Frequency response function adaptation for reconstruction of combustion signature in a 9-L diesel engine

An accelerometer as a low-cost non-intrusive transducer for sensing the combustion events in a diesel engine was investigated via the reconstruction of in-cylinder pressure using an adapting frequency response function (FRF). As the noise introduced into the accelerometer signal and the response to combustion vary with the operating condition, the FRF computed from a single operating condition only works for the same or similar conditions. To overcome this limitation, an adaptation process for the FRF was explored. Robustness of FRF over additional operational conditions with start of injection, start of combustion, and load variations was greatly improved. Frequency domain analysis shows that only the low-frequency content is determinant for the in-cylinder pressure reconstruction, and the adaptation of the first and second (0 Hz and 121 Hz) harmonics of the FRF results in the greatest improvement for the in-cylinder pressure estimation accuracy. The 0 Hz harmonic is adjusted based on the pre-measured in-cylinder pressure offset and the online measured accelerometer signal offset. Particle swam optimization as a computational algorithm is applied to adapt the 121 Hz harmonic of FRF. The results show that the adapted FRF, in comparison to the unadapted FRF, can reduce the phase error up to 1.3 crank angle degrees and reduce the amplitude error by up to 90%.