<<

The Road Toward Greener by Using Neuro- Fuzzy Modeling of Spark-Ignition with Variable Overlap

Mashhour M. Bani Amer1 and Yousef S. H. Najjar2,* 1Department of Biomedical Engineering 2Department of Mechanical Engineering Jordan University of Science and Technology, P.O.Box 3030, 22110 Irbid, JORDAN.

Received May 22, 2013; Accepted June 29, 2013

Abstract: Greening technology is gaining more attention in different aspects, especially the surface transport. This technology aims to improve performance and reduce global greenhouse gas emissions. In this work, a typical automotive spark ignition with four cylinders and 1600 cc displacement was considered. This work introduces a neuro-fuzzy approach that can be used to design an intelligent system for controlling the overlap angle in the spark ignition engine cycle. The simulation results showed that the neuro-fuzzy-based control system for varying the overlap angle achieved increased engine power and reduced fuel consumption in comparison with the conventional engine that used fi xed overlap angle. The power was improved by about 2.2–12.6 %, while the BSFC was reduced by about 2%-11%.

Keywords: Green cars, neuro-fuzzy modeling, variable valve-timing

1 Introduction

Greening technology is gaining more attention in different aspects, especially the surface transport. This technology aims to improve performance and reduce global greenhouse gas emissions. Alternative approaches comprise the use of hybrid vehicles [1, 2], Lithium ion batteries [3], artifi cial neural control of the air conditioning system [4], and improve engine performance by investigating strati- fi ed charge design [5, 6] and using alternative fuels [7] such as ethanol-gasoline [8–19], oxygenated fuels [11], fuels and combustion [12], liquefi ed petroleum gas LPG [13, 14], natural gas [15, 16], and hydrogen [17]. Variable geometry include length and [18], as well as variable valve-timing and lift

*Corresponding author: [email protected] DOI: 10.7569/JSEE.2013.629518

J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 275 Mashhour M. Baniamer et al.: The Road Toward Greener Cars by Using Neuro-Fuzzy Modeling to improve the volumetric effi ciency [19]. Detailed studies are being carried out for variable inlet and exhaust valve-timing and valve geometry covering valve lift in addition to variable valve, independently by the author. The volumetric effi ciency is defi ned as the mass of fuel and air induced into the divided by the mass that would occupy the displaced volume at the den- sity in the intake manifold. It is desirable to maximize the volumetric effi ciency of an engine because the amount of fuel that can be burned, hence power produced for a given is maximized (whereas size and weight are mini- mized). The volumetric effi ciency depends on the intake manifold confi guration, valve size, lift, and timing [20–23]. The shape and location of the peaks of the volumetric effi ciency curve are very sensitive to the manifold confi guration, type of fuel delivery system, and the type of cooling system [24, 25]. If the inlet pressure is less than the exhaust pressure, the engine is throttled. In this case, there is fl ow from the cylinder into the intake port when the intake valve opens. In the initial portion of the intake stroke, the induced gas is primarily composed of combustion products that have previously fl owed into the intake port. In the latter portion of the stroke, the mixture fl owing in is fresh charge, undiluted by any combustion products. If the inlet pressure is greater than the exhaust pressure, then there is fl ow from the intake port into the engine until the pressure equilibrates [20, 25]. In actual engines, because of valve overlap, there may be a fl ow of fresh mixture from the inlet to the exhaust port, which can waste fuel and be a source of hydro- carbon exhaust emissions. The second case is when inlet and exhaust pressures are equal; the engine is then said to be un-throttled. The throttling also hurts the volu- metric effi ciency, mainly because of an increase in the residual fraction. The residual fraction decreases with increasing compression ratio, as one would expect [20]. Introducing γ=1.4 and the inlet Mach index Z defi ned by Taylor [24]

2 Z = (π/4)b Up/Af ci (1) For good volumetric effi ciency, one should keep the Mach index down to less than about Z=0.6. Based on the analyses, we can interpret this to mean that the average gas speed through the inlet valve should be less than the sonic velocity, so that the intake fl ow is not chocked. Hence, inlet can be sized on the basis of the maximum speed for which the engine is designed. Likewise for effi cient expulsion of the , the average effective area Ae, of the exhaust valves should be such that their Mach index is less than about 0.6, in which case, relative to intake conditions.

