【作者】 张丽霞；

【导师】 赵又群；

【作者基本信息】 南京航空航天大学 ， 车辆工程， 2007， 博士

【摘要】 汽车操纵动力学特性影响汽车驾驶的操纵轻便程度,也是高速汽车安全行驶的一个主要性能。汽车在高速转向行驶时,驾驶员模型参数不易确定,从而导致驾驶员模型建立困难。为了避开驾驶员建模困难问题,本文运用操纵逆动力学方法,反求出驾驶员的操纵输入。针对目前路面不平度研究方法的局限性,本文运用逆问题的思想研究了路面不平度的识别方法,为汽车平顺性分析研究奠定了基础,为路面性能的分析提供了依据。(1)提出了一种基于径向基函数神经网络在频域范围内识别路面不平度的仿真研究方法。该方法以4自由度和7自由度汽车振动模型为基础,以Matlab软件仿真得到的汽车车身质心垂直加速度和俯仰角加速度作为神经网络理想输入样本,拟合的路面不平度为神经网络理想输出样本,应用RBF神经网络建立汽车车身质心垂直加速度、俯仰角加速度和路面不平度之间的非线性映射模型。另取一组仿真得到的车身质心垂直加速度和俯仰角加速度代入已训练好的网络进行路面不平度的识别。结果表明,该方法具有较强的抗噪声能力和较理想的识别精度,识别的路面不平度与拟合的路面不平度吻合较好。(2)提出了一种基于径向基函数神经网络在时域范围内识别路面不平度的仿真研究方法。在时域范围内建立RBF神经网络非线性映射模型,另取一组仿真得到的车身质心垂直加速度和俯仰角加速度代入已训练好的网络识别路面不平度。最后以ADAMS/View虚拟试验仿真得到的车身质心垂直加速度和俯仰角加速度来识别路面不平度,从而验证了运用RBF神经网络识别路面不平度的有效性。(3)提出了在不同汽车跟踪同一指定路径的情况下,汽车操纵逆动力学角输入识别和力输入识别的仿真研究方法。该方法以线性2自由度汽车方向盘转角输入和线性3自由度汽车方向盘转矩输入为数学模型,运用最优控制理论识别方向盘转角输入和方向盘转矩输入。用直接配置方法将最优控制问题转化为非线性规划问题,用序列二次规划方法对转化后的非线性规划问题进行求解。结果表明:利用该方法计算出来的路径跟踪性良好,且可以比较跟踪同一路径的不同汽车的操纵性能,且仿真结果与实车试验结果运动趋势相似。(4)提出了一种解决汽车最速操纵问题的仿真研究方法。该方法基于最优控制理论,以驾驶员对汽车施加的转角输入和驱动力/制动力为控制变量,以最短时间完成双移线和蛇行线过程为控制目标。通过直接配置方法将最优控制问题转化为非线性规划问题之后,运用序列二次规划方法求解。采用该方法计算了不同汽车在同样给定路径边界时的Matlab仿真结果。结果表明,该方法能够解决汽车的最速操纵问题,可以比较不同汽车以最短时间完成双移线和蛇行线过程的操纵性能,且仿真结果与ADAMS/Car虚拟样机试验结果具有良好的一致性。更多还原

【Abstract】 The character of vehicle handling dynamics affects vehicle handling handiness, and is a main performance index determining safe running for high speed vehicle. In the case of vehicle high speed turning, driver model parameters are not easy to determine, which makes driver model difficult to bulid. In order to avoid the problem, driver handling input is calculated by the handling inverse dynamics method.Aiming at the limitation of current research method for road surface roughness, the road surface roughness identification method is studied by the thinking of inverse problem. It establishs the foundation for vehicle ride comfort , and provides the basis for the anlysis of road surface performance.(1)Based on RBF neural networks, a simulation research method of road surface roughness identification in the field of frequency is put forward. Based on four degree-of-freedom and seven degree-of-freedom vehicle vibration model, the vehicle body centroid vertical acceleration and pitching angular acceleration which are got through Matlab simulation are regarded as neural networks ideal input sample, the imitated road surface roughness is regarded as neural networks ideal output sample. The nonlinear mapping relations among vehicle body centroid vertical acceleration, pitching angular acceleration and the road surface roughness are found by RBF neural networks. Another vehicle body centroid vertical acceleration and pitching angular acceleration which are calculated by simulation are used to identify road surface roughness by trained networks. Simulated results show that the method has better ability of anti-noise and ideal identification accuracy, the road surface roughness of identification fits the imitated road surface roughness.(2)Based on RBF neural networks, a simulation research method of road surface roughness identification in the field of time is put forward. The nonlinear mapping relations in the field of time are found by RBF neural networks . Another vehicle body centroid vertical acceleration and pitching angular acceleration which are calculated by simulation are used to identify road surface roughness by trained networks. Finally another vehicle body centroid vertical acceleration and pitching angular acceleration which are got by ADAMS/View virtual experment simulation are used to identify road surface roughness. So the availability of road surface roughness identification by RBF neural networks is validated .(3)A simulation research method for identifying the angle input and the force input in vehicle handling inverse dynamics is proposed under the condition of different vehicles tracking the same given path. In this method, the linear vehicle model of steering angle input with two degree-of-freedom and steering momemt input with three degree-of-freedom are adopted, and the optimal control theory is used to identify the steering angle input and the steering moment input. By using the direct parallel method, the optimal control problem is converted into a nonlinear programming problem that is then solved by means of the sequential quadratic programming. Simulated results show that the proposed method is of good path-tracking ability; is able to compare the maneuverability of different vehicles that track the same path; and the movement trend of simulated results is similar to that of actually experiment.(4)A simulation research method for solving vehicle minimum time maneuver problem is proposed. Based on optimal control theory, steering angle input and traction/brake force imposed by driver are control variables, the minimum time required to complete the double lane change and slalom is control object. By using the direct parallel method, the optimal control problem is converted into a nonlinear programming problem that is then solved by means of the sequential quadratic programming. Matlab simulation results are obtained for two different vehicles performing similar given path boundary by the method. Simulated results show that the proposed method can solve vehicle minimum time maneuver problem; compare maneuverability of two different vehicles that complete double lane change and slalom with the minimum time; and the results fit the results of ADAMS/Car virtual experment.更多还原