Multiplexed extremum seeking for calibration of spark timing in a CNG-fuelled engine

The compositional variability of many alternative fuels, coupled with fuel agnostic behaviour like engine ageing and vehicle-to-vehicle differences, leads to the desire for some form of online calibration in order to optimise fuel efficiency. This has led to the incorporation of extremum seeking techniques within the field in order to continually fine tune engine performance. These typically address steady state engine performance and are characterised by slow convergence times, hindering their deployment in typical dynamic driving scenarios. To address this potential shortcoming, in this paper a novel multiplexed extremum seeking scheme is proposed to track a time-varying extremum caused by a measurable disturbance. It consists of multiple extremum seeking agents that are individually activated based on the disturbance. The multiplexed approach accommodates the rigorous practical stability results of the “traditional” extremum seeking approaches, but offers improved results in dynamic scenarios. The proposed approach is implemented both in simulation and experimentally on a compressed natural gas (CNG) engine operating over a drive cycle. The experimental results show that under proper tuning, the proposed controller can improve the engine fuel efficiency for unknown natural gas compositions without requiring gas composition sensing at little additional calibration effort.