CN103234749A - Automobile clutch control comfort evaluation method based on artificial neural network - Google Patents
Automobile clutch control comfort evaluation method based on artificial neural network Download PDFInfo
- Publication number
- CN103234749A CN103234749A CN2013101213195A CN201310121319A CN103234749A CN 103234749 A CN103234749 A CN 103234749A CN 2013101213195 A CN2013101213195 A CN 2013101213195A CN 201310121319 A CN201310121319 A CN 201310121319A CN 103234749 A CN103234749 A CN 103234749A
- Authority
- CN
- China
- Prior art keywords
- neural network
- value
- function
- input
- net
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Hydraulic Clutches, Magnetic Clutches, Fluid Clutches, And Fluid Joints (AREA)
Abstract
The invention relates to an automobile clutch control comfort evaluation method based on an artificial neural network. The method is characterized in that the computer simulation capability is utilized for setting up an error back propagation neural network model, acquired comfort subjective scoring of industry experts is mapped to each measurable objective parameter index via the neural network by constructing a test platform, and accordingly, quantized scoring of comfort is achieved.
Description
Technical field
The invention belongs to the mechanical test field, be specifically related to a kind of method that quantizes scoring based on the automobile clutch handling comfort of neural network.
Background technology
Nowadays automobile is to become the requisite vehicles in our life, and the thing followed then is a large amount of traffic hazard.Just imagine, for the public transport, the taxi driver that drive all day on the urban road of traffic congestion, to step on " thousands of " even " thousands of pin " clutch coupling in one day, if maneuvering performance is not good, increase the traffic hazard incidence thereby can accelerate tired driver.At home, most clutch coupling manufacturer all is the static separation characteristic that detects clutch coupling with clutch cover and plate assembly all-round property testing machine or stalling characteristic test machine." comfortableness " ignored in this test, is difficult point and how this fuzzy subjective feeling of clutch coupling comfortableness is quantized test.
By making up an automobile clutch Function detection platform, collect a large amount of objective parameter indexs, and by the expert each index comfortableness is experienced marking.Import database, accumulated a large amount of related datas.And neural network has strong non-linear mapping ability and good fault-tolerance, can well approach the time of day data of system, and its approximation accuracy height, training study speed are fast.Therefore adopt neural net method to set up automobile clutch comfortableness scoring model, handling comfort to automobile clutch is given a mark, can realize that subjective feeling quantizes, can improve the clutch coupling quality after feeding back to relevant industries enterprise, lifting enterprise class brings the driver simultaneously and better drives to experience.
Summary of the invention
The present invention proposes a kind of based on artificial neural network theories, combine error backpropagation algorithm and relevance theory, utilize computing machine that automobile clutch is controlled the method and apparatus that comfort level is carried out quantitatively evaluating, this invention comprises the real vehicle operation simulation system of a cover clutch coupling altogether, the testing apparatus of a cover automobile clutch serviceability and a cover objective evaluation system.Real vehicle operation simulation system comprises adjustable seat, clutch pedal, correlate meter, switch, monitor, clutch master cylinder, wheel cylinder (or power-assisted wheel cylinder).Testing apparatus comprises simulating flywheel, simulation clutch drived disk assy, simulation release bearing and pilot sleeve, pedal force and angular displacement sensor, separating force and separates stroke sensor, Separation sensor, hydraulic-pneumatic measuring instrument.Evaluation system comprises computing machine, data collecting card, database, analysis software, analyzes performance of clutch in order to gather the clutch coupling running parameter.The concrete steps of this method are:
Step (1) preference pattern variable.
Adopt neural network to set up automobile clutch comfortableness scoring model, for guaranteeing the validity based on test data, avoid the blindness of black case modeling, at first utilize Analysis on Mechanism and prior imformation, input variable and the output variable of choose reasonable scoring model.
Selecting the expert to experience marking based on Analysis on Mechanism is the output variable of neural network model, selects seven input variables that principal element is neural network model: 1. maximum pedal force; 2. minimum pedal force; 3. final pedal force; 4. semi-linkage point pedal force; 5. semi-linkage point damping; 6. minimum pedal force position; 7. separate stroke.
