present invention proposes modified certainly
Adapt to fuzzy neural network gear method for predicting residual useful life convergence, error precision, in terms of better than tradition
Fuzzy neural network prediction technique.
Detailed description of the invention
Fig. 1 is the real-time method for predicting residual useful life flow chart of middle gear of the embodiment of the present invention;
Fig. 2 is the prediction of modified Adaptive Fuzzy Neural-network gear predicting residual useful life model in the embodiment of the present invention
Flow chart;
Fig. 3 is modified Adaptive Fuzzy Neural-network gear predicting residual useful life model structure in the embodiment of the present invention
Figure, wherein M indicates the 2nd layer, and T indicates the 3rd layer, and N indicates the 4th layer, and O indicates the 5th layer, and X indicates the input terminal of memory unit, S table
Show the hidden layer of memory unit, Z-1Indicate the feedback of memory unit, xi (1)It is the 1st layer of input variable, xi (2)It is that the 2nd layer of input becomes
Amount, θji (2)It is the weight fed back in the 2nd layer, U(2)It is weight of the input layer to hidden layer;
Fig. 4 is modified Adaptive Fuzzy Neural-network structure chart in the embodiment of the present invention;
It is the model ratio that the output of training data and model export in different frequency of training that Fig. 5, which is in the embodiment of the present invention,
Compared with figure, pn2 is the reality output of training data in embodiment in figure, and Y- is model output;Fig. 5 (a) is that frequency of training is 200
When, the output of the training data figure compared with model output, Fig. 5 (b) is frequency of training when being 800, the output of training data with
Figure is compared in model output, and Fig. 5 (c) is frequency of training when being 1000, figure compared with the output of training data is exported with model;
Fig. 6 is modified Adaptive Fuzzy Neural-network gear predicting residual useful life model in the embodiment of the present invention in difference
When frequency of training, the Error Graph of the output of training data and model output;Fig. 6 (a) be frequency of training be 200 when, training data
Output with model output Error Graph, Fig. 6 (b) be frequency of training be 800 when, training data output with model output error
Figure;When Fig. 6 (c) frequency of training is 1000, the Error Graph of training data output and model output;
Fig. 7 is modified Adaptive Fuzzy Neural-network gear predicting residual useful life model in the embodiment of the present invention in difference
Test data output figure compared with model output when frequency of training, y1 are that test data reality output Y11 is model output;Figure
7 (a) be frequency of training be 200 when, test data output with model output compared with figure, Fig. 7 (b) be frequency of training be 800 when,
Test data output figure compared with model output;When Fig. 7 (c) frequency of training is 1000, test data output is exported with model
Comparison figure;
Fig. 8 is modified Adaptive Fuzzy Neural-network gear predicting residual useful life model in the embodiment of the present invention in difference
When frequency of training, the Error Graph of test data output and model output;Fig. 8 (a) be frequency of training be 200 when, test data is defeated
Out with model output Error Graph, Fig. 8 (b) be frequency of training be 800 when, test data output with model output Error Graph;
When Fig. 8 (c) frequency of training is 1000, the Error Graph of test data output and model output;
Fig. 9 is modified Adaptive Fuzzy Neural-network gear predicting residual useful life model interception in the embodiment of the present invention
In 200 groups of data, permission precision is 1 × e-4, test data output figure compared with model output, x2 is that test data is actually defeated
Out, Y11 is model output;
Figure 10 is modified Adaptive Fuzzy Neural-network gear predicting residual useful life model interception in the embodiment of the present invention
200 groups of data in, permission precision be 1 × e-4, the Error Graph of test data output and model output.
