Research ArticleResearch on Application of Regression LeastSquares Support Vector Machine on PerformancePrediction of Hydraulic Excavator
Zhan-bo Chen12
1 Zhongnan University of Economics and Law Wuhan 430074 China2Hubei University of Economics Wuhan 430205 China
Correspondence should be addressed to Zhan-bo Chen chenzhanbo2014163com
Received 15 July 2014 Revised 11 October 2014 Accepted 22 October 2014 Published 11 November 2014
Academic Editor Onur Toker
Copyright copy 2014 Zhan-bo Chen This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
In order to improve the performance prediction accuracy of hydraulic excavator the regression least squares support vectormachineis applied First the mathematical model of the regression least squares support vector machine is studied and then the algorithmof the regression least squares support vector machine is designed Finally the performance prediction simulation of hydraulicexcavator based on regression least squares support vector machine is carried out and simulation results show that this methodcan predict the performance changing rules of hydraulic excavator correctly
1 Introduction
The hydraulic excavator belongs to the construction machin-ery which has been applied in many fields successfully suchas transportation industry mining industry constructionindustry and hydraulic engineeringThe hydraulic excavatoris made up of three parts which are working equipmentupper turntable and traveling gear Normally working con-ditions of the hydraulic excavator are bad and the loadsbeing applied to the hydraulic excavator are big thereforethe engine will deviate from operating mode with low fuelconsumption and then hydraulic excavator will exhibit poorperformance In addition the energy consumption of thehydraulic pressure system is big which can lead to thebig energy wasting Therefore it is necessary to predict theperformance of the hydraulic excavator correctly and thenthe working efficiency of the hydraulic excavator can beimproved [1]
The performance predicting procession is nonlinearwhich is affected by many uncertain factors therefore aneffective predicting technology should be chosen At presentthere are many performances prediction methods such asartificial neural network technology grey prediction tech-nology and extension technology [2] However the current
prediction technologies have some disadvantages the pre-dicting precision is low the predicting efficiency is low andthe operation is difficult [2]
Generally the support vector machine has the nonlinearand uncertain characteristics In recent years the supportvector machine was established by Vapnik which has beenconcerned by many scientists and the support vector hasstrong learning ability Some scientists have improved thesupport vector machine The least squares support vectormachine is put forward by SukKensThe least squares supportmachine introduces the least squares system into the supportvector while the traditional support machine applies the twoplanning methods to deal with function estimation problemtherefore this method has higher predicting ability than thetraditional support machine which can be applied in theperformance prediction of the hydraulic excavator
2 Mathematical Model of Least SquaresSupport Vector Machine
The support vector machine is a machine learning technol-ogy which is better than artificial neural network technol-ogy grey prediction technology and extension technologyin predicting ability The support vector machine applies
Hindawi Publishing CorporationJournal of Control Science and EngineeringVolume 2014 Article ID 686130 4 pageshttpdxdoiorg1011552014686130
2 Journal of Control Science and Engineering
to the performance prediction of small samples and thepredicting reliability is good at the same time it has goodsuperior ability and then the high predicting precision canbe obtained Therefore the good predicting effect can beobtained based on least squares support vector machine forthe hydraulic excavator [3]
The training samples (119909119894 119910119894) (119909119894 119910119894isin 119877 119894 119895 = 1 2
119899) are known and the least squares support vector machinemainly applies a nonlinear mapping function to make thedata transfer to a high dimensional space and then bemappedback to the original space to achieve the linear regression ofinput space the linear regression model is shown as follows
