A Statistical Approach to Spark Advance Mapping

Engine performance and efficiency are largely influenced by combustion phasing. Operating conditions and control settings influence the combustion development over the crankshaft angle; the most effective control parameter used by electronic control units to optimize the combustion process for spark ignition engines is spark advance (SA). SA mapping is a time-consuming process, usually carried out with the engine running in steady state on the test bench, changing SA values while monitoring brake mean effective pressure, indicated mean effective pressure (IMEP), and brake specific fuel consumption (BSFC). Mean values of IMEP and BSFC for a test carried out with a given SA setting are considered as the parameters to optimize. However, the effect of SA on IMEP and BSFC is not deterministic, due to the cycle-to-cycle variation; the analysis of mean values requires many engine cycles to be significant of the performance obtained with the given control setting. Finally, other elements such as engine or components aging, and disturbances like air-to-fuel ratio or air, water, and oil temperature variations could affect the tests results; this facet can be very significant for racing engine testing. This paper presents a novel approach to SA mapping with the objective of improving the performance analysis robustness while reducing the test time. The methodology is based on the observation that, for a given running condition, IMEP can be considered a function of the combustion phasing, represented by the 50% mass fraction burned (MFB50) parameter. Due to cycle-to-cycle variation, many different MFB50 and IMEP values are obtained during a steady state test carried out with constant SA. While MFB50 and IMEP absolute values are influenced by disturbance factors, the relationship between them holds, and it can be synthesized by means of the angular coefficient of the tangent line to the MFB50-IMEP distribution. The angular coefficient variations as a function of SA can be used to feed a SA controller, able to maintain the optimal combustion phasing. Similarly, knock detection is approached by evaluating two indexes; the distribution of a typical knock-sensitive parameter (maximum amplitude of pressure oscillations) is related to that of $CHRNET$ (net cumulative heat release), determining a robust knock index. A knock limiter controller can then be added in order to restrict the SA range to safe values. The methodology can be implemented in real time combustion controllers; the algorithms have been applied offline to sampled data, showing the feasibility of fast and robust automatic mapping procedures.