中文摘要：

介绍了一种新的轴系扭振优化手段。利用有限元结合多体动力学的方法，对某X8170ZC型柴油机轴系扭振情况进行研究。建立了轴系扭振仿真虚拟样机，其中主轴颈与主轴承、连杆大头与曲柄销采用非线性连接单元，使轴系扭振仿真更接近实际情况。其次，建立了扭振减振器优化目标函数，引入BP（Back Propagation）神经网络对减振器进行参数优化设计。其中，BP神经网络的学习过程采用仿真提供的数据。优化结果表明，利用该方法进行轴系扭振优化设计切实可行，降低了轴系工作转速范围内的共振振幅，由优化前的0．14°下降到0．08°。

英文摘要：

A new method to optimize the torsional vibration of diesel engine crankshaft is introduced. The torsional vibration level of X8170ZC diesel engine crankshaft is investigated based on the finite element method and the muhi-body dynamics method. The virtual prototype of the crankshaft is given and the nonlinear units are used to connect main journal with bearing,and crank pin with big end of connecting-rod, making the model more simulative to the real condition. The optimum design objective function of torsional damper is founded and it is solved by using the back propagation neural network（BPNN）. Simulation results are provided to the learning process of BPNN. The result shows that this method is feasible to reduce torsional vibration. The amplitude of resonance vibration is decreased from 0.14° to 0.08°.