Engine idle-speed system modelling and control optimization using artificial intelligence

This paper proposes a novel modelling and optimization approach for steady state and transient performance tune-up of an engine at idle speed. In terms of modelling, Latin hypercube sampling and multiple-input and multiple-output (MIMO) least-squares support vector machines (LS-SVMs) are proposed to build an engine idle-speed model based on experimental sample data. Then, a genetic algorithm (GA) and particle swarm optimization (PSO) are applied to obtain an optimal electronic control unit setting automatically, under various user-defined constraints. All of the above techniques mentioned are artificial intelligence techniques. To illustrate the advantages of the MIMO LS-SVM, a traditional multilayer feedforward neural network (MFN) is also applied to build the engine idle-speed model. The modelling accuracies of the MIMO LS-SVM and MFN are also compared. This study shows that the predicted results using the estimated model from the LS-SVM are in good agreement with the actual test results. Moreover, both the GA and PSO optimization results show an impressive improvement on idle-speed performance in a test engine. The optimization results also indicate that PSO is more efficient than the GA in an idle-speed control optimization problem based on the LS-SVM model. As the proposed methodology is generic, it can be applied to different engine modelling and control optimization problems.