Compensation Map Calibration of Engine Management Systems Using Least-Squares Support Vector Committee Machine and Evolutionary Optimization

Nowadays, automotive engine compensation control is carried out electronically by utilizing many compensation maps in engine management systems; such that the engine can sustain its performance under the variations in engine operating conditions and environmental parameters. In traditional engine compensation map calibration, the parameters are normally set by a trial and error method because the exact mathematical engine model has not been derived. In this paper, a new framework, namely multiinput/output least-squares support vector committee machine, is proposed to construct the engine compensation control system (ECCS) models based on experimental data. As the number of adjustable parameters involved in the ECCS is very huge, the model accuracy and training time are usually degraded. Nonlinear regression is therefore employed to perform dimension reduction before modelling. The ECCS models are then embedded in an objective function for parameter optimization. Two widely-used evolutionary optimization algorithms, Genetic algorithm (GA) and particle swarm optimization (PSO), are applied to the objective function to determine the optimal calibration maps automatically. Experimental results show that the proposed modelling and optimization framework is effective and PSO is superior to the GA in compensation map calibration.