research on the optimization design of motorcycle engine based on doe methodology

Peer-review under responsibility of the organizing committee of the 13th Global Congress on Manufacturing and Management doi: 10.1016/j.proeng.2017.01.216 ScienceDirect 13th Global Cong

Procedia Engineering 174 ( 2017 ) 740 – 747

1877-7058 © 2017 The Authors Published by Elsevier Ltd This is an open access article under the CC BY-NC-ND license

( http://creativecommons.org/licenses/by-nc-nd/4.0/ ).

Peer-review under responsibility of the organizing committee of the 13th Global Congress on Manufacturing and Management

doi: 10.1016/j.proeng.2017.01.216

ScienceDirect

13th Global Congress on Manufacturing and Management, GCMM 2016

Research on the Optimization Design of Motorcycle Engine Based

on DOE Methodology

Li Yongfana, Zhang Shuaia, Wang Jingb*

a Mechanical Electrial Engineering School, Beijing Information Science & Technology University, Beijing 100192, China

b Business School, Nankai University, Tianjin 300071, China

Abstract

The optimization design of engine is always one of the top concerns in motorcycle industry In this paper, effective torque and fuel consumption ratio are defined as the performance evaluation indexes of engine, while air-fuel ratio, intake valve timing angle, exhaust valve timing angel, pressure, and temperature are defined as input variables With the application of DOE methodology, a full factorial DOE is conducted to estimate the regression model and identify the statistical significant factors And then, with the selection of additional experimental points, RSM is introduced to construct the precise regression model between input variables and performance indexes Based on that, an optimum solution that can satisfy both performance requirements are brought forward and testified

© 2016 The Authors Published by Elsevier Ltd

Peer-review under responsibility of the organizing committee of the 13th Global Congress on Manufacturing and Management

Keywords: Optimization Design, DOE, RSM, Motorcycle Engine

1 Introdution

Just as heart is to human body, engine is the most important part of motorcycle As an example, a four-stroke engine can provide continuous power for motorcycle with the circulation of four strokes: intake stroke, compression stroke, power stroke and exhaust stroke With the application of Variable Valve Timing (VVT) technology [1] and Variable Intake Manifold (VIM) System [2], engineers can enhance the engine power by controlling the switching time of valves, and then improve the dynamic performance Generally, the dynamic performance can be described

* Corresponding author: Wang Jing Tel.: +86-22-2350-0108

E-mail address: wangjingteda@nankai.edu.cn

© 2017 The Authors Published by Elsevier Ltd This is an open access article under the CC BY-NC-ND license

( http://creativecommons.org/licenses/by-nc-nd/4.0/ ).

Peer-review under responsibility of the organizing committee of the 13th Global Congress on Manufacturing and Management

as effective torque, effective power, average effective pressure, and rotational speed, in which effective torque is preferred as the key index to reflect the engine’s working ability Besides that, fuel economy, compactness, reliability and durability should be considered synthetically when evaluating an engine’s performance Considering the complicated interaction among different evaluation indexes, engineers usually rely on simulation software (such

as CAE, GT series, Converge, and etc.) to imitate the operation process of engine [3], and then optimize the design However, because most of variables that influence the performance of engine are of numerical data, there will be infinite variable portfolios Though the application of simulation software can significantly reduce the experiment cost and enhance the experiment efficiency when each variable has a specific value, it can hardly deal with the situation of infinite potential variable combinations So, we must consider carefully how to design the experiment plan properly, by which we can identify the key variables and their optimum value intervals with as few number of experiments as possible

To solve this problem, in this paper, we introduce DOE (Design of Experiment) methodology into the R&D process of motorcycle engine, to arrange and conduct the experiments rationally [4] With the statistical analysis of simulational experiment data, we set up the functional relationship model, discuss about the influences of different variables on the performance of engine, then search out and test the optimum variables combination

2 Functional Model Definition and Description

2.1 The Definition of Outputs (Response) “Y”

Generally, the evaluation indexes of motorcycle engine include power, fuel economy, strength, compactness, reliability, durability, and etc Considering the function design requirement, we mainly focus on the dynamic and economic performance of engine We define Effective Torque (ET) as Y1 to represent the dynamic performance (measurement unit is “NЬm”), while Fuel Consumption Rate (FCR) as Y2 to represent the economic performance (measurement unit is “g/(kwЬh)”) Basically, greater torque of engine means higher acceleration performance and stronger off-road ability, so Y1 is a the-larger-the-better (LTB) type of characteristic; the smaller the fuel consumption rate is, the more cost can be saved, so Y2 is a the-smaller-the-better (STB) type of characteristic