1/2 Ae/Ai ~ ci /ce = (Ti/Te) (2)

A smaller exhaust valve diameter and lift (l~d/4) can be used because the speed of sound is higher in the exhaust gases than in the inlet gas fl ow. Current practice dictates that the exhaust to intake valve area ratio is on the order of 70 to 80% [26–27].

DOI: 10.7569/JSEE.2013.629518

276 J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 Mashhour M. Baniamer et al.: The Road Toward Greener Cars by Using Neuro-Fuzzy Modeling

During overlap it is assumed that the gases are pushed into the intake mani- fold only to return later. Notice that heat loss increases the residual fraction and is important here because the exhaust gases are considerably hotter than the cylinder walls. This is the main reason for the discrepancy between fuel-air cycle calcula- tions with ideal intake and exhaust and the experiments. Inlet and exhaust mani- folds are sized or “tuned” to use the pressure waves to optimize the volumetric effi ciency at a chosen engine speed. A tuned intake manifold will have a locally higher pressure when the intake valve is open, increasing the exhaust outfl ow. To minimize engine size and produce a given torque versus speed curve (with torque proportional to the volumetric effi ciency at fi xed thermal effi ciency), it is clearly desirable to be able to vary valve-timing with engine speed. Variable valve- timing (VVT) is a technique that can address the problem of obtaining optimal engine performance over a range of and engine speed. Variable valve- timing allows the intake and exhaust valves to open and close at varying angles, depending on the speed and load conditions. At idle, with a nearly closed throttle, the intake and exhaust valve overlap is minimized to reduce exhaust back fl ow. At low speed, the intake valves are closed earlier to increase volumetric effi ciency and torque. At high speed, with an open throttle, the intake valves are closed later to increase volumetric effi ciency and power. Data obtained from real spark-ignition engines illustrates the effect of varying valve-timing on the volumetric effi ciency versus speed. Earlier than normal inlet η valve closing reduces back fl ow losses at low speed and increases v. The penalty is reduced airfl ow at a high speed later than normal inlet valve closing, which is only advantageous at very high speeds. To date, relatively little attention has been devoted to the modelling and control of variable valve-timing using intelligent techniques, such as artifi cial neural net- works, fuzzy logic, or neural fuzzy approaches [28, 29]. In reference [29], the control of valve-timing is based on conventional (not intelligent) control techniques. In ref- erence [30], the authors used intelligent approach using artifi cial neural network to determine the effects of intake valve-timing on the engine performance. However, they did not suggest an intelligent approach for controlling the valve-timing to achieve optimum engine performance in terms of power and fuel consumption. Furthermore, the combining of neural networks and fuzzy logic to form a hybrid neuro-fuzzy approach enhances the performance of the approach in terms of learn- ing, adaptation and accuracy. This is why attempts have been made, as described in this paper, to develop a neural fuzzy model suitable for designing an intelligent control approach of variable valve-timing in order to improve the performance of the spark-ignition engine in terms of power and fuel consumption [31].

2 Experimental Results and Discussion There is a valve overlap period at top dead center where the exhaust and intake valves are both open. This creates a number of fl ow effects. With a spark-ignition

