| Cylinder-piston assembly is the power source of a diesel engine. Though failure rate of the assembly isn't the highest, once it appears abnormal wear, the operation quality of the diesel engine is decreased than before, or even the diesel engine is destroyed. Thus, monitoring the working state of cylinder-piston assembly is very important for keeping diesel engine in a good operation condition.Based on the reasons above, a research on the identification of working state of the cylinder-piston assembly is carried out in this thesis. Firstly, the thesis analyzes the vibration signal of diesel engine surface and the crashing process between piston and cylinder, establishes the mapping relationship between them, and masters the influence factors that is the diesel engine models, operating conditions and so on. Secondly, considering that the vibration signal is non-stationary stochastic process, the thesis mainly uses wavelet packets to deal with the vibration signal, plots the three-dimensional time-frequency energy chart of the wavelet packets, and proposes using characteristic parameter of the main impact energy as an important basis to monitor cylinder-piston assembly working state. Thirdly, the thesis combines genetic algorithm and neural network, which will improve the network performance. On the basis of primarily optimizing the weight distribution with genetic algorithm, the thesis trains and recognizes the BP network, realizes the identification of the cylinder-piston working state assembly by means of the nonlinear mapping ability of neural network. Finally, the thesis validates the effectiveness of extracting characteristic parameters by experiments taken on cylinder inner parts of different machine types and different failure forms.The result shows that vibration analysis method can accurately identify the working state of cylinder-piston assembly without unseam machine.The method, which is simple and reliable, shows us a good prospect in the engineering application. |
PDF Full Text Request