119891 (119909) = [ 120601 (119909)] + 119887 (1)
where 120596 denotes the weighted vector 120601(119909) denotes themapping function and 119887 denotes the threshold value
The regression least squares support vector machineapplies two-time penalty function to transfer the regressionproblem into two-time optimization problem and the corre-sponding objective function is expressed as follows
where119877ref[119891] denotes the structural risk119877emp[119891] denotes theempiric risk 1205952 denotes the confidence risk 119862(sdot) denotesthe loss function 120578 denotes regularization factor 119890
119897= 119891(119909
119897)minus
119910119897 and 119899 denotes the sample sizeThe 120576 insensitive loss function is put forward by Vapnik
which has an insensitive zone that offers any loss value 120576 zoneThe information of sample points in 120576 zone cannot appearin regression function and then the regression function issparse and simple
The optimal function is constructed based on the idea of120576 insensitive loss function which is expressed as follows [4]
min120596119887120591120591
lowast
119869 =1
2120596119879120596 + 119862
119899
sum119894=1
(120591119894+ 120591lowast
119894) (3)
st
119910119894minus 120596119879120601 (119909119894) minus 119887 le 120583 + 120591
119894
120596119879120601 (119909119894) + 119887 minus 119910
119894le 120583 + 120591
lowast
119894
120591119894 120591lowast
119894ge 0
(4)
where 120591119894and 120591lowast119894denote the relaxation factors 119862 denotes the
penalty function 119862 gt 0 and 120583 denotes the precisionThe core function is introduced and the dual function of
expression (4) can be expressed as follows [5]
max120573120573lowast
119869 = minus1
2
119899
sum119894119895
(120573119894minus 120573lowast
119894) (120573119895minus 120573lowast
119895) sdot 119870 (119909
119894 119909119895)
minus 120583
119899
sum119894=1
(120573119894+ 120573lowast
119894) +
119899
sum119894
119910119894(120573119894minus 120573lowast
119894)
st119899
sum119894=1
(120573119894minus 120573lowast
119894) 120573
119894 120573lowast
119894isin (0 119862]
(5)
where 120573119894and 120573
lowast
119894are also Lagrange operators and 119870(119909
119894 119909119895)
denotes the core function
The estimating function of the least squares supportvector machine can be expressed as follows [6 7]
119891 (119909) =
119899
sum119894=1
(120573119894minus 120573lowast
119894)119870 (119909 119909
119896) + 119887 (6)
The parameters optimization of the least squares supportvectormachine is very importantThe core function generallyapplies the radial function There is an unknown factor 120574and improving the value of 120574 can improve the convergencerate of the algorithm In addition there is another parameterwhich is the parameter of the least squares support vectormachine 120578 the parameters Gamma and 120578 can decide thelearning ability of the least squares support vector machinetogether
3 Algorithm of the Least Squares SupportVector Machine
The algorithm procedure of the least squares support vectormachine is listed as follows
Step 1 The range of the two parameters 120574 and 120578 is confirmedbased on basic principles of the least squares support vectormachine and the empirical range of the two parameters islisted as follows 120574 isin [001 01] 120578 isin [002 10000]
Step 2 The values of the two parameters 120574 and 120578 areconfirmed in range then the two-dimensional plane (120574
119894 120578119894)
119894 = 1 2 119898 1 2 119899 can be constructed The value ofGamma can be chosen according to the real situation oftraining samples and relating experiences
Step 3 The pairs of parameters (120574119894 120578119894) in different plane grid
nodes are input into the least squares support vectormachinethe corresponding training is carried out based on trainingsamples and then the training error is output (120574
119894 120578119894) with
least error is used as the most optimal results [8]
Step 4 When the training precision of algorithm cannotsatisfy the real requirement the optimal parameter is usedas center to construct the new plane grids and confirm thenear parameter then the new training is carried out againthen the precision of the algorithm can be improved and theprocedures mentioned above can be carried out repeatedlythen the multilayers parameter optimization plane networkcan be formed finally the optimal parameters of the leastsquares support vector machine can be obtained then theideal training precision can be obtained [9]