DOI: 10.7569/JSEE.2013.629518

J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 277 Mashhour M. Baniamer et al.: The Road Toward Greener Cars by Using Neuro-Fuzzy Modeling engine at part throttle, there will be back fl ow of the exhaust into the inlet mani- fold since the exhaust pressure is greater than the throttled intake pressure. This will reduce the part load performance because the volume available to the intake charge is less, reducing the volumetric effi ciency. Rough idle can also occur due to unstable combustion. On the other hand, because this dilution will reduce the peak combustion temperatures, the NOx pollution levels will also be reduced [32]. At wide open throttle, with both valves open, there will be some short circuit- ing of the inlet charge directly to the exhaust, because, in this case, the intake pres- sure is greater than the exhaust pressure. This will reduce the full load performance because a fraction of the fuel is not burning in the cylinder. The higher performance engine operates at much higher piston speeds at wide open throttle, with power and volumetric effi ciency as the important factors, whereas the conventional engine operates at lower speeds (rpm), with idle and part load performance of importance. Therefore, the high performance intake valve opens about 25 before the conventional intake valve, and closes about 30 after the conventional intake valves. As the engine design speed increases, to maintain a maximum valve opening during the intake and exhaust strokes, the intake and the exhaust valves are open for a longer dura- tion, from about 230 to about 285. Early opening of the exhaust valve will reduce the expansion ratio, but will also reduce the exhaust stroke pumping work [33, 34]. In this work, a typical automotive spark-ignition engine with four cylinders and 1600 cc displacement was considered. As shown in Figure 1, the engine design overlap angle θ=30° in the speed range of 3500–4500 rpm. The overlap angle lies between IVO and EVC around the TDC in the cycle. The maximum volumetric

55

50

45

40

35

30

25

20 Overlap angle (Degree) 15

10

5 0 1000 2000 3000 4000 5000 6000 Engine speed (RPM)

Figure 1 The overlap angle as a function of engine speed.

DOI: 10.7569/JSEE.2013.629518

278 J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 Mashhour M. Baniamer et al.: The Road Toward Greener Cars by Using Neuro-Fuzzy Modeling

86

84

82

80

78 (%) efficiency Volumetric 76

74 0 1000 2000 3000 4000 5000 6000 Engine speed (RPM)

Figure 2 The volumetric effi ciency as a function of engine speed.

η η effi ciency v occurs at engine speed around 4000 rpm (Figure 2). v then decreases for both lower (idle) and high (wide-open throttle wot) speeds. To overcome this problem and enhance power and effi ciency over the whole operating range, a vari- θ ° able valve-timing (VVT) system is used to decrease idle to 10 at relatively low speeds 3000rpm and lower. It also increases θ to 50∞ at higher speeds more than η wotη 5000 rpm to increase v, hence sustaining high v over the whole range of operat- ing speeds. Consequently the power output is boosted between 2.2–12.6% and the fuel economy is improved by 2–11%, thereby reducing resultant pollutants, especially CO2, contributing towards green transportation. Accordingly, it is sug- gested that an electronic signal is sent from the engine control module (ECM) to the actuators, which change the valve-timing, hence the valve angle according to the engine loading conditions expressed by the engine speed.

3 Neuro-fuzzy Control Approach and Results The control approach was developed using adaptive neuro-fuzzy inference sys- tem (ANFIS) [35, 36]. In comparison with a purely fuzzy approach, the ANFIS removes the requirement for manual accomplishment of fuzzy rules and manual optimization of the fuzzy system parameters and also automatically tune the sys- tem parameters to achieve negligible prediction error. Furthermore, in comparison with a purely artifi cial neural networks approach, it is able to include the uncer- tainty that exists in the experimental data. Thus, the use of a neuro-fuzzy approach will result in a more accurate modeling and enable control of variable valve-timing

DOI: 10.7569/JSEE.2013.629518

J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 279 Mashhour M. Baniamer et al.: The Road Toward Greener Cars by Using Neuro-Fuzzy Modeling through the control of the overlap angle because it includes the features of both fuzzy logic and artifi cial neural networks. To design the neuro-fuzzy control (NFC) approach of valve-timing, the speed was considered as an input for the controller while its output is the overlap angle (θ). The experimental data were divided into two data sets: training (21 measurements) and checking (seven measurements). As shown in Table 1, the NFC approach was devel- oped using four fuzzy rules and four Bell-shaped membership functions for the input speed (Figure 3). The NFC approach was generated using the grid partitioning method and trained using 21 data set measurements. Then, the performance of the developed NFC approach was tested using the checking data that are not included in the training stage. The obtained results showed (Figure 4) that the developed NFC approach is capable of predicting the overlap angle with a root-mean square error (RMSE) of 0.62 for a number of epochs higher than 200. The RMSE for checking data set is slightly higher (0.67) but their difference is statically insignifi cant. 1 ⎡⎤n 2 =θ−θ⎢⎥1 2 RMSE ∑()i,, des.. i pred (3) ⎣⎦⎢⎥n i = 1