The collection of step (2) input/output variable is specially:
The manipulation environment that the test board of building is driven simulation, by related sensor, the objective value of seven indexs of survey, and charge to correspondence database.The experience impression of clutch coupling is given a mark to the comfortableness degree according to self by the multidigit industry specialists, for reducing owing to indivedual experts experience inaccurate error.The Grubbs check is carried out in expert's marking, picked marking data.Then remain the average of non-unusual scoring as the comfortableness score value of corresponding clutch coupling, charge to associated databases.
Wherein
Be non-unusual marking value, n is non-unusual marking data number,
Be the subjective marking value of the final expert of correspondence.
For reducing the complicacy of parameter, further improve neural network efficient.Seven objective input pointers and a subjective marking index being carried out principal component analysis (PCA), determine new major component, is the neural network input variable with new major component, and output variable is constant.
Step (3) data normalization is handled.
Output data in the training sample comprise seven, and it is bigger that the order of magnitude differs, and eat decimal for avoiding big number, and add rapid convergence, and data are carried out normalized, are converted into [0,1] interval value range
Wherein
Be the maximal value in the input pointer,
Be the minimum value in the input pointer,
Be the input data,
Be the numerical value after the normalization.
Step (4) is built the BP neural network framework.
Call matlab R2011 Neural Network Toolbox newff function and set up the BP neural network, NET=newff (PR, [
],
, BTF, BLF, PF); NET is the BP neural network framework, and PR is input matrix maximal value and scope of minimum value,
Be the neuronic number of i layer,
It is the transport function of i layer.
, N is the total number of plies of neural network.BTF represents the training function of neural network, and BLF is weights and bias, and PF is the network performance function.
Step (5) neural network training.Concrete grammar is:
I, initialization BP neural network, the value assignment of utilizing random function to produce is given weights and bias, calls init function initialization neural network.
II, neural metwork training upper limit number of times and target error are set.
III, training data is set is input matrix P, and it is matrix T that desired value is set, and the train function that calls in the matlab R2011 Neural Network Toolbox carries out the data training until convergence to BP neural network NET, and NET=train (NET, P, T).
The test of step (6) BP neural network.
The BP neural network of finishing training is tested, seven achievement datas are formed matrix P_test, directly call the sim function, (NET P_test), carries out emulation to test matrix to D=sim, wherein the objective function of D.
The grading of step (7) fuzzy c cluster.Concrete grammar is:
The comfortableness score value of different clutch couplinges to the test gained carries the fcm function by matlab, and (D c) carries out progressive operation to the objective function that simulates to [center, U, obj_fcn]=fcm.Wherein center is cluster centre; U is the degree of membership matrix; Obj_fcn is desired value; D is the objective function that simulates; C is batch total.Just tell the comfortableness degree grading of all kinds of clutch couplinges according to the score value of cluster centre.
The present invention utilizes the Computer Simulation ability, build an error back propagation (BP) neural network model, by making up test platform, the industry specialists that the collects subjective scoring situation to comfortableness is mapped in each the objective parameter index that can survey by neural network, realized the quantification scoring of comfortableness.
The beneficial effect of the inventive method:
1, the BP neural network has the ability of approaching any Nonlinear Mapping function in theory, therefore utilizes this method than traditional linear regression method scoring precision height.
2, avoided original rationalistic to the very complicated analysis of comfortableness based on the model application of BP neural network, simplified process.And can utilize the subjective marking of expert preferably, just not need to be given a mark by the expert again after setting up credible mapping relations by abundant data, practical.
3, increase the quantity of BP neural metwork training data, can better improve precision, obtain one more with the more high effect of precision.
Description of drawings
Fig. 1 is Analysis on Mechanism figure in the pilot control clutch coupling process.
Fig. 2 is the detection platform system layout.
Embodiment
Be example with a test clutch coupling for truck wherein, carry out the comfortableness marking modeling embodiment based on the BP neural network.
