摘要
在传统的防抱死系统中,由于没有额外的传感器测试更多的参数,以及路面的不平度、信号的干扰等原因,致使路面辨识非常困难.针对以上问题,通过对整车制动模型和 ABS 制动模型的研究,建立了带有偏差单元的递归神经网络路面辨识模型,此路面辨识模型是通过对标准信号进行反复学习记忆,用非线性映射方法进行特征抽取并存储于网络的各节点上.工作时,当输入信号存在变形或噪声时,该模型通过联想记忆对输入信号进行聚类分析,得出正确的判断.实验表明,该辨识过程是一种高层次的认知过程且便于批量生产、升级、更新和维护.
In traditional system of ABS,a discriminate of road surface is very difficult because of no more ex- tra-sensors for more parameters,undulate degree of road surface,interference of signals etc.Through study of whole vehicle brake model and ABS anti-brake model,a model for resolving these questions was built,it is a discriminate model of neural network with delivery and regression of deviation's residential unit.The model's theory is that extract characters with non-linear reflectance methods and save them in joints of network through continual studying for standard signal.When having signal's shape changed or noise interference in the work,the model can classify and analyze through reminding and recalling and give right result.Through tests,we know this model is benefit for batch produce,promote,renew and uphold of ABS.
出处
《中国工程机械学报》
2004年第1期92-95,104,共5页
Chinese Journal of Construction Machinery
基金
国家自然科学基金(59835050)
关键词
路面辨识
防抱死
车辆工程
神经网络
自动控制
数学模型
discriminate of road surface
anti-skid brake
vehicular engineering
neural network
automatic control
mathematics model
