基于多种群竞争进化的车辆超载网络检测  

Network detection for vehicle overload based on multi-population competition evolutionary

作　　者：申燚[1] 张雨[2] 袁明新[1] 王琪[1,3] 

机构地区：[1]江苏科技大学机电与汽车工程学院,江苏张家港215600 [2]南京工程学院车辆工程系,江苏南京211167 [3]江苏悦达专用车有限公司,江苏盐城224007

出　　处：《江苏科技大学学报（自然科学版）》2011年第4期364-368,共5页Journal of Jiangsu University of Science and Technology:Natural Science Edition

基　　金：江苏省自然科学基金资助项目(BK2008576);江苏科技大学博士科研启动基金资助项目(35271004);江苏科技大学张家港校区青年基金资助项目

摘　　要：针对车辆超载检测中复杂的数据处理,提出了基于BP(Back Program)网络的非线性检测模型.为了克服神经网络在梯度学习中存在的收敛速度慢,容易陷入局部极小等不足,将多种群竞争机制引入到免疫进化中,提出了一种新的基于实数编码的多种群竞争免疫算法,并基于车辆轴重检测数据实现了BP网络的权值和阈值优化.实验结果验证了该方法能有效地降低因系统非线性和时变性所导致的测量误差,提高了测量精度.To solve the complicated data processing in the vehicle overload detection,a nonlinear system modeling based on BP（Back Program） network is presented.In order to overcome the shortages of gradient-based study in neural network such as slow convergence speed,easily trapping into local minimum,the multi-population competition mechanism is introduced in the immune evolutionary,and a new real-coded multi-population competition immune algorithm（RMCIA） is presented.Furthermore,the weights and threshold values of BP network are optimized based on the RMCIA through the real detecting data.The experimental results prove the effectiveness of proposed method.The method can reduce the measure error for nonlinearity and time varying of system and improve the detecting precision.

关 键 词：多种群竞争 免疫算法 BP网络 超载 

分 类 号：U467.5[机械工程—车辆工程]