4 Prediction Simulation of HydraulicExcavator Based on Least Squares SupportVector Machine
Thepredictionmodel of the hydraulic excavator performanceis constructed based on the following basic steps
(1) The characteristics and performance indexes of thehydraulic excavator are confirmed And the learning samplesof the regression least squares support vector machine areobtained
Journal of Control Science and Engineering 3
Table 1 Data sample of performance prediction for the hydraulicexcavator
(2) The core function is chosen the mapping relationbetween the input and output parameters is obtained throughsample learning
(3)Thenew parameters are input into the regression leastsquares support vectormachine to carry out prediction of thehydraulic excavator performance
(4) The new learning samples of the hydraulic excavatorsystem are added to the regression least squares supportvector machine then the prediction ability of model can beimproved
The prediction effect of the least squares support vectormachine can be judged by the root-mean-square error whichis expressed as follows [10]
119864 = radic1
119899
119899
sum119894=1
(119910119894minus 119910119894
119910119894
)
2
times 100 (7)
Through field test for the hydraulic excavator 100 groupsamples are obtained and the former 60 groups of samplesare used as performance predicting samples of the hydraulicexcavator and the other 40 groups of samples are used asverification samples of prediction model the performanceprediction simulation programmer of the hydraulic excavatoris compiled by MATLAB software
The input parameters of performance prediction modelfor the hydraulic excavator are listed as follows rotating speedof engine 119899
119868119862119864 the rotating speed of motor 119899
119864119872 rotating
speed of the hydraulic pump 119899119875 the torque of the engine119879
119868119862119864
the torque of the motor 119879119864119872
and the torque of the hydraulicpump 119879
119875 the output parameters of the model are output
power of the engine 119873119868119862119864
and the pilot controlling pressureof hydraulic control valve119875 respectively the part samples areshown in Table 1
The performance predicting curve and actual measure-ment curve of output power for the engine are shown inFigure 1 As seen from Figure 1 the output power of theengine changes in the moment with time and output power
75
45
15
10 20 30 4000
60
30
Time (s)
Measured valuePrediction value
Out
put p
ower
of e
ngin
e (kW
)
Figure 1 Performance predicting curve and measured curve ofoutput power for engine
Table 2 Output power prediction results of the hydraulic excavatorbased on different prediction model
changes irregularly the predicting results agree with theactual measurements and these results show that the leastsquares support vector machine based on regression hasbetter prediction ability
During the working process of the hydraulic excavatorthe input parameter can be regulated based on the predictionresults of the least squares support vector machine based onregression then the hydraulic excavator can work steadilyand it can be in the optimal performance point the workingefficiency of the hydraulic excavator can be improved accord-ingly
The output power prediction results of the hydraulicexcavator from the regression least square support vectormachine traditional support vector machine and BP neutralnetwork are compared which are shown in Table 2
As seen from Table 2 the prediction results from theregression least support vectormachine are closer to themea-sured value than those from tradition support vectormachineand BP neutral network The regression least support vectormachine can obtain best performance prediction results ofthe hydraulic excavator
4 Journal of Control Science and Engineering
5 Conclusions
The least squares support vectormachine based on regressionis applied in the performance prediction of the hydraulicexcavator the model of the least squares support vectormachine based on regression is established the correspond-ing algorithm procedure is designed and the performanceprediction model of the hydraulic excavator is establishedThe performance prediction simulation is carried out andresults show that the regression least squares support vectormachine has higher prediction precision which can predictthe changing rules of the performance for the hydraulicexcavator and improve theworking efficiency which haswideapplication space