θ θ Where i,des. and i,pred. represent the desired and predicted values of overlap angle, respectively and n represents the number of data points Figure 5 illustrates a comparison between the desired (measured) and pre- dicted (from the NFC approach) overlap angle from which it can be concluded

Table 1 The parameters of the developed NFC approach. Number of nodes 20 Number of linear parameters 8 Number of nonlinear parameters 12

Total number of parameters 20 Number of training data pairs 21 Number of checking data pairs 7 Generation Method of Fuzzy Inference System Grid Partition Training Method of Fuzzy inference system Hybrid Number of fuzzy rules 4 Training RMSE 0.62 Checking RMSE 0.67

DOI: 10.7569/JSEE.2013.629518

280 J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 Mashhour M. Baniamer et al.: The Road Toward Greener Cars by Using Neuro-Fuzzy Modeling

Very low Low High Very High 1

0.5

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 Input variable “Speed.(RPM)”

Figure 3 Plot of membership function for the speed.

1.4 Training error

1.2

1 Error

0.8

0 50 100 150 200 250 300 Epochs Figure 4 The training root-mean square error (RMSE) as a function of epochs.

60 Training data: 0 FIS output: *

40 Output 20

0 0 5 10 15 20 25 Index Figure 5 Plot showing a comparison between the desired (o) and predicted (*) overlap angle.

DOI: 10.7569/JSEE.2013.629518

J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 281 Mashhour M. Baniamer et al.: The Road Toward Greener Cars by Using Neuro-Fuzzy Modeling

14

12

10

8

6

4

Power increase (%) Power 2

0

–2 0 1000 2000 3000 4000 5000 6000 Engine speed (RPM)

Fig ure 6 The percentage increase of engine power as a result of using the proposed NFC approach.

14

12

10

8

6

4 BSFC decrease (%) 2

0

–2 0 1000 2000 3000 4000 5000 6000 Engine speed (RPM)

Fig ure 7 The percentage decrease in BSFC as a result of using the proposed NFC approach. that the predicted results agree closely with the measured ones. Figures 6 and 7 demonstrate the percentage increase in power and decrease in BSFC produced by the spark-ignition engine as a result of applying variable overlap angle instead of fi xed one. As shown in these Figures, the use of neuro-fuzzy control

DOI: 10.7569/JSEE.2013.629518

282 J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 Mashhour M. Baniamer et al.: The Road Toward Greener Cars by Using Neuro-Fuzzy Modeling results in a good improvement of the engine power and BSFC. The increase of engine power and decrease of BSFC imply that a signifi cant amount of fuel is saved with better engine power, thus achieving the greening technology aspects.

4 Conclusions 1. Volumetric effi ciency of conventional engines and, consequently, power and thermal effi ciency drop when keeping valve overlap θ fi xed. θ ° 2. Variable valve-timing is used to decrease idle to 10 at lower speeds, and θ ° η increases wot to 50 at higher speeds, thus sustaining high v over the whole operating range. 3. The simulation results showed that the neuro-fuzzy-based control system for varying overlap angle achieved increased engine power and reduced fuel consumption in comparison with the conventional engine that uses a fi xed overlap angle. 4. The power was improved by about 2.2–12.6 %, while the BSFC was reduced by about 2 -11%.

