针对现有主动悬架在应用最优控制时缺乏路面扰动识别内容的问题，提出一种识别路面扰动反馈的最优控制器。该控制器在传统系统状态反馈最优控制的基础上引入扰动反馈项，并通过粒子群算法优化加权系数，同时采用直线电机作为作动器。考虑到路面不平度与系统状态响应获取存在先后顺序，采用开环带有外部输入的非线性自回归(Nonlinear Auto-regressive Model with Exogenous Inputs，NARX)神经网络预测与逆模型相结合的方法来识别路面不平度。神经网络离线训练在线识别，识别模块实时将结果传输给控制器。在整车模型上对控制策略进行仿真。结果表明，粒子群优化使平顺性指标显著改善;采用的路面识别方法可有效提高识别的精确性;与不识别扰动控制相比，本策略可有效降低悬架动挠度的恶化，并改善整体控制效果。
Optimal Control Strategy with Identification of Road Disturbance Feedback for Active Suspensions
To address the problem that the existing active suspensions lack road disturbance identification when apply-ing optimal control,an optimal controller with identifying road disturbance feedback was proposed.The disturbance feed-back term was introduced into the controller based on traditional system state feedback optimal control.The weighting coeffi-cient was optimized using the particle swarm optimization algorithm,and a linear motor was used as the actuator.Consider-ing the order of road roughness and system state response acquisition,the combination of open-loop nonlinear auto-regres-sive model with exogenous inputs (NARX) neural network prediction and inverse model was used to identify the road rough-ness.The neural network was trained offline for online recognition,and the results were transmitted to the controller in real-time through the recognition module.The control strategy was simulated on the whole vehicle model.It is shown by the re-sults that the ride comfort index is significantly improved through particle swarm optimization.The accuracy of recognition is effectively enhanced by the proposed road surface recognition method.Compared with the control method without distur-bance recognition,this strategy can effectively reduce the deterioration of the suspension dynamic deflection and improve the overall control effectiveness.
vibration and waveactive suspensionoptimal controlparticle swarm algorithmroad roughness identificationNARX neural networkinverse model
万方数据
维普
