针对故障诊断规则属性的复杂性,将粗糙集核属性参数的先验知识作为免疫疫苗引入抗体编码,以提高其在线学习及优化能力。将属性的分类近似质量作为适应度(目标)函数进行优化,在获取多个属性约简的同时,仍能快速地求得最小约简。把优化后的属性集作为支持向量机分类器的输入参数对故障样本进行训练与分类,提高了SVM故障诊断的能力。通过实验表明,该方法能快速、高效地对属性集进行数据压缩,有效提高了故障诊断准确率。
In view of the complexity of fault diagnosis attribute, the prior knowledge of core attributes in rough sets is introduced into antibody coding as immune vaccine to enhance its ability of learning and optimization on line. The classification approximation quality is taken as the fitness (objective) function to be optimized, so the minimum reduction can be quickly obtained while more reductions in attribute sets are acquired. The optimized attribute sets are then input into SVM classifier to train and classify the fault sample. Experimental results show that the approach is efficient and quick in data compression of attribute sets, and is effective in improving the precision of fault diagnosis.