Specific embodiment
An embodiment of the present invention is described further with reference to the accompanying drawing:
In the embodiment of the present invention, the gear method for predicting residual useful life based on fuzzy neural network, method flow diagram such as Fig. 1
It is shown, comprising the following steps:
Step 1 carries out fatigue test to gear, obtains the Real-time Monitoring Data that characterization gear is degenerated:
Gear fatigue life test uses power flow blocking test rack, and the center of testing stand is away from for 150mm, motor
Revolving speed is 1200r/min, and test process is monitored cabinet vibration, oil temperature and noise etc., uses material for alloy in test
Steel, tooth face hardness are the hardened face gear of 58-61HRC, are surface-treated as carburizing and quenching, staggeredly overlap engagement side using front and back sides
Formula, main examination case module m=3, number of teeth z1=z2=50, pressure angle α=20 °, facewidth 29mm, the real work facewidth 13~
14mm;Accompanying examination case number of gear teeth is z3=z4=24, and pressure angle α=20 °, work facewidth 20mm, and gear mesh frequency has 2,
Main examination case gear is 1000Hz, and accompanying examination case gear is 480Hz, and test lubricating oil uses L-CKC320 Industrial Closed gear oil;
11 sensors are arranged in test altogether, and acceleration transducer is mounted on the bearing block position of main examination case, in gear-box
Mounting temperature sensor installs noise transducer in the surface of main examination case, and No. 1~No. 4 acceleration transducers are arranged in main examination
The radial direction of axle box bearing seat, No. 7 and No. 8 acceleration transducers are arranged in the axial direction of main examination case, No. 5 and No. 6 acceleration transducer cloth
It sets in the radial direction for accompanying examination axle box bearing seat;No. 9 and No. 10 sensors are sonic transducer, be arranged in main examination case and accompanying try case just on
At side about 40cm;No. 11 sensors are the temperature sensors for testing lubricating oil temperature, are arranged in main examination cabinet, test in test
Lubricating oil temperature, sample frequency 25.6kHz, sampling time 60s, sampling interval 9min are using conventional method in groups in this test
The mode of dead load carries out, torque 822.7N.M, determines the gear failure when testing gear and broken teeth occurring;
Step 2 carries out feature extraction to the fatigue state of gear, carries out slump evaluations to gear wear degraded performance;
Detachment tooth position in gear fatigue test is selected to set recently and in bearing block location arrangements when carrying out life prediction
463 groups of vibration signals of No. 4 sensors output carry out feature extraction, and sample frequency 25.6kHz in test, sampling time 60s are adopted
Sample interval 9min converts sampled point number for monitoring time, and the square amplitude curve of No. 4 sensors is after removing individual singular points
Overall variation trend can reflect corresponding with the vibrational energy relationship of test each monitoring point abrasion condition of gear, the curve contain from
Start the square amplitude of vibration signal that completion fatigue test broken teeth occurs at 77.2h for running in stage, by gear fatigue test
Know broken teeth Shi Junfang amplitude xrms(T*)=77.375 are denoted as degradation, T*For the time for reaching fault threshold for the first time, T*=
77.2h, using 4# sensing data as modified Adaptive Fuzzy Neural-network mode input, square amplitude (Root Mean
Square, RMS) as model output xrms(Δ t) assesses gear degradation characteristics;
Step 3 combines fuzzy system and neural network, not with the self-study mechanism Compensation Fuzzy system of neural network
Foot, establishes a kind of fuzzy neural network of fuzzy message;
The building of step 4, modified Adaptive Fuzzy Neural-network gear predicting residual useful life model is divided into model training
Stage and model measurement stage were trained training data input prediction model for model training using preceding 200 groups of data
Journey acquires the optimal value of each parameter of network, and 100 groups of data are used for model measurement, and allowable error precision is 1 × e-4Carry out gear fortune
The prediction of row state,
Training learns that the number of membership function value influences training result, and training error can be reduced by increasing number, but will increase
Calculation amount, this test chooses 5 and obtains preferable training result, when being individually below 200,800,1000 for frequency of training, instruction
Practice output and model output (Fig. 5), the training data output and model output error (Fig. 6), test data output and mould of data
Type output (Fig. 7), test data exports and model output error (Fig. 8),
The modified Adaptive Fuzzy Neural-network gear predicting residual useful life model known to Fig. 5,6,7,8 has good
Fitting effect, with the increase of frequency of training, error has good convergence, under parameter the same terms, fuzznet
The error of network gear predicting residual useful life model and modified Adaptive Fuzzy Neural-network gear predicting residual useful life model essence
Degree is as shown in table 1:
12 kinds of model training error precisions of table compare
As seen from the above table, in the identical situation of frequency of training, modified Adaptive Fuzzy Neural-network gear remaining life
Prediction model is lower than the error precision of traditional fuzzy neural network gear predicting residual useful life model, permission error be 1 ×
e-4When, traditional training error needs iteration 864 times, and modified Adaptive Fuzzy Neural-network gear predicting residual useful life need to change
Generation 424 times.
Step 5 carries out gear predicting residual useful life according to gear condition estimation and known gear distress threshold value;
After can be to the monitoring point of system using modified Adaptive Fuzzy Neural-network gear predicting residual useful life model
Degenerate state predicted, can solve and arrive for the first time by the degenerate state value and known degenerate state fault threshold that predict
Up to the time of fault threshold, the gear remaining life as predicted.
In conclusion the present invention proposes modified Adaptive Fuzzy Neural-network gear predicting residual useful life algorithm, in mould
Memory unit is added in paste layer node, the facility information for including of last moment can be remained into subsequent time and be applied to output
On, convergence, error precision, in terms of be better than traditional fuzzy neural network, the remaining longevity in real time can be improved
Order prediction accuracy.
