摘要
为了实现列车车轮踏面损伤识别,提出了一种基于GA-RBFNN算法的货车车轮踏面损伤识别方法。该算法采用浮点数编码将RBFNN的中心参数和宽度进行了编码,利用GA的选择、交叉和变异操作优化网络参数,权值采用最小二乘法确定。利用该算法和BP算法、传统的RBFNN算法进行了剥离和擦伤识别的对比实验,结果表明:GA-RBFNN算法对剥离、擦伤和非损伤三类样本的测试集的识别率高于传统的RBFNN算法和BP算法,而且GA-RBFNN算法的进化代数远远小于BP算法和传统的RBFNN算法迭代次数。
In order to the recognition of the train wheel tread damages, the pattern recognition method of the train wheel tread damages is developed. The algorithm uses float encoding to encode learning parameters of network, establishes fitness function, optimizes learn- ing parameters by using the operation of selection, crossover, mutation. Compared with the traditional RBFNN and BP, the experimental results show that the recognition rate of testing samples is higher than traditional RBFNN and BP, the evolutional generations of GA-RBFNN algorithm are less than recursive times of traditional RBFNN and BP.
出处
《计算机工程与应用》
CSCD
2012年第8期32-34,共3页
Computer Engineering and Applications
基金
陕西省自然科学基础研究计划资助项目(No.2011JQ8013)
