在对粗糙集理论和遗传算法研究的基础上,提出一种基于知识依赖度为启发信息的改进自适应遗传约简算法,并将其应用于雷达故障诊断.在该算法中,对随机产生的二进制初始种群用属性核加以限制,在适应度函数中引入了决策属性对条件属性的依赖度,对交叉概率和变异概率进行了新的设计,并且对产生的新一代个体增加修正校验算子.利用该算法对雷达故障进行诊断,获取简单而又能体现故障征兆与故障原因对应的诊断规则,避免了传统基于故障树的专家故障诊断系统准确性差、效率低的缺点. Based on the research of rough set theory and genetic algorithm, an improved adaptive genetic reduction algorithm based on knowledge dependence is proposed and applied to radar fault diagnosis.In this algorithm, Binary initial population is restricted by attributes kernel. The dependence degree of decision attributes on conditional attributes is introduced into fitness function, and new design of crossover probability and mutation probability is introduced, and a new generation of modified calibration operator The algorithm is used to diagnose the radar fault and obtain the diagnostic rules which are simple and can reflect the fault symptom and the fault reason, thus avoiding the shortcomings of the traditional fault diagnosis system based on the fault tree, such as poor accuracy and low efficiency.