Claims (1)
1. a kind of Adaptive Fuzzy Neural-network gear method for predicting residual useful life, it is characterised in that implementation steps are:
Step 1 installs acceleration transducer in gear-box, obtains the Real-time Monitoring Data of characterization gear condition;
Acceleration transducer is mounted on the bearing block position of main examination case, the mounting temperature sensor in gear-box, in main examination case
Noise transducer is installed in surface,
Step 2 carries out feature extraction to the fatigue state of gear, carries out slump evaluations to gear wear degraded performance;
Slump evaluations are carried out to gear wear degraded performance using square amplitude, it is discrete in each sampling time Δ t/length
The square amplitude x of the time series of random signalrms(Δ t) may be expressed as:
Δ t is the sampling time in formula;N=Fs× Δ t, FsFor sample frequency, n is sampling number, and ∑ indicates to sum, s ∈ (1,2,
3....n), xsIt (t) is sampled value;
Step 3 combines fuzzy system and neural network, with the deficiency of the self-study mechanism Compensation Fuzzy system of neural network,
Establish a kind of fuzzy neural network that can handle fuzzy message;
Step 4 establishes a kind of fuzzy neural network that can handle fuzzy message, and it is single that memory is added in all nodes of Fuzzy Processing layer
Member by last moment imformation memory and is applied in the output at current time, information is made to continue to save, reinforces the front and back of information
Association, establishes a kind of modified Adaptive Fuzzy Neural-network, self-adaptive processing fuzzy message, neural network input marking is
{x0,x1,…,xt, the output token of hidden layer is { s0,s1,…,st, output token is { y0,y1,…,yt, x0,x1,…,xt
Indicate the input layer variable that neural network is walked from 0 to t, s0,s1,…,stIndicate the hidden layer variable walked from 0 to t, y0,
y1,…,ytIndicate that from input layer to hidden layer, finally to output layer, memory is added in the output layer variable walked from 0 to t, information flow
After unit, meeting guidance information is returned from output unit hides layer unit, and the input of hidden layer not only has this layer input, also comprising upper
The state of one hidden layer, the i.e. node interconnection of hidden layer can also connect certainly, stState, s are walked for the t of hidden layert=f (Uxt+
Wst-1), wherein f is activation primitive, and U is weight of the input layer to hidden layer, and W is weight of the hidden layer to hidden layer, is being calculated
s1That is first hiding layer state, needs to use s0, but be not present, 0 is set in the implementation,
1st layer: choosing 4 variable { xt-3r,xt-2r,xt-r,xt, wherein xt-3r,xt-2r,xt-r,xtFor 4 input state variables,
Aj,BjFor each input distribute 2 subordinating degree function values, i.e. j=1,2, produce 16 if-then fuzzy rules, fuzzy rule
For the linear combination of input variable, if x1It is Aj, x2It is Bj, then yl=c1,lx2+c2,lx1+c3,l, l=1 in formula, 2,3,4,
c1,l,c2,l,c3,lFor consequent parameter, ylFor by the output of the l articles fuzzy rule as a result,It is by the defeated of the 1st layer of fuzzy rule
Out as a result, input variable is blurred are as follows:
X in formulaiInput variable is represented,It is arbitrary parametrization membership function, is taken as sigmoid function,
After memory unit is added,
The membership function value for being added and inputting after memory unit is represented, indicates input variable xiIt is under the jurisdiction of AjDegree, exp
Indicate that the truth of a matter is the exponential function using natural constant e the bottom of as, bji, mjiIt is premise parameter, its value variation will affect
The shape of sigmoid function, xi (1)It is the 1st layer of input variable, xi (2)It is the 2nd layer of input variable, θji (2)It is to be fed back in the 2nd layer
Weight, initial value 0, is continued to optimize in an iterative process,It is a delay cell, it can be by last moment data
Included in device status information remain into subsequent time,
2nd layer: fuzzy operator calculates the relevance grade of each rule,
It is denoted as by the output result of the 2nd layer of fuzzy ruleAlso simplification is denoted as ωl, Π is product signs,
3rd layer: the relevance grade of each rule is normalized,
It is denoted as by the output result of the 3rd layer of fuzzy ruleAlso simplification is denoted as∑ is summation number,
4th layer: the output of each rule is calculated,
Wherein it is denoted as by the output result of the 4th layer of fuzzy rulec1,l,c2,l,c3,l,c4,l,c5,lReferred to as consequent parameter, fl=
