類神經網路估測器應用在機車稀混合汽控制之研究

Research on the Neural Network Estimator Applied on the Lean Air/Fuel Mixture Control for Motorcycles

黃瀚宣(Han-Hsuan Huang)

指導教授 ： 鄒忠全

崑山科技大學/工程學院/機械工程研究所/碩士(2011年)

摘要

本文提出類神經網路應用在機車稀混合汽控制之研究，估測引擎在不同負載下暫態時之空燃比變化，將空燃比控制在16：1主要是使油耗降低，利用空燃比估測器改善因為節氣門開啟瞬間空燃比18：1過稀導致動力輸出延遲的現象。
首先利用Matlab/Simulink軟體撰寫機車的閉迴路噴油控制系統，並將空燃比設定由原本的14.7：1更改為16：1，以執行稀混合汽噴油控制。然後利用Matlab/Neural Network發展混合汽空燃比的估測技術，利用引擎的噴油嘴噴油脈衝、進氣歧管壓力、節氣門開度及引擎轉速信號經類神經空燃比估測器就能正確地估測真實空燃比。最後在測試台上實驗驗證噴油控制策略和空燃比估測器的功能與其正確性。結果顯示在引擎穩態時噴油控制策略可讓空燃比控制在設定的16：1附近；空燃比估測器也能準確地估測出引擎暫態時的空燃比。

關鍵字

稀混合汽控制 ； 空燃比估測器 ； 類神經網路

並列摘要

This thesis proposes a research on the neural network estimator applied on the lean air/fuel mixture control for motorcycles. This NN estimator can estimate the air/fuel ratio under different transient load conditions and help the engine control the air/fuel ratio at 16:1. Using NN estimator can result in the reduction of fuel consumption and improve the power output delay due to instantaneous throttle opening and air/fuel ratio too lean (near 18:1).
First, using Matlab/Simulink software, the closed-loop fuel injection control system was written and the air/fuel ratio was changed from 14.7:1 to 16:1, in order to carry out lean air/fuel mixture control for motorcycles. Then, the Neural Network air/fuel ratio estimator was developed by using Matlab/Neural Network package to estimate actual air/fuel ratio from the parameters of fuel injection pulse width, manifold absolute pressure, throttle opening, and engine speed signal. At last, the lean air/fuel mixture control strategy and NN air/fuel estimator were validated by the experimental data from the test rig in the lab of KSU.
The results show the engine fuel injection control strategy can regulate the air/fuel ratio to 16:1 in the steady state and the NN estimator can estimate the actual air/fuel ratio in the transient state.

並列關鍵字

Lean Air/Fuel Mixture Control ； Air/Fuel Ratio Estimator ； Neural Network

參考文獻

[6] 甘宏基，“機車噴射引擎稀混合汽控制”，碩士論文，崑山科技大學機械工程系智慧車輛組， 2008。

[9] Jason S. Souder,“Powertrain Modeling and Nonlinear Fuel Control,”Masters of Science, Mechanical Engineering, University of California, Berkeley, 1998.

[10] Taglialatela F., Cesario N. and Lavorgna M. , “Soft Computing Mass Air Flow Estimator for a Single-Cylinder SI Engine,” SAE paper, No.2006-01-0010.

[11] Moskwa, J. J.,“Automotive Engine Modeling for Real Time Control,”Ph.D. thesis, Massachusetts Institute of Technology, pp. 45-71, 1988.

[12] Moskwa, J. J. and Hedrick, J. K.,“Modeling and Validation of Automotive Engine for Control Algorithm Development,” Transations of the ASME Journal of Dynamic System, Measurement, and Control, Vol.114,June,1992, pp. 278-285.