Step (1) preference pattern variable.
Analysis on Mechanism.Model is output as the expert to the score value of concrete clutch coupling.According to the manipulation impression of brainstrust to clutch coupling, to give a mark as " very light ", " gently ", " lighter ", " comfortable ", " heavier ", " weight ", " very heavy ", score value is between [0,10].By reference to the accompanying drawings 1, be that pedal force is enough little on the one hand, be pedal position, stroke are suitable on the other hand, trample conveniently.Handling accurately implication is that clutch coupling engages starting point and thoroughly the burble point position is clear, engages regulatory region and controls easily, reacts fast.The requirement of clutch coupling is also had: work will stablize, reliably, long service life; Joint is smooth-going, soft, makes the gentle start of automobile energy, does not impact and shake; Structure is simple, compact, and anufacturability is good, dismounting is easy to maintenance etc.Small-sized passenger car generally requires control at 100-130N, and Heavy Duty Commercial Car and offroad vehicle generally require control at 140-170N.Excessive or too small pedal force equally also can make the people produce discomfort.The too fast reduction of pedal force and too small paddy peak all can make the people produce the sensation of " stepping on sky " than (namely crossing peak value bumper step power reduces too much), increase the sense of discomfort of human body, the difficulty of " semi-linkage " control when strengthening the clutch coupling joint.Under most of situation, the frictional dissipation in the mechanical system is as far as possible little, but concerning the control system of automobile clutch, too small damping can increase driver's labour intensity on the contrary---especially when the road that blocks up travels; Excessive damping then can exert an influence to the accurate control of clutch coupling.Consider if use the left foot solenoidoperated cluthes, so pedal position should be according to human body center line certain distance that moves to left; Maximum pedal travel is subjected to clutch release travel, separating force and ratio of gear three's restriction, the too small meeting of range clutch coupling can not be separated or pedal force excessive, the excessive human comfort that then influences of range; Step height also has certain requirement---and should satisfy the requirement of range, it is too high also will to make human body lift pin, can make human body sensory comfortable.Determine be input as 7 of neural network model based on above-mentioned Analysis on Mechanism; Maximum pedal force, minimum pedal force, final pedal force, semi-linkage point pedal force, the damping of semi-linkage point, minimum pedal force position and separate stroke.
The collection of step (2) input/output variable.
20 samples with certain truck clutch coupling are example, by detection system as shown in Figure 2, collect the objective value of seven input pointers, and charge in the objective detection database.
Make that input matrix is P, then concrete arrangement mode is:
Wherein, represent j index of i sample, seven desired values once are respectively: maximum pedal force, minimum pedal force, final pedal force, semi-linkage point pedal force, the damping of semi-linkage point, minimum pedal force position and separate stroke.
Wherein, Gn and Gn ' are respectively critical value;
Judge score value for each expert before not checking;
Be the scoring average before not checking, t is for calculating the standard deviation of back gained.Concrete checking procedure is:
The Grubbs check table
B, carry out data qualifier and judge.
Make that output matrix is Q, then concrete arrangement mode is:
Step (3) data normalization is handled.
Since output matrix have only index of score value, so only need utilize formula that input matrix is carried out normalized
For each row is according to normalized among the input matrix P, the input matrix after obtaining handling is P '.
Step (4) makes up the BP neural network.
Build BP neural network model framework, call the newff function in the matlab function library.
NET=newff(maxmin,[1000,7],{'tansig','purelin'},'traingd')
Wherein, maxmin represents to import the maximin of input vector; [1000,7] expression ground floor has 1000 neurons, and the second layer has seven neurons; Tansig is the input layer transport function; Pirelin is the output layer transport function; Traingd is the training function of algorithm.
Step (5) training BP neural network.
