改进的决策树生成算法在电路板故障诊断中的应用

Improved decision tree generation algorithms and its application to fault diagnosis of the printed circuit board

用决策表表示印刷电路板的测试诊断知识．针对传统的决策树生成算法之不足，通过分析决策树分枝的生成情况，提出了改进算法１；用试验子序列代替传统方法的单个试验，提出了改进算法２．实例说明，改进算法具有更好的优化效果，且证明了传统算法是改进算法２ 的特例．将改进算法应用到印刷电路板测试诊断中，使故障诊断的平均测试开销最小，并构造了电路板故障诊断专家系统．

Decision tables are used to present the knowledge of PCBs. In view of existing issues in the traditional algorithm for decision tree generation, two improved algorithms are proposed by taking into account the following state of tree, and replacing the single test with a sub sequence of tests. The efficiency of improved algorithms is virified. Moreover, the traditional algorithm is proved to be a special case of improved algorithm Ⅱ. The improved algorithms are applied to the test and diagnosis of PCBs, and the diagnosis cost is minimized. Finally, the frame of a test & diagnosis expert system for PCBs is constructed.