Modelling And Control Of The Quasi-Homogeneous Lean Burn Engine

It is necessary to apply electronic control technology to the engine in order to economize on energy and reduce the emission pollution. The new technology, quasi-homogeneous lean burn, can give full play to the advantage of lean burn, and thus help to reduce the fuel consumption and emission of the engine. However, higher requirements are raised for the engine control system. The quasi-homogeneous lean burn engine model is presented in this paper for the characteristics of the lean burn engine. On the basis of the mean value engine model, this lean burn engine model can be built up of three subsystems: manifold air mass flow, fuel vapor and fuel film, and crank shaft and loading. Thus scientific basis is provided for the air-fuel ratio (AFR) control scheme for the quasi-homogeneous lean burn engine and for the emission reduction control scheme for the quasi-homogeneous lean burn engine by AFR switching. To meet the special requirement of the air-fuel ratio control for lean burn engine, a sliding mode-neural network control scheme for the air-fuel ratio of the quasi-homogeneous lean burn engine is presented in this paper based on the quasi-homogeneous lean burn engine model. In this scheme, the sliding variable structure control strategy, in which the feedback control is its key idea, is designed using fuzzy exponential approach and adopted to control the air-fuel ratio under the condition of quasi-steady state. At the same time, the neural network method is applied to build the intake air flow observer and realize the instantaneous fuel compensation. To reduce the emission of the quasi-homogeneous lean burn engine, a comprehensive control scheme for the quasi-homogeneous lean burn engine is proposed in this paper. According to the working principle of NOx storage-reduction catalyst converter and its special demand for the control system, the sliding mode-neural network control scheme is adopted to control the injection pulse width, while the MAP converting method combined with online correction is adopted to realize the control of the throttle opening and the torque-based knocking closed-loop control scheme is adopted to control the spark ignition timing. The experiments show that by this scheme, the fluctuation of the output power can be confined to a smaller range, and it can be further reduced after certain transformation by the learning control. Thus the AFR step process of the engine can be optimized. To satisfy the special requirements of control experiments of gasoline engine, the author has developed an electronic control system suitable for lean burn gasoline engine with the self-developed hardware and software. The quasi-homogeneous lean burn engine air-fuel ratio control experiments and NOx emission comprehensive control experiments are carried out on a Toyota 8A gasoline engine with this system. By the experiments results, it can be shown that applying sliding mode-neural network scheme to control the air-fuel ratio of the quasi-homogeneous lean burn engine, the overshoot can be reduced to 0.2 AFR unit. At the same time, the combination of calibration and iterative leaning control scheme is practicable to control the AFR step process of a lean burn engine with the NOx storage-reduction catalyst. The results also show that the electronic control system has reliable performances and can completely fulfill the requirements of the project.