Combustion parameters estimation based on multi-channel vibration acceleration signals

Highlights: • Multi-channel vibration signals are tested on a two-cylinder diesel engine. • MHD is introduced to compute the weighted coefficient of each sensor. • The weighted coefficient is used to reconstruct useful vibration signal. • The reconstructed vibration signal is denoised with a low-pass filter. • Combustion parameters are identified from the denoised vibration signal. -- Abstract: The vibration signals are often used to extract combustion information. And the vibration sensor is generally affixed at the location where it is sensitive to the combustion information. When the vibration sensor can't be mounted at the most sensitive location or the vibration sensor failure occurs, it is hard to extract enough combustion information from the vibration signal. A new methodology is presented to reconstruct the signal closely related to the combustion information based on multi-channel vibration acceleration signals. The modified Hausdorff distance is introduced to describe the similarity and to compute the weighted coefficient among the collected vibration acceleration signals. The analyzed results indicate that at least three sensors are needed to reconstruct vibration acceleration signal closely related to the combustion information. The reconstructed signal is firstly denoised and then is used to identify combustion parameters such as the start of combustion, the crank angle at which 50% of the fuel mass fraction has burned (MFB50) and the end of combustion (MFB95). Test results show that the maximum error between the predicted value and that obtained from the cumulated heat release (CHR) is to be within 1.0 deg, 1.2 deg and 1.0 deg, respectively.