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] MHagaWHiroshi andK Fujishima ldquoDigging control systemfor hydraulic excavatorrdquo Mechatronics vol 11 no 6 pp 665ndash676 2001
[2] Z-Y Jia J-W Ma F-J Wang andW Liu ldquoHybrid of simulatedannealing and SVM for hydraulic valve characteristics predic-tionrdquo Expert Systems with Applications vol 38 no 7 pp 8030ndash8036 2011
[3] Q P Ha Q H Nguyen D C Rye and H F Durrant-Whyte ldquoImpedance control of a hydraulically actuated roboticexcavatorrdquoAutomation in construction vol 9 no 5 pp 421ndash4352000
[4] Y Hong and C W W Ng ldquoBase stability of multi-proppedexcavations in soft clay subjected to hydraulic upliftrdquo CanadianGeotechnical Journal vol 50 no 2 pp 153ndash164 2013
[5] J Jiang C Song and L Bao ldquoForwardGene selection algorithmbased on least squares support vector machinerdquo Journal ofBionanoscience vol 7 no 3 pp 307ndash312 2013
[6] M-Y Ye and X-DWang ldquoChaotic time series prediction usingleast squares support vector machinesrdquo Chinese Physics vol 13no 4 pp 454ndash458 2004
[7] S Swaddiwudhipong K K Tho Z S Liu J Hua and N S BOoi ldquoMaterial characterization via least squares support vectormachinesrdquo Modelling and Simulation in Materials Science andEngineering vol 13 no 6 pp 993ndash1004 2005
[8] Y Wan ldquoPump performance analysis based on least squaressupport vector machinerdquo Transactions of the Chinese Society ofAgricultural Engineering vol 25 no 8 pp 115ndash118 2009
[9] X Wang X Tian and Y Cheng ldquoValue approximation withleast squares support vector machine in reinforcement learningsystemrdquo Journal of Computational and Theoretical Nanosciencevol 4 no 7-8 pp 1290ndash1294 2007
[10] K Huang H JWang H R Xu J PWang and Y B Ying ldquoNIRspectroscopy based on least square support vector machines forquality prediction of tomato juicerdquo Spectroscopy and SpectralAnalysis vol 29 no 4 pp 931ndash934 2009
to the performance prediction of small samples and thepredicting reliability is good at the same time it has goodsuperior ability and then the high predicting precision canbe obtained Therefore the good predicting effect can beobtained based on least squares support vector machine forthe hydraulic excavator [3]
The training samples (119909119894 119910119894) (119909119894 119910119894isin 119877 119894 119895 = 1 2
119899) are known and the least squares support vector machinemainly applies a nonlinear mapping function to make thedata transfer to a high dimensional space and then bemappedback to the original space to achieve the linear regression ofinput space the linear regression model is shown as follows
119891 (119909) = [ 120601 (119909)] + 119887 (1)
where 120596 denotes the weighted vector 120601(119909) denotes themapping function and 119887 denotes the threshold value
The regression least squares support vector machineapplies two-time penalty function to transfer the regressionproblem into two-time optimization problem and the corre-sponding objective function is expressed as follows
where119877ref[119891] denotes the structural risk119877emp[119891] denotes theempiric risk 1205952 denotes the confidence risk 119862(sdot) denotesthe loss function 120578 denotes regularization factor 119890
119897= 119891(119909
119897)minus
119910119897 and 119899 denotes the sample sizeThe 120576 insensitive loss function is put forward by Vapnik
which has an insensitive zone that offers any loss value 120576 zoneThe information of sample points in 120576 zone cannot appearin regression function and then the regression function issparse and simple
The optimal function is constructed based on the idea of120576 insensitive loss function which is expressed as follows [4]
min120596119887120591120591
lowast
119869 =1
2120596119879120596 + 119862
119899
sum119894=1
(120591119894+ 120591lowast
119894) (3)
st
119910119894minus 120596119879120601 (119909119894) minus 119887 le 120583 + 120591
119894
120596119879120601 (119909119894) + 119887 minus 119910
119894le 120583 + 120591
lowast
119894
120591119894 120591lowast
119894ge 0
(4)
where 120591119894and 120591lowast119894denote the relaxation factors 119862 denotes the
penalty function 119862 gt 0 and 120583 denotes the precisionThe core function is introduced and the dual function of
