基于事件触发的汽车主动悬架振动控制器设计  

Design of Event Trigger-based Vibration Control for Active Suspension System

作　　者：庞辉[1] 王明祥 王磊[1] 郑理哲 PANG Hui;WANG Mingxiang;WANG Lei;ZHENG Lizhe(School of Mechanical and Precision Instrument Engineering,Xi’an University of Technology,Xi’an 710048,Shaanxi,China)

机构地区：[1]西安理工大学机械与精密仪器工程学院,陕西西安710048

出　　处：《兵工学报》2024年第8期2698-2711,共14页Acta Armamentarii

基　　金：陕西省自然科学基础研究计划重点项目(2023-JC-ZD-31)。

摘　　要：为提升车辆的行驶平稳性和安全性,针对具有输入死区和饱和特性的汽车主动悬架系统,设计一种基于事件触发(Event Trigger,ET)和长短期记忆(Long Short-Term Memory,LSTM)神经网络的智能振动控制器。在构建四分之一车辆主动悬架模型基础上,提出一种高效的ET控制器,能够有效缓解通信紧张问题和避免控制器固有的芝诺现象,提高控制器的稳定性和可靠性。为了进一步增强控制器的智能性和自适应性引入了LSTM神经网络,利用径向基函数神经网络来模拟和生成不同条件下的响应数据实现LSTM神经网络的训练,以精确预测并补偿作动器的死区和饱和影响,进而使主动悬架系统的垂直加速度趋近于0 m/s^(2),从而极大地提升了乘坐舒适性。通过数值仿真验证所设计控制器的适用性和有效性。研究结果表明,所设计的控制器在多种工况下能够有效提高主动悬架系统的动态性能。To improve the ride smoothness and safety of vehicles,an intelligent vibration controller based on event trigger(ET)and long short-term memory(LSTM)neural network is devised for automotive active suspension systems characterized by input dead zone and saturation.On the basis of building a quarter-car active suspension model,an appropriate ET controller is proposed,which effectively mitigates the communication bottlenecks and avoids the inherent Zeno phenomenon of controllers,thus enhancing their stability and reliability.A LSTM neural network is introduced to further improve the intelligence and adaptability of controller.A radial basis function neural network is utilized to simulate and generate the required response data for training LSTM neural network and compensate for input dead zone and saturation,making the vertical acceleration of the active suspension system get closer to 0 m/s^(2)and thus improve the vehicle ride comfort.The applicability and effectiveness of the designed controller are verified by numerical simulation.The research findings indicate that the controller can effectively enhance the dynamic performance of active suspension system under diverse operating conditions.

关 键 词：主动悬架系统 事件触发 长短期记忆网络 径向基函数神经网络 

分 类 号：U461.4[机械工程—车辆工程]