A review of model-based offline and online optimization control systems for combustion engines

Abstract

To meet the development requirements of ultra-high efficiency, ultra-low emissions, and strong working condition adaptability for internal combustion engines, the collaborative optimization of engine performance and emissions has become an urgent problem to be solved. Nevertheless, several performance indicators of engines are mutually restricted, and there is a complex coupling relationship between controllable variables and performance parameters. All these make it difficult to rely on experience to achieve the multi-objective performance optimization for engines, promoting the research on model-based multi-objective engine performance optimization. Since the effectiveness of optimization relies on fast and accurate predictive models, the advantages of machine learning (ML) algorithms have been fully exploited. In this paper, the literature on engine performance prediction models based on ML algorithms is reviewed. Moreover, the technical fields and significant achievements with regard to multi-objective engine performance optimization based on intelligent optimization algorithms are analyzed and summarized. The most commonly used intelligent algorithms in engine performance optimization are evaluated and compared in detail, which provides guidance for the analysis and selection of proper optimization strategies, modeling methods, and optimization algorithms. Finally, this paper discusses the research hotspots and future development direction in this field, in order to provide more ideas for the future research on engine multi-objective performance optimization.