Nomenclature Symbols Description Unit –A Average area [m2] ANFIS Adaptive neuro-fuzzy inference controller [ - ] b Cylinder [m]

BSFC Brake specifi c fuel consumption [kg /kW h] c Flow speed [m/s] ECM Engine control module [ - ] EVC Exhaust valve closes [ - ] IVO Inlet valve opens [ - ] mfs Membership functions [ - ] N Engine speed [rpm] NFC Neuro-fuzzy control [ - ] -

NOx Nitrogen oxide [ - ] RMSE Root-mean square error [ - ] T Gas temperature [K] TDC Top dead centre [ - ] U Piston speed [ m/s] VVT Variable valve-timing [ - ] Z Mach index [ - ] Superscript – Average

DOI: 10.7569/JSEE.2013.629518

J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 283 Mashhour M. Baniamer et al.: The Road Toward Greener Cars by Using Neuro-Fuzzy Modeling

Subscript i inlet e exhaust f fl ow p piston Greek letters η v Volumetric effi ciency [%] θ Overlap angle [degree] θ idle Overlap angle at idle θ wot Overlap angle at wide open throttle

References

1. C. Silva, M. Ross, and T. Farias, Evaluation of energy consumption, emissions and cost of plug-in hybrid vehicles. Energ. Convers. Manage. 50, 1635–1643 (2009). 2. K. Bayinder, M. Gozukucuk, and A. Teke, A comprehensive overview of hybrid electric vehicle: Power train confi gurations, power train control techniques and electronic con- trol units. Energ. Convers. Manage. 52, 1305–1313 (2011). 3. K.M. Tsang, L. Sun, and W.L.Chan, Identifi cation and modelling of lithium ion battery. Energ. Convers. Manage. 5, 2857–2862 (2010). 4. M. Hosoz and H.M. Ertunc, Artifi cial neural network analysis of an automobile air con- ditioning system. Energ. Convers. Manage. 47, 1574–1587 (2006). 5. M. Lounici, K. Loubar, M. Balistrou, and M.Tazerout, Investigation on heat transfer evaluation for a more effi cient two-zone combustion model in the case of natural gas SI engines. Appl. Therm. Eng. 31, 319–328 (2011). 6. Y.S.H. Najjar, Comparison of performance of a direct-injection stratifi ed charge engine (DISC) with a spark-ignition engine. Energy 36, 136–143 (2011). 7. A. Demirbas, Present and future transportation fuels. Energ. Source., Part A: Recovery, Utilization, and Environmental Effects 30, 1473–1483 (2008). 8. R.C. Costa and J.R. Sodre, Compression ratio effects on an ethanol/gasoline fuelled engine performance. Appl. Therm. Eng. 31, 278–283 (2011). 9. M.A. Ceviz and F.Yuksel, Effects of ethanol-unleaded gasoline blends on cyclic variabil- ity and emissions in an SI engine. Appl. Therm. Eng. 25, 917–925 (2005). 10. D. Hung, G. Zhu, and H. Schock, Combustion characteristics of a single-cylinder spark ignition gasoline and ethanol dual-fuelled engine. Proc. IMechE. Part D: J. Automob. Eng. 224, 850–858 (2009). 11. M. Nabi, Theoretical investigation of engine thermal effi ciency, adiabatic fl ame tem-

perature, NOx emission and combustion-related parameters for different oxygenated fuels. Appl. Therm. Eng. 30, 839–844 (2010). 12. D. Bradley, Combustion and the design of future engine fuels. Proc. IMechE. Part C:J. Mech. Eng. Sci. 223, 320–331 (2009). 13. P. Price, G. Guo, and M.Hischmann, Performance of an evaporator for a LPG powered vehicle. Appl. Therm. Eng. 24, 1179–1194 (2004). 14. M.A. Said, T.F. Yusaf, and I. Hussein, Performance and emission investigation of a four- stroke liquefi ed petroleum gas spark-ignition engine generator used in a Malaysian night market. Proc. IMechE, Part A: J. Power Energy 224, 650–658 (2009).

