AUTOMATED ENGINE CALIBRATION OPTIMIZATION USING ONLINE EXTREMUM SEEKING

Emissions compliance during engine start-up conditions is a major obstacle for current automotive manufacturers across global markets. The challenges to meeting emissions targets are both due to increasingly stringent regulations and the difficulty in developing control strategies for a high degree-of freedom and highly non-linear system. Online extremum-seeking (ES) methods offer a promising alternative to traditional optimization based on design-of-experiment based automotive calibration. With extremum-seeking methods, results from all prior experiments are used to intelligently and efficiently generate the next iteration of the control parameter(s). In this work, the applicability of the online extremum-seeking method is explored to optimize five performance variables (injection timing for two injection events, the injection fuel mass divided between the first and second injection events, air-fuel equivalence ratio and exhaust cam timing) to minimize brake specific fuel consumption while imposing different constraints on NOx emissions. The experiments were conducted using a production turbocharged four-cylinder gasoline engine with an advanced fuel injection system. The results show the utility of the ES strategy to quickly identify optimal control parameter combinations and the emissions and engine performance improvements during the calibration process. The results also demonstrate the dramatic benefit of the ES calibration strategy in terms of test time required.