【摘要】 提出了一种基于特征集第一主元方向及其信息保留率的状态识别方法。首先计算得到多路信号的特征集,继而计算该特征集的第一主元方向在特征集所在的高维空间中的方向角向量,并以该方向角向量和第一主元方向上的信息保留率作为状态识别的新特征集。提出了一种较为宽泛的神经网络收敛准则和判据方法,该方法中网络输出节点数为1,并结合非对称交叉式遗传算法作为网络训练中寻优方法,可有效实现网络的快速训练。试验表明,所提方法可对高维多元数据进行高度抽象,且计算简便,识别准确率较高,泛化能力较强,具有一定的应用价值。更多还原

【Abstract】 Based on the first principal direction and information remain rate,a condition identification method is proposed. The features set of multi-channel signals was calculated and the direction angel vector of the first principal direction was obtained. A new feature set was built by direction angel vector and the information remain rate. Then,a convergence rule and judging procedure of neural works were further proposed combining the anisomerous crossover genetic algorithm,and the training speed of the neural works was increased. The results show that the proposed method can extract the feature from the high dimension multivariable features-set,the calculation process is simple,and the accuracy of identification and the generalization capability of the method are remarkable.更多还原