expression (4) can be expressed as follows [5]
max120573120573lowast
119869 = minus1
2
119899
sum119894119895
(120573119894minus 120573lowast
119894) (120573119895minus 120573lowast
119895) sdot 119870 (119909
119894 119909119895)
minus 120583
119899
sum119894=1
(120573119894+ 120573lowast
119894) +
119899
sum119894
119910119894(120573119894minus 120573lowast
119894)
st119899
sum119894=1
(120573119894minus 120573lowast
119894) 120573
119894 120573lowast
119894isin (0 119862]
(5)
where 120573119894and 120573
lowast
119894are also Lagrange operators and 119870(119909
119894 119909119895)
denotes the core function
The estimating function of the least squares supportvector machine can be expressed as follows [6 7]
119891 (119909) =
119899
sum119894=1
(120573119894minus 120573lowast
119894)119870 (119909 119909
119896) + 119887 (6)
The parameters optimization of the least squares supportvectormachine is very importantThe core function generallyapplies the radial function There is an unknown factor 120574and improving the value of 120574 can improve the convergencerate of the algorithm In addition there is another parameterwhich is the parameter of the least squares support vectormachine 120578 the parameters Gamma and 120578 can decide thelearning ability of the least squares support vector machinetogether
3 Algorithm of the Least Squares SupportVector Machine
The algorithm procedure of the least squares support vectormachine is listed as follows
Step 1 The range of the two parameters 120574 and 120578 is confirmedbased on basic principles of the least squares support vectormachine and the empirical range of the two parameters islisted as follows 120574 isin [001 01] 120578 isin [002 10000]
Step 2 The values of the two parameters 120574 and 120578 areconfirmed in range then the two-dimensional plane (120574
119894 120578119894)
119894 = 1 2 119898 1 2 119899 can be constructed The value ofGamma can be chosen according to the real situation oftraining samples and relating experiences
Step 3 The pairs of parameters (120574119894 120578119894) in different plane grid
nodes are input into the least squares support vectormachinethe corresponding training is carried out based on trainingsamples and then the training error is output (120574
119894 120578119894) with
least error is used as the most optimal results [8]
Step 4 When the training precision of algorithm cannotsatisfy the real requirement the optimal parameter is usedas center to construct the new plane grids and confirm thenear parameter then the new training is carried out againthen the precision of the algorithm can be improved and theprocedures mentioned above can be carried out repeatedlythen the multilayers parameter optimization plane networkcan be formed finally the optimal parameters of the leastsquares support vector machine can be obtained then theideal training precision can be obtained [9]
4 Prediction Simulation of HydraulicExcavator Based on Least Squares SupportVector Machine
Thepredictionmodel of the hydraulic excavator performanceis constructed based on the following basic steps
(1) The characteristics and performance indexes of thehydraulic excavator are confirmed And the learning samplesof the regression least squares support vector machine areobtained
Journal of Control Science and Engineering 3
Table 1 Data sample of performance prediction for the hydraulicexcavator
(2) The core function is chosen the mapping relationbetween the input and output parameters is obtained throughsample learning
(3)Thenew parameters are input into the regression leastsquares support vectormachine to carry out prediction of thehydraulic excavator performance
(4) The new learning samples of the hydraulic excavatorsystem are added to the regression least squares supportvector machine then the prediction ability of model can beimproved
The prediction effect of the least squares support vectormachine can be judged by the root-mean-square error whichis expressed as follows [10]
119864 = radic1
119899
119899
sum119894=1
(119910119894minus 119910119894
119910119894
)
2
times 100 (7)
Through field test for the hydraulic excavator 100 groupsamples are obtained and the former 60 groups of samplesare used as performance predicting samples of the hydraulicexcavator and the other 40 groups of samples are used asverification samples of prediction model the performanceprediction simulation programmer of the hydraulic excavatoris compiled by MATLAB software