A. initialization network
Net.initfcn is used for determining the initialization function of whole network; The parameter net.layer{i}.initfcn initialization function that decides each layer; The initweb function according to the initialization function of each layer oneself (net.inputweights{i, j}.initfcn) initializes weights matrix and biasing arrange initialization rands immediately usually, and concrete grammar is as follows:
net.layers{1}.initfcn=’initwb’;
net.inputweights{1,1}.initfcn=’rands’;
net.layerweights{2,1}.initfcn=’rands’;
net.biases{1,1}.initfcn=’rands’;
net.biases{2,1}.initfcn=’rands’;
net=init(net);
Net.IW{1,1} are that input layer is to the weight matrix of hidden layer;
Net.LW (2,1) is that hidden layer is to the weight matrix between output layer;
Net.b{1,1} are the threshold values vector of hidden layer;
Net.b{2,1} are the threshold values of output node;
B., network training number of times, training objective error and display frequency are set
net.trainparam.epochs=2000;
net.trainparam.goal=0.00025;
net.trainparam.show=25;
It was 2000 steps that the network training number of times is set; Training objective error 0.00025; Show that the training step number is 25.
C. utilize input matrix P and objective matrix Q, by calling the train function, (met, P ' Q) carry out the training of comfortableness judge network model until convergence to net=train.
Step (6) network test.
To form according to the input matrix form of step (1) for seven achievement datas of test and be used for clutch coupling comfortableness testing evaluation matrix P_test, and carry out normalized according to step (3) again, the test matrix after the normalization is P_test '.Call matlab sim function, bring network simulation into.Code is:
D=sim (net, P_test '); Every value is the computing machine marking value after the training in the D matrix.
The grading of step (7) fuzzy c cluster.
The comfortableness scoring that records is carried out the comfort level grading according to matlab fcm function, and code is:
[center,U,fcn]=fcm(D,3)。Center is cluster centre; U is the degree of membership matrix; Fcn is desired value; 3 is grade class number.Be respectively: excellent, good, poor.Obtain comfortableness marking and the concrete grade of concrete clutch coupling by aforesaid operations.
Claims (1)
1. control the comfort level evaluation method based on the automobile clutch of artificial neural network, it is characterized in that this method may further comprise the steps:
Step (1) preference pattern variable is specially:
Utilize Analysis on Mechanism and prior imformation, input variable and the output variable of choose reasonable scoring model;
Selecting the expert to experience marking based on Analysis on Mechanism is the output variable of neural network model, selects seven input variables that principal element is neural network model: maximum pedal force; Minimum pedal force; Final pedal force; 4. semi-linkage point pedal force; 5. semi-linkage point damping; 6. minimum pedal force position; 7. separate stroke;
The collection of step (2) input/output variable is specially:
The manipulation environment that the test board of building is driven simulation by related sensor, records the objective value of seven indexs, and charges to correspondence database; The experience impression of clutch coupling is given a mark to the comfortableness degree according to self by the multidigit industry specialists, for reducing owing to indivedual experts experience inaccurate error; The Grubbs check is carried out in expert's marking, picked marking data; Then remain the average of non-unusual scoring
As the comfortableness score value of corresponding clutch coupling, charge to associated databases;
Wherein
Be non-unusual marking value, n is non-unusual marking data number,
Be the subjective marking value of the final expert of correspondence;
For reducing the complicacy of parameter, further improve neural network efficient; Seven objective input pointers and a subjective marking index being carried out principal component analysis (PCA), determine new major component, is the neural network input variable with new major component, and output variable is constant;
Step (3) data normalization is handled, and is specially:
Output data in the training sample comprise seven, and it is bigger that the order of magnitude differs, and eat decimal for avoiding big number, and add rapid convergence, and data are carried out normalized, are converted into [0,1] interval value range