c1,lx4+c2,lx3+c3,lx2+c4,lx1+c5,lFor the linear combination of input variable,
5th layer: the output of computing system,
Wherein it is denoted as by the output result of the 5th layer of fuzzy rule
Then modified Adaptive Fuzzy Neural-network prediction model output result is denoted as Y:
Training data input network is corrected into each parameter using hybrid algorithm training network, first gives bji, mjiInitial value is assigned, by minimum
Square law estimates c1,l,c2,l,c3,l,c4,l,c5,l, finally using gradient descent method backpropagation systematic error to correct bji, mji,
Parameter θ is added during blurring in modified Adaptive Fuzzy Neural-network prediction modelji (2), system is in first operation
Each parameter, θ are corrected in the manner described aboveji (2)It is 0;After iteration starts, the value that last moment is blurred can be rolled into this moment
Blurring output in, value 0.9, utilize Neural Network Self-learning ability automatically from training sample study amendment weight become
Amount adjusts subordinating degree function, fuzzy rule is generated, by constantly learning to make the response of model constantly to approach reality output;
Step 5 predicts gear condition using trained modified Adaptive Fuzzy Neural-network prediction model input test data,
The time for reaching fault threshold for the first time can be solved by the degenerate state value predicted and known degenerate state fault threshold.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810304816.1A CN108645615B (en) | 2018-04-08 | 2018-04-08 | A kind of Adaptive Fuzzy Neural-network gear method for predicting residual useful life |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810304816.1A CN108645615B (en) | 2018-04-08 | 2018-04-08 | A kind of Adaptive Fuzzy Neural-network gear method for predicting residual useful life |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108645615A CN108645615A (en) | 2018-10-12 |
CN108645615B true CN108645615B (en) | 2019-11-22 |
Family
ID=63745841
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810304816.1A Active CN108645615B (en) | 2018-04-08 | 2018-04-08 | A kind of Adaptive Fuzzy Neural-network gear method for predicting residual useful life |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108645615B (en) |
Families Citing this family (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109740809B (en) * | 2018-12-27 | 2021-03-23 | 北京五维星宇科技有限公司 | Equipment cluster residual life prediction method and terminal based on single-mounted life evolution |
CN110174261B (en) * | 2019-05-23 | 2020-09-08 | 太原科技大学 | Gear real-time residual life prediction method based on multi-degradation monitoring |
CN110245391B (en) * | 2019-05-28 | 2023-07-18 | 上海发电设备成套设计研究院有限责任公司 | Method for predicting service life based on hardness of artificial neural network |
CN110341591A (en) * | 2019-07-10 | 2019-10-18 | 太原科技大学 | It is a kind of for pure electric vehicle or the acoustic management system of hybrid vehicle active safety |
CN110543869A (en) * | 2019-09-10 | 2019-12-06 | 哈工大机器人(山东)智能装备研究院 | Ball screw service life prediction method and device, computer equipment and storage medium |
CN113092103B (en) * | 2021-04-08 | 2022-08-19 | 湖南交通工程学院 | Data-driven gear meshing stiffness real-time prediction device and prediction method |
CN113361189B (en) * | 2021-05-12 | 2022-04-19 | 电子科技大学 | Chip performance degradation trend prediction method based on multi-step robust prediction learning machine |
CN113704898B (en) * | 2021-07-19 | 2023-09-05 | 宁波东力传动设备有限公司 | System parameter decision method for evaluating nonlinear dynamics comprehensive performance of gear |
CN113658415B (en) * | 2021-07-30 | 2024-03-26 | 江苏湛德医疗用品有限公司 | Early warning method and system of intelligent gateway |
CN114021784A (en) * | 2021-10-22 | 2022-02-08 | 华能(浙江)能源开发有限公司清洁能源分公司 | Method and device for determining residual service life of equipment and electronic equipment |
CN114057053B (en) * | 2022-01-18 | 2022-04-26 | 杭州浅水数字技术有限公司 | Method for monitoring fatigue degree of component of special machine |
CN114896873B (en) * | 2022-04-26 | 2023-04-07 | 盐城工学院 | Traveling wave ultrasonic motor service life prediction method and system based on convolutional neural network |
WO2024021108A1 (en) * | 2022-07-29 | 2024-02-01 | Siemens Aktiengesellschaft | Method and device for predicting service life of rolling bearing and computer readable storage medium |
CN116257972B (en) * | 2022-11-29 | 2024-02-20 | 元始智能科技(南通)有限公司 | Equipment state evaluation method and system based on field self-adaption and federal learning |
