Abstract: Aiming at the problem that it is difficult to determine the input control current by calculating the expected damping force of semi-active suspension with automotive magnetorheological damper, the forward model was established by adopting the generalized Spencer phenomenon model, firstly considering the obvious nonlinear hysteresis characteristics of the magnetorheological damper forward dynamics model. Combined with the BP neural network, the magnetorheological damper inverse model was established. Then the BP neural network with strong mapping ability was optimized by using the rapid convergence and overall optimization ability of particle swarm optimization algorithm to improve the accuracy of input control current. Finally, the optimized magneto-rheological damper inverse model was combined with the semi-active suspension control system for simulation verification. The simulation results show that: compared with the actual control current, the optimized magneto-rheological damper inverse model can calculate the input control current more accurately, and its relative error is greatly reduced; in addition, all the performance indexes of the optimized suspension system are improved.

Key words: vehicle engineering; magnetorheological damper; forward inverse model; BP neural network; optimization; semi-active suspension

摘要： 针对通过汽车磁流变阻尼器半主动悬架期望阻尼力反求输入控制电流难以确定的问题。首先考虑到磁流变阻尼器正向动力学模型存在明显的非线性滞回特性，采用通用性较强的Spencer现象模型建立正向模型，并结合BP神经网络建立磁流变阻尼器逆模型；然后利用粒子群算法的极速收敛整体寻优能力优化具有强映射能力的BP神经网络，以提高输入控制电流的准确性；最后基于优化后的磁流变阻尼器逆模型并结合半主动悬架控制系统进行了仿真验证。仿真结果表明：与实际的控制电流相比，优化后的磁流变阻尼器逆模型更能准确计算输入控制电流，其相对误差大幅降低；此外，经优化后悬架各项性能指标均得到了改善。

关键词: 车辆工程；磁流变阻尼器；正向逆模型；BP神经网络；优化；半主动悬架