The input parameters of performance prediction modelfor the hydraulic excavator are listed as follows rotating speedof engine 119899
119868119862119864 the rotating speed of motor 119899
119864119872 rotating
speed of the hydraulic pump 119899119875 the torque of the engine119879
119868119862119864
the torque of the motor 119879119864119872
and the torque of the hydraulicpump 119879
119875 the output parameters of the model are output
power of the engine 119873119868119862119864
and the pilot controlling pressureof hydraulic control valve119875 respectively the part samples areshown in Table 1
The performance predicting curve and actual measure-ment curve of output power for the engine are shown inFigure 1 As seen from Figure 1 the output power of theengine changes in the moment with time and output power
75
45
15
10 20 30 4000
60
30
Time (s)
Measured valuePrediction value
Out
put p
ower
of e
ngin
e (kW
)
Figure 1 Performance predicting curve and measured curve ofoutput power for engine
Table 2 Output power prediction results of the hydraulic excavatorbased on different prediction model
changes irregularly the predicting results agree with theactual measurements and these results show that the leastsquares support vector machine based on regression hasbetter prediction ability
During the working process of the hydraulic excavatorthe input parameter can be regulated based on the predictionresults of the least squares support vector machine based onregression then the hydraulic excavator can work steadilyand it can be in the optimal performance point the workingefficiency of the hydraulic excavator can be improved accord-ingly
The output power prediction results of the hydraulicexcavator from the regression least square support vectormachine traditional support vector machine and BP neutralnetwork are compared which are shown in Table 2
As seen from Table 2 the prediction results from theregression least support vectormachine are closer to themea-sured value than those from tradition support vectormachineand BP neutral network The regression least support vectormachine can obtain best performance prediction results ofthe hydraulic excavator
4 Journal of Control Science and Engineering
5 Conclusions
The least squares support vectormachine based on regressionis applied in the performance prediction of the hydraulicexcavator the model of the least squares support vectormachine based on regression is established the correspond-ing algorithm procedure is designed and the performanceprediction model of the hydraulic excavator is establishedThe performance prediction simulation is carried out andresults show that the regression least squares support vectormachine has higher prediction precision which can predictthe changing rules of the performance for the hydraulicexcavator and improve theworking efficiency which haswideapplication space
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] MHagaWHiroshi andK Fujishima ldquoDigging control systemfor hydraulic excavatorrdquo Mechatronics vol 11 no 6 pp 665ndash676 2001
[2] Z-Y Jia J-W Ma F-J Wang andW Liu ldquoHybrid of simulatedannealing and SVM for hydraulic valve characteristics predic-tionrdquo Expert Systems with Applications vol 38 no 7 pp 8030ndash8036 2011
[3] Q P Ha Q H Nguyen D C Rye and H F Durrant-Whyte ldquoImpedance control of a hydraulically actuated roboticexcavatorrdquoAutomation in construction vol 9 no 5 pp 421ndash4352000
[4] Y Hong and C W W Ng ldquoBase stability of multi-proppedexcavations in soft clay subjected to hydraulic upliftrdquo CanadianGeotechnical Journal vol 50 no 2 pp 153ndash164 2013
[5] J Jiang C Song and L Bao ldquoForwardGene selection algorithmbased on least squares support vector machinerdquo Journal ofBionanoscience vol 7 no 3 pp 307ndash312 2013
[6] M-Y Ye and X-DWang ldquoChaotic time series prediction usingleast squares support vector machinesrdquo Chinese Physics vol 13no 4 pp 454ndash458 2004
[7] S Swaddiwudhipong K K Tho Z S Liu J Hua and N S BOoi ldquoMaterial characterization via least squares support vectormachinesrdquo Modelling and Simulation in Materials Science andEngineering vol 13 no 6 pp 993ndash1004 2005
[8] Y Wan ldquoPump performance analysis based on least squaressupport vector machinerdquo Transactions of the Chinese Society ofAgricultural Engineering vol 25 no 8 pp 115ndash118 2009
