Research On The Method Of Combustion Characteristic Observation With Acoustic Signal For CR Diesel Engine

At present, it is difficult to observe the combustion quality deteriorationoriginated from fuel quality, operating conditions and system ageing, simply byemploying the conventional open-loop combustion control. Stimulated by theincreasingly demand of low fuel consumption and emissions and the rapiddevelopment of alternative fuels, closed-loop combustion control is now regarded as apossible and necessary means to meet future regulations. However, limited by therelatively high cost, low duration and difficult implementation of the pressuretransducer, the ordinary pressure transducer-based observation method andclosed-loop control are still not widely applied in practice. Therefore, in this paper, acontrol-oriented acoustic-signal-based combustion process observation method wasdeveloped for common rail diesel engines.Since the acoustic are easily disturbed by noises, leading to low signal-noiseratios (SNR), in this research, wavelet transforms were applied in the time-frequencyanalysis of the acoustic obtained from a medium/heavy-duty common rail dieselengine equipped with an original ECU. Two frequency bandwidths (4-6KHz and1-1.5KHz) with high SNR and signal intensity were proposed as the characteristicbandwidths for combustion analysis.In this research, a method was developed for diagnosis of misfire and roughcombustion cycles on common rail diesel engines by using acoustic. Investigationswere carried out into the differences between the acoustic obtained from misfire,rough combustion and normal operating cycles within the time and time-frequencydomains. Characteristic variables of acoustic, sensitive to both normal and abnormalcombustion processes,were defined. Acoustic-signal-based models for misfire andrough combustion cycles diagnosis were developed by employing discriminatoryanalysis. The model test results show that correct identification rate for misfirediagnosis is99%and for rough combustion cycles diagnosis is over85%.A method for the observation of combustion timings was proposed. Aftertime-frequency analysis of the acoustic, appropriate frequency bandwidths forcombustion observation were selected, within which characteristic variables weredefined. Correlation analysis was conducted between the variables and the measuredtimings of the combustion onset and the peak in-cylinder pressure rise. Polynomialregression model was then employed to developing the observation models of the aforementioned combustion timings. For ordinary operating conditions, predictionswith errors around1°CA could be achieved.An embedded online implementation method for combustion characteristicobservation based on acoustic signals was proposed based on parallel processing ofthe hardware and software resources of the ECU. It takes only5.31ms to complete asingle combustion characteristic observation task on the controller developed in thisresearch for a common rail diesel engine. It is proved that the method can beimplemented in the ECU.