2.2 The Definition of Input Variables (Factor) “X”

Considering the feasibility of modeling and function analysis, we choose Air-Fuel Ratio (A/F), Intake Valve Timing Angle (IVTA), Exhaust Valve Timing Angel (EVTA), Pressure (P), and Temperature (T) as the main input variables after many discussions with R&D staffs of motorcycle companies, and define them as X1, X2, X3, X4, X5

separately More descriptions about the X are as follows

Air-Fuel Ratio (A/F) It means the ratio between air quality and fuel quality in engine A/F can indicate the status of air-fuel mixture It is a very important variable for the operation of engine The measurement unit of A/F is

“%”

Intake Valve Timing Angle (IVTA) VVT technology can adjust the angle of intake valve dynamically to improve the combustion of fuel The measurement unit of IVTA is “Degree”

Exhaust Valve Timing Angel (EVTA) VVT technology can also adjust the angle of exhaust valve dynamically

to improve the combustion of fuel The measurement unit of EVTA is also “Degree”

Pressure (P) Pressure can promote the combination of air and atomized fuel when engine is working However, too much pressure can also cause the deflagration of mixture, and then raise the fuel consumption The measurement unit of P is “Bar”

Temperature (T) The temperature in cylinder can reflect the combustion state of mixture, and its measurement unit is “°C”

2.3 Functional Model

Assuming that there are linear relationships between X and Y, we can set up the functional models as follows:









     

 I ; ; ; ; ;

< (1)

H









     

 I ; ; ; ; ;

< (2)

In which, “ε” means the error, including both experimental error, and lack of fit

Next, we conduct a full factorial design and testify the validity of our assumption by ANOVA analysis

3 Full Factorial DOE and ANOVA Analysis

3.1 Design of Full Factorial Experiment

The first step of experiment design is to set the values for input variables Considering the realistic value intervals,

we assign two levels for each factor as Table 1

Table 1 Description of Factor Level Factor Name Low Level High Level Measurement Unit

According to orthogonal experimental design, a two-level full factorial DOE with five factors needs 25 (equal to 32) experimental runs And we added 4 centre-point experiments to evaluate the error “ε” So the whole experiment plan includes 36 experimental runs and Simulation results are as Table 2 (randomized when conducting experiments)

Table 2 Experiment Plan and Results of Two-level Full Factorial DOE

14.25 122.5 235 1 286.5 44.6578 242.506

3.2 ANOVA Results and Discussion

The ANOVA Analysis Results are as Table 3

Table 3 ANOVA Result: ET/FCR Versus Factors

For the practical explanation and application, we just consider the main effects and two-way interactions among factors According to the ANOVA analysis results, we notice that the P-Values of “Curvature” in both two tables are almost zero, which means that the linear assumption between the factors and responses are statistically rejected Therefore, based on the previous experiment results, we choose additional experimental points to construct response surface, and describe the curve relationships between X and Y precisely

4 Analysis and Optimization Based on Response Surface Methodology

4.1 The Selection of Additional Experimental Points

Rotatability should be considered sincerely when choosing additional experimental points to construct response surface model [5] It means that the variance of predicted value is only related with the distance from the experimental point to centre point, and is independent of its location According to central composite circumscribed (CCC) method [6], we choose ten additional experimental points (two points in each dimension of X symmetrically), and conduct another six experiments in the centre point to achieve uniform precision

The additional experimental points (including another centre points) and their simulation results are as Table 4

Table 4 the Additional Experimental Points and Their Results

14.25 122.5 199.3238 1 286.5 30.3585 261.908

4.2 Response Surface Regression Analysis and Discussion

The Results of Response Surface Regression are as Table 5

Table 5 Response Surface Regression Result: ET/FCR Versus Factors

According to the response surface regression results, the P-Values of “Lack-of-Fit” are “0.409” and “0.559”, so

we can judge that the models are fitted well Besides, factors of A/F, EVTA, IVTA, P, T, (A/F)2, (EVTA)2, (IVTA)2, EVTA*IVTA, and IVTA*P have statistically significant impact on ET; while factors of A/F, EVTA, IVTA, P, (A/F)2, (EVTA)2, and (IVTA)2 have statistically significant impact on FCR

The contour plots of responses (as showed in Figure 1 and Figure 2) also indicate intuitively the curve relationships between factors and responses

', 

+<:' 

/<:' 

: 

.URJ<GR[KY

+<:'',













/<:'',













6',













:',













/<:'+<:'













6+<:'













:+<:'













6/<:'













:/<:'













:6













$

Ȍ

Ȍ

Ȍ

Ȍ

"

 

 

 

 



+:

)UTZU[X6RUZYUL+:

Figure1 Contour Plots of ET

', 

+<:' 

/<:' 

.URJ<GR[KY

+<:'',















/<:'',















6',

















/<:'+<:'















6+<:'

















6/<:'

















$

Ȍ

Ȍ

Ȍ

Ȍ

"

 

 

 

 



,)8

)UTZU[X6RUZYUL,)8

Figure2 Contour Plots of FCR

4.3 The Optimization of the Regression Model

The regression model can be optimized with the deletion of insignificant factors The comparison between original and optimized model is showed in Table 6

Table 6 Regression Performance of Original Vs Optimized Model

ET

FCR

With the optimization, the final regression equation can be defined as follows

3 (97$



 ,97$

(97$



 ,97$

,97$







(97$

(97$





$)

$)



 7







3



 ,97$



 (97$





$)







(7

(3)

,97$

,97$



 (97$

(97$





$)

$)







3



 ,97$



 (97$





$)







)&5

(4)

5 The Calculation of Optimum Solutions and Further Discussion

5.1 The Calculation of Optimum Solutions

Giving the same weight and importance to both ET and FCR, we draw an optimization plot as Figure 3

Figure 3 Optimization Plot of X to Y

According to the calculation results, Effective Torque can be maximized to 57.9511 and Fuel Consumption Ratio

be minimized to 234.3743 simultaneously, when A/F is 14.1239, EVTA is 124.8424, IVTA is 225.9909, Pressure is 1.1189, and Temperature is 254.3914 95% Confidence Interval (CI) of ET is (56.870, 59.032), while CI of FCR is (231.14, 237.61) And 95% Prediction Interval (EI) of ET is (55.898, 60.004), while EI of FCR is (226.25, 242.50) All the conclusions are testified by several times simulation and operations in reality

5.2 Further Discussion

The optimization design of engine is always of the most important research area in motorcycle industry In this paper, we mainly focus on the effective torque and the fuel consumption ratio, and discuss about the approaches to the optimum solutions for the best performance of engine However, there still exist another “Y” that can be used to evaluate engine performance, and with the increase of “Y” numbers, the problem will become complicated exponentially [7] Some new development in multiple response surface methodology [8], [9] may be introduced to deal with these challenges in future research

Acknowledgements

This work was supported by grants from the National Natural Science Foundation of China (71302016, 71102047), Philosophy and Social Science Planning of Tianjin (TJTQ11-018), Fundamental Research Funds for the Central Universities (NKZXB1164), and the Social Science Foundation of Beijing Municipal Education Commission (71E1610980)

References

[1] P Mohammad, K Amir, Precise lift control in a new variable valve actuation system using discrete-time sliding mode control, Mechanism and machine theory, 2016, 99: 217-235

[2] F Zhao, A review of research about variable intake manifold, machinery, materials science and engineering applications, 2011, 228:

299-302

[3] R Alessandro, T Marco, Dynamic analysis of a motorbike engine timing system: Experimental and numerical investigation of the geartrain, Mechanical systems and signal processing, 2014, 48 (1-2):325-338

[4] T Hadi, J Samad, T Hamid, N Ali, Application of DOE evaluation to introduce the optimum injection strategy-chamber geometry of diesel engine using surrogate epsilon-SVR, Applied thermal engineering, 2016, 106: 56-66

[5] K Sudhir, D Anoop K, Response surface algorithms for engine mount optimization in motorcycles, ASME International Design Engineering Technical Conferences/Computers and Information in Engineering Conference, 2010, 711-726

[6] G.Stelios D., S Stella, A Manohar, A class of composite designs for response surface methodology, Computational statistics & data analysis, 2014, 71: 1124-1133

[7] B M Kemal, S Cenk, S Murat, Optimization of the operating parameters based on Taguchi method in an SI engine used pure gasoline, ethanol and methanol, Fuel, 2016, 180: 630-637

[8] L Shi, D Lin, J Peterson, A confidence region for the ridge path in multiple response surface optimization, European journal of operational research, 2016, 252(3): 829-836

[9] J Wang, Y Ma, L Ouyang, Y Tu, A new Bayesian approach to multi-response surface optimization integrating loss function with posterior probability, Eu ropean journal of operational research, 2016, 249(1): 231-237

4.3 The Optimization of the Regression Model

The regression model can be optimized with the deletion of insignificant factors The comparison... response surface, and describe the curve relationships between X and Y precisely

4 Analysis and Optimization Based on Response Surface Methodology

4.1 The Selection of. .. design of engine is always of the most important research area in motorcycle industry In this paper, we mainly focus on the effective torque and the fuel consumption ratio, and discuss about the