[9] X Wang X Tian and Y Cheng ldquoValue approximation withleast squares support vector machine in reinforcement learningsystemrdquo Journal of Computational and Theoretical Nanosciencevol 4 no 7-8 pp 1290ndash1294 2007
[10] K Huang H JWang H R Xu J PWang and Y B Ying ldquoNIRspectroscopy based on least square support vector machines forquality prediction of tomato juicerdquo Spectroscopy and SpectralAnalysis vol 29 no 4 pp 931ndash934 2009
(2) The core function is chosen the mapping relationbetween the input and output parameters is obtained throughsample learning
(3)Thenew parameters are input into the regression leastsquares support vectormachine to carry out prediction of thehydraulic excavator performance
(4) The new learning samples of the hydraulic excavatorsystem are added to the regression least squares supportvector machine then the prediction ability of model can beimproved
The prediction effect of the least squares support vectormachine can be judged by the root-mean-square error whichis expressed as follows [10]
119864 = radic1
119899
119899
sum119894=1
(119910119894minus 119910119894
119910119894
)
2
times 100 (7)
Through field test for the hydraulic excavator 100 groupsamples are obtained and the former 60 groups of samplesare used as performance predicting samples of the hydraulicexcavator and the other 40 groups of samples are used asverification samples of prediction model the performanceprediction simulation programmer of the hydraulic excavatoris compiled by MATLAB software
The input parameters of performance prediction modelfor the hydraulic excavator are listed as follows rotating speedof engine 119899
119868119862119864 the rotating speed of motor 119899
119864119872 rotating
speed of the hydraulic pump 119899119875 the torque of the engine119879
119868119862119864
the torque of the motor 119879119864119872
and the torque of the hydraulicpump 119879
119875 the output parameters of the model are output
power of the engine 119873119868119862119864
and the pilot controlling pressureof hydraulic control valve119875 respectively the part samples areshown in Table 1
The performance predicting curve and actual measure-ment curve of output power for the engine are shown inFigure 1 As seen from Figure 1 the output power of theengine changes in the moment with time and output power
75
45
15
10 20 30 4000
60
30
Time (s)
Measured valuePrediction value
Out
put p
ower
of e
ngin
e (kW
)
Figure 1 Performance predicting curve and measured curve ofoutput power for engine
Table 2 Output power prediction results of the hydraulic excavatorbased on different prediction model
changes irregularly the predicting results agree with theactual measurements and these results show that the leastsquares support vector machine based on regression hasbetter prediction ability
During the working process of the hydraulic excavatorthe input parameter can be regulated based on the predictionresults of the least squares support vector machine based onregression then the hydraulic excavator can work steadilyand it can be in the optimal performance point the workingefficiency of the hydraulic excavator can be improved accord-ingly
The output power prediction results of the hydraulicexcavator from the regression least square support vectormachine traditional support vector machine and BP neutralnetwork are compared which are shown in Table 2
As seen from Table 2 the prediction results from theregression least support vectormachine are closer to themea-sured value than those from tradition support vectormachineand BP neutral network The regression least support vectormachine can obtain best performance prediction results ofthe hydraulic excavator
4 Journal of Control Science and Engineering
5 Conclusions
The least squares support vectormachine based on regressionis applied in the performance prediction of the hydraulicexcavator the model of the least squares support vectormachine based on regression is established the correspond-ing algorithm procedure is designed and the performanceprediction model of the hydraulic excavator is establishedThe performance prediction simulation is carried out andresults show that the regression least squares support vectormachine has higher prediction precision which can predictthe changing rules of the performance for the hydraulicexcavator and improve theworking efficiency which haswideapplication space