CN115982896B (en) * | 2023-03-21 | 2023-06-30 | 山东金帝精密机械科技股份有限公司 | Bearing retainer service life detection method and device |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20040103027A (en) * | 2003-05-30 | 2004-12-08 | 국방과학연구소 | Method for estimating an inductance profile of srm using adaptive neuro-fuzzy inference system |
CN101793887A (en) * | 2010-01-09 | 2010-08-04 | 中国水产科学研究院黄海水产研究所 | Construction method of fuzzy neural network expert system for water quality assessment in turbot culture |
CN105046389A (en) * | 2015-02-13 | 2015-11-11 | 国家电网公司 | Intelligent risk assessment method for electric power security risk assessment, and system thereof |
CN105740625A (en) * | 2016-01-31 | 2016-07-06 | 太原科技大学 | Real time residual life prediction method of gear |
CN107370169A (en) * | 2017-05-12 | 2017-11-21 | 沈阳工业大学 | Extensive energy-accumulating power station peak regulation controller and method based on ANFIS short-term load forecastings |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120166363A1 (en) * | 2010-12-23 | 2012-06-28 | Hongbo He | Neural network fault detection system and associated methods |
-
2018
- 2018-04-08 CN CN201810304816.1A patent/CN108645615B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20040103027A (en) * | 2003-05-30 | 2004-12-08 | 국방과학연구소 | Method for estimating an inductance profile of srm using adaptive neuro-fuzzy inference system |
CN101793887A (en) * | 2010-01-09 | 2010-08-04 | 中国水产科学研究院黄海水产研究所 | Construction method of fuzzy neural network expert system for water quality assessment in turbot culture |
CN105046389A (en) * | 2015-02-13 | 2015-11-11 | 国家电网公司 | Intelligent risk assessment method for electric power security risk assessment, and system thereof |
CN105740625A (en) * | 2016-01-31 | 2016-07-06 | 太原科技大学 | Real time residual life prediction method of gear |
CN107370169A (en) * | 2017-05-12 | 2017-11-21 | 沈阳工业大学 | Extensive energy-accumulating power station peak regulation controller and method based on ANFIS short-term load forecastings |
Also Published As
Publication number | Publication date |
---|---|
CN108645615A (en) | 2018-10-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108645615B (en) | A kind of Adaptive Fuzzy Neural-network gear method for predicting residual useful life | |
Saon et al. | Predicting remaining useful life of rotating machinery based artificial neural network | |
Liu et al. | A multi-step predictor with a variable input pattern for system state forecasting | |
Yang et al. | Multi-step ahead direct prediction for the machine condition prognosis using regression trees and neuro-fuzzy systems | |
CN105740625B (en) | A kind of real-time method for predicting residual useful life of gear | |
Yan et al. | A dynamic multi-scale Markov model based methodology for remaining life prediction | |
Zhang et al. | Multivariable trend analysis using neural networks for intelligent diagnostics of rotating machinery | |
Van Tung et al. | Machine fault diagnosis and prognosis: The state of the art | |
CN109883691B (en) | Gear residual life prediction method integrating kernel estimation and random filtering | |
CN111783362B (en) | Method and system for determining residual service life of electric gate valve | |
CN110175369A (en) | A kind of gear method for predicting residual useful life based on two-dimensional convolution neural network | |
Wang et al. | An evolving neuro-fuzzy technique for system state forecasting | |
Salunkhe et al. | Prediction of Remaining Useful Life of mechanical components-a Review | |
Soualhi et al. | Long-term prediction of bearing condition by the neo-fuzzy neuron | |
CN109447305B (en) | Trend prediction method based on quantum weighted long-time and short-time memory neural network | |
Shao et al. | Fault prognosis and diagnosis of an automotive rear axle gear using a RBF-BP neural network | |
He et al. | Machine prognostics under varying operating conditions based on state-space and neural network modeling | |
Sreepradha et al. | Neural network model for condition monitoring of wear and film thickness in a gearbox | |
Feng et al. | Online residual useful life prediction of large-size slewing bearings—A data fusion method | |
Zhang et al. | A multi-fault modeling approach for fault diagnosis and failure prognosis of engineering systems | |
Sarwar et al. | Time series method for machine performance prediction using condition monitoring data | |
Magnus et al. | Vibration-based condition monitoring of heavy duty machine driveline parts: Torque converter, gearbox, axles and bearings | |
Yao et al. | A two-stage data quality improvement strategy for deep neural networks in fault severity estimation | |
CN113988118A (en) | Wind power equipment fault prediction method based on LSTM network | |
Gao et al. | A neural network approach to bearing health assessment |
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 |