Wherein
Be the maximal value in the input pointer,
Be the minimum value in the input pointer,
Be the input data,
Be the numerical value after the normalization;
Step (4) is built the BP neural network framework, is specially:
Call matlab R2011 Neural Network Toolbox newff function and set up the BP neural network, NET=newff (PR, [
],
, BTF, BLF, PF); NET is the BP neural network framework, and PR is input matrix maximal value and scope of minimum value,
Be the neuronic number of i layer,
It is the transport function of i layer;
, N is the total number of plies of neural network; BTF represents the training function of neural network, and BLF is weights and bias, and PF is the network performance function;
Step (5) neural network training; Be specially:
I, initialization BP neural network, the value assignment of utilizing random function to produce is given weights and bias, calls init function initialization neural network;
II, neural metwork training upper limit number of times and target error are set;
III, training data is set is input matrix P, and it is matrix T that desired value is set, and the train function that calls in the matlab R2011 Neural Network Toolbox carries out the data training until convergence to BP neural network NET, and NET=train (NET, P, T);
The test of step (6) BP neural network is specially:
The BP neural network of finishing training is tested, seven achievement datas are formed matrix P_test, directly call the sim function, (NET P_test), carries out emulation to test matrix to D=sim, wherein the objective function of D;
The grading of step (7) fuzzy c cluster; Be specially:
The comfortableness score value of different clutch couplinges to the test gained carries the fcm function by matlab, and (D c) carries out progressive operation to the objective function that simulates to [center, U, obj_fcn]=fcm; Wherein center is cluster centre; U is the degree of membership matrix; Obj_fcn is desired value; D is the objective function that simulates; C is batch total; Just tell the comfortableness degree grading of all kinds of clutch couplinges according to the score value of cluster centre.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013101213195A CN103234749A (en) | 2013-04-09 | 2013-04-09 | Automobile clutch control comfort evaluation method based on artificial neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013101213195A CN103234749A (en) | 2013-04-09 | 2013-04-09 | Automobile clutch control comfort evaluation method based on artificial neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103234749A true CN103234749A (en) | 2013-08-07 |
Family
ID=48882800
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2013101213195A Pending CN103234749A (en) | 2013-04-09 | 2013-04-09 | Automobile clutch control comfort evaluation method based on artificial neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103234749A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104680335A (en) * | 2015-03-24 | 2015-06-03 | 北京闪信鼎中技术有限公司 | Evaluation method for new energy source automobile |
CN106202898A (en) * | 2016-07-05 | 2016-12-07 | 同济大学 | A kind of evaluation methodology based on analytic hierarchy process (AHP) brake pedal feel |
CN108734386A (en) * | 2017-04-20 | 2018-11-02 | 发那科株式会社 | Machined surface quality evaluation device |
CN109541943A (en) * | 2018-12-07 | 2019-03-29 | 西南交通大学 | A kind of tramcar on-line optimizing and controlling method |
CN110069894A (en) * | 2019-05-09 | 2019-07-30 | 同济大学 | A kind of objective mapping test method for intelligent automobile traffic coordinating |
CN112711191A (en) * | 2020-12-28 | 2021-04-27 | 北京理工大学 | Method for improving speed regulation accuracy of hydro-viscous speed regulation clutch based on neural network training |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060155734A1 (en) * | 2005-01-07 | 2006-07-13 | Grimes Michael R | Apparatus and methods for evaluating a dynamic system |
CN101118620A (en) * | 2007-09-18 | 2008-02-06 | 吉林大学 | Vehicle gear shifting quality evaluation method based on nerval net |
CN101793602A (en) * | 2010-03-31 | 2010-08-04 | 浙江铁流离合器股份有限公司 | Subjective assessment test bench and test method for operating performance of clutches |
CN102393875A (en) * | 2011-07-01 | 2012-03-28 | 合肥工业大学 | Operating force comfortableness grade classification method of operating device |
CN102494889A (en) * | 2011-11-30 | 2012-06-13 | 杭州电子科技大学 | Computer evaluation method for automobile clutch comfort |