DOI: 10.7569/JSEE.2013.629518

284 J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 Mashhour M. Baniamer et al.: The Road Toward Greener Cars by Using Neuro-Fuzzy Modeling

15. X. Zhao, L. Fu, S. Zhang, Y. Li, Y. Jiang, H. Li, and Z. Sun, Performance study of an inno- vative natural gas CHP system. Energ. Convers. Manage. 52, 321–328 (2011). 16. D. Lowers, S. Aceves, C. Westerbrook, and J. Smith, Detailed chemical kinetic simu- lation of natural gas HCCL combustion: Gas composition effects and investigation of control strategies. J. Eng. Gas Turb. Power 123, 433 (2001). 17. A.R. Maher and S. Al-Baghdadi, Effect of compression ratio, equivalence ratio and engine speed on the performance and emission characteristics of a spark ignition engine using hydrogen as a fuel. Renew. Energ. 29, 2245–2260 (2004). 18. J. Yamin and M. Dado, Performance simulation of a four-stroke engine with variable stroke-length and compression ratio. Appl. Therm. Eng. 77, 447–463 (2004). 19. M. Golcu,Y. Sekmen, P. Erduranlu, and S. Salman, Artifi cial neural-network based mod- eling of variable valve-timing in a spark-ignition engine. Appl. Energ. 81, 187–197 (2005). 20. T.G. Asmus, Perspectives on application of variable valve-timing. SAE Paper, 910445, (1991). 21. E. Sher and T. Bar-Kohany, Optimization of variable valve-timing for maximizing per- formance of an un-throttled SI engine: A theoretical study. Energy 27, 757–775 (2002). 22. S. Nagumo and S. Hara, Study of fuel-economy improvement through control of intake valve-closing timing: Cause of combustion deterioration and improvement. JSAE Rev. 16, 13–19 (1995). 23. T. Kohany and E. Sher, Using the 2nd law of thermodynamics to optimize variable valve- timing for maximizing torque in a throttled SI engine. SAE Paper, 1999–2028 (1999). 24. C.F. Taylor, The Internal Combustion Engine in Theory and Practice, The MIT Press, Cambridge, Mass. (1985). 25. C.R. Ferguson and A.T. Kirpatrick, Internal Combustion Engines-Applied Thermo-Sciences, Wiley and Sons, New York, NY. (2001). 26. K. Maekawa and N. Ohsawa, Development of a valve-timing control system. SAE Paper, 890680, (1989). 27. T. Dresner and P. Barkan, A review and classifi cation of variable valve-timing, SAE paper, 890674, (1989). 28. M. Golcu, Y. Sekmen, P. Erduranli, and M.S. Salman, Artifi cial neural network based modeling of variable-timing in a spark-ignition engine. Appl. Energ. 81, 187–197 (2005). 29. S.R. Jang, C.T. Sung, and E. Mizutani, Neurofuzzy and Soft Computing, A Computational Approach to Learning and Machine Intelligence, Prentice-Hall, Englewood Cliffs, NJ, USA, (1992). 30. T. Hosaka and M. Hamazaki, Development of the variable valve-timing and lift (VTEC) engine for the NSX. SAE Paper, 910008, (1991). 31. T. Nakayasu, H. Yamada, T. Suda, N. Iwase, and K. Takahashi, Intake and exhaust sys- tems equipped with a variable valve-control device for enhancing of engine power. SAE Paper, 0247, (2001). 32. F. Bozza, A. Gimelli, A. Senatore, and A. Caraceni, A theoretical comparison of various VVA systems for performance and emission improvements of SI-engines. SAE paper, 0235, (2001). 33. J.B. Heywood, Internal Combustion Engine Fundamentals, McGraw-Hill, (1988). 34. R. Babuška, Fuzzy Modeling for Control, Kluwer Academic Publishers, Dordrecht, The Netherland, (1998). 35. R. Babuška and H. Verbruggen, Neuro-fuzzy methods for nonlinear system identifi ca- tion. Annu. Rev. Control 27, 73–85 (2003). 36. M. Brown and C. Harris, Neuro-Fuzzy Adaptive Modeling and Control, Prentice-Hall, Englewood Cliffs, NJ, USA, (1994).

DOI: 10.7569/JSEE.2013.629518

J. Sustainable Energy Eng., Vol. 1, No. 4, January 2014 285