Research Of Diesel Engine Cylinder Pressure Reconstruction Methods Based On Cyclostationary Properties

Cylinder pressure is one of important index for characterizing diesel operation states. Therefore under the situation of being unable to gain cylinder pressure directly from the minitype high-speed diesel engine, indirect identification of cylinder pressure by the cylinder heads' vibration respond signals has become main research direction of diesel engine fault diagnosis.The author carried out studies on the issue of identifing diesel engine cylinder pressure by cylinder head surface vibration signals,Study content is as follows:(1)The paper has studied generation mechanism of 4135 non-supercharged four-cylinder four-stroke diesel engine surface vibration,showed the main diesel engine vibration excitation source and their timing, and explain the intrinsic relationship between the excitation sources and the engine surface vibration signals; introduced cyclostationary theory, pointed out cyclostationary properties of cylinder heads vibration signals, establishmented a statistical model of the vibration signals ,given the reason why the sampling mothod based on the model must be equal time intervals.(2)The paper has studied a RBF neural network identification methods- based on non hypothetical. After processing the vibration signals and the cylinder pressure signals respectively by a synchronous average method, treated them as the network inputs and outputs, trained neural networks, and establishmented the nonlinear mapping relations between cylinder pressure signals and vibration signals , and used this relationship to identify the cylinder pressure . The results show that the identification curves of the pressure and the measured pressure curves are almost a coincidence ,that the neural network method has not only high recognition accuracy, but also the advantages of small errors, and good adaptablity to the working condition.At the same time,thepaper pointed out that synchronization average of the signals effectively enhance the signal to noise ratio.(3)The paper has study a transfer function identification method based on the linear assumption. Through pre-processing, low-pass filtering and adaptive cancellation processing, removed the interferences from the other driving forces and random noise, and then set up the transfer function between the vibrations and the cylinder pressure signals, and use the transfer function to identify the cylinder pressure The results showed that the identified pressure is close to the measured pressure,just with a small error; By comparison , the adaptive cancellation was verified that it could effectively highlight the outbreak of vibration response signals, remove a large number of interferences and noise, improved the accuracy of transfer function model; transfer function model can be intuitively judged through mapping the coherence function curves of the model inputs and outputs.