-
2013
- 2013-04-09 CN CN2013101213195A patent/CN103234749A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060155734A1 (en) * | 2005-01-07 | 2006-07-13 | Grimes Michael R | Apparatus and methods for evaluating a dynamic system |
CN101118620A (en) * | 2007-09-18 | 2008-02-06 | 吉林大学 | Vehicle gear shifting quality evaluation method based on nerval net |
CN101793602A (en) * | 2010-03-31 | 2010-08-04 | 浙江铁流离合器股份有限公司 | Subjective assessment test bench and test method for operating performance of clutches |
CN102393875A (en) * | 2011-07-01 | 2012-03-28 | 合肥工业大学 | Operating force comfortableness grade classification method of operating device |
CN102494889A (en) * | 2011-11-30 | 2012-06-13 | 杭州电子科技大学 | Computer evaluation method for automobile clutch comfort |
Non-Patent Citations (1)
Title |
---|
扈静: "基于主成分分析的汽车操纵力舒适性分析与评价", 《中国机械工程》, vol. 2, no. 20, 31 October 2011 (2011-10-31), pages 2456 - 2459 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104680335A (en) * | 2015-03-24 | 2015-06-03 | 北京闪信鼎中技术有限公司 | Evaluation method for new energy source automobile |
CN106202898A (en) * | 2016-07-05 | 2016-12-07 | 同济大学 | A kind of evaluation methodology based on analytic hierarchy process (AHP) brake pedal feel |
CN108734386A (en) * | 2017-04-20 | 2018-11-02 | 发那科株式会社 | Machined surface quality evaluation device |
CN109541943A (en) * | 2018-12-07 | 2019-03-29 | 西南交通大学 | A kind of tramcar on-line optimizing and controlling method |
CN110069894A (en) * | 2019-05-09 | 2019-07-30 | 同济大学 | A kind of objective mapping test method for intelligent automobile traffic coordinating |
CN112711191A (en) * | 2020-12-28 | 2021-04-27 | 北京理工大学 | Method for improving speed regulation accuracy of hydro-viscous speed regulation clutch based on neural network training |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103234749A (en) | Automobile clutch control comfort evaluation method based on artificial neural network | |
CN108407816B (en) | Method and system for evaluating driving of automobile driver | |
CN101793602B (en) | Subjective assessment test bench and test method for operating performance of clutches | |
US20220196459A1 (en) | Real-time vehicle overload detection method based on convolutional neural network | |
CN111841012A (en) | Automatic driving simulation system and test resource library construction method thereof | |
CN104573343B (en) | A kind of Asphalt Pavement Surface Evenness comfortableness field evaluation method and method of tire | |
CN105547711A (en) | Motorcycle acceleration fatigue durability test method based on road simulation stand | |
CN109493566A (en) | A kind of fatigue driving prediction technique based on GPS data | |
CN106682768A (en) | Prediction method, system, terminal and server of test score | |
CN116029555B (en) | Bridge risk identification early warning system based on lightweight neural network and application method | |
Rowley et al. | Examining the driverless future: An analysis of human-caused vehicle accidents and development of an autonomous vehicle communication testbed | |
CN112109715B (en) | Method, device, medium and system for generating vehicle power output strategy | |
CN114323853A (en) | Intelligent monitoring method for asphalt pavement compactness based on decision tree model | |
CN108388730A (en) | The method of vehicle development | |
CN110704789B (en) | Population dynamic measurement and calculation method and system based on 'urban superconcephalon' computing platform | |
CN111806443A (en) | Pure electric heavy truck unmanned energy consumption optimization method | |
CN115422707A (en) | Scene-driven driver danger prediction capability test and evaluation method | |
KR101816441B1 (en) | Device and method for evaluating driving sensitivity of vehicle | |
Elseifi et al. | A mechanistic approach to utilize traffic speed deflectometer (TSD) measurements into backcalculation analysis | |
Şener | Finite element based vehicle component fatigue life assessment according to a customer usage profile | |
CN105496369A (en) | Angry driving state detection method | |
Ab Aziz et al. | Monitoring of hand-arm vibration | |
Nair et al. | Realistic pavement serviceability equations using the 690D surface dynamics profilometer | |
Abdullah et al. | Design, Analysis and Fabrication of Fixed-base Driving Simulator Frame | |
KR101349640B1 (en) | Evaluation apparatus and process for driver's cognitive performance on variable message sign |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130807 |