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] MHagaWHiroshi andK Fujishima ldquoDigging control systemfor hydraulic excavatorrdquo Mechatronics vol 11 no 6 pp 665ndash676 2001
[2] Z-Y Jia J-W Ma F-J Wang andW Liu ldquoHybrid of simulatedannealing and SVM for hydraulic valve characteristics predic-tionrdquo Expert Systems with Applications vol 38 no 7 pp 8030ndash8036 2011
[3] Q P Ha Q H Nguyen D C Rye and H F Durrant-Whyte ldquoImpedance control of a hydraulically actuated roboticexcavatorrdquoAutomation in construction vol 9 no 5 pp 421ndash4352000
[4] Y Hong and C W W Ng ldquoBase stability of multi-proppedexcavations in soft clay subjected to hydraulic upliftrdquo CanadianGeotechnical Journal vol 50 no 2 pp 153ndash164 2013
[5] J Jiang C Song and L Bao ldquoForwardGene selection algorithmbased on least squares support vector machinerdquo Journal ofBionanoscience vol 7 no 3 pp 307ndash312 2013
[6] M-Y Ye and X-DWang ldquoChaotic time series prediction usingleast squares support vector machinesrdquo Chinese Physics vol 13no 4 pp 454ndash458 2004
[7] S Swaddiwudhipong K K Tho Z S Liu J Hua and N S BOoi ldquoMaterial characterization via least squares support vectormachinesrdquo Modelling and Simulation in Materials Science andEngineering vol 13 no 6 pp 993ndash1004 2005
[8] Y Wan ldquoPump performance analysis based on least squaressupport vector machinerdquo Transactions of the Chinese Society ofAgricultural Engineering vol 25 no 8 pp 115ndash118 2009
[9] X Wang X Tian and Y Cheng ldquoValue approximation withleast squares support vector machine in reinforcement learningsystemrdquo Journal of Computational and Theoretical Nanosciencevol 4 no 7-8 pp 1290ndash1294 2007
[10] K Huang H JWang H R Xu J PWang and Y B Ying ldquoNIRspectroscopy based on least square support vector machines forquality prediction of tomato juicerdquo Spectroscopy and SpectralAnalysis vol 29 no 4 pp 931ndash934 2009
The least squares support vectormachine based on regressionis applied in the performance prediction of the hydraulicexcavator the model of the least squares support vectormachine based on regression is established the correspond-ing algorithm procedure is designed and the performanceprediction model of the hydraulic excavator is establishedThe performance prediction simulation is carried out andresults show that the regression least squares support vectormachine has higher prediction precision which can predictthe changing rules of the performance for the hydraulicexcavator and improve theworking efficiency which haswideapplication space
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] MHagaWHiroshi andK Fujishima ldquoDigging control systemfor hydraulic excavatorrdquo Mechatronics vol 11 no 6 pp 665ndash676 2001
[2] Z-Y Jia J-W Ma F-J Wang andW Liu ldquoHybrid of simulatedannealing and SVM for hydraulic valve characteristics predic-tionrdquo Expert Systems with Applications vol 38 no 7 pp 8030ndash8036 2011
[3] Q P Ha Q H Nguyen D C Rye and H F Durrant-Whyte ldquoImpedance control of a hydraulically actuated roboticexcavatorrdquoAutomation in construction vol 9 no 5 pp 421ndash4352000
[4] Y Hong and C W W Ng ldquoBase stability of multi-proppedexcavations in soft clay subjected to hydraulic upliftrdquo CanadianGeotechnical Journal vol 50 no 2 pp 153ndash164 2013
[5] J Jiang C Song and L Bao ldquoForwardGene selection algorithmbased on least squares support vector machinerdquo Journal ofBionanoscience vol 7 no 3 pp 307ndash312 2013
[6] M-Y Ye and X-DWang ldquoChaotic time series prediction usingleast squares support vector machinesrdquo Chinese Physics vol 13no 4 pp 454ndash458 2004
[7] S Swaddiwudhipong K K Tho Z S Liu J Hua and N S BOoi ldquoMaterial characterization via least squares support vectormachinesrdquo Modelling and Simulation in Materials Science andEngineering vol 13 no 6 pp 993ndash1004 2005
[8] Y Wan ldquoPump performance analysis based on least squaressupport vector machinerdquo Transactions of the Chinese Society ofAgricultural Engineering vol 25 no 8 pp 115ndash118 2009
[9] X Wang X Tian and Y Cheng ldquoValue approximation withleast squares support vector machine in reinforcement learningsystemrdquo Journal of Computational and Theoretical Nanosciencevol 4 no 7-8 pp 1290ndash1294 2007
[10] K Huang H JWang H R Xu J PWang and Y B Ying ldquoNIRspectroscopy based on least square support vector machines forquality prediction of tomato juicerdquo Spectroscopy and SpectralAnalysis vol 29 no 4 pp 931ndash934 2009
