摘要
为实现装甲车辆电源系统调压器智能化检测和隔离故障开展了测试性设计;首先建立了调压器仿真模型,对调节电路进行故障模式影响分析,得到所有故障模式,对两种工作状态分别构建其相关性模型并得到改进后的诊断树,在此基础上利用基于L-M改进学习算法的神经网络对模型进行网络训练和验证,并与其它学习算法进行了比较;从验证结果来看,该学习算法相对优越,且此方案能够正确检测和隔离调压器各故障模式,表明其正确性和可行性,实现了对其测试性设计的目的。
To realize the detection and isolation of these faults intelligently on the voltage regulator of armored vehicle power system, the testability research is performed. By building the voltage regulator electrocireuit simulation model and analyzing failure mode and effects, ac- quires all fault modes, designs the dependency matrix about the two kinds of work states, and then obtains fault tree model that improved, afterwards, uses the neural network based on L--M improved algorithm to train and validate the mode, which is compared with other algo- rithms. As a result, the algorithm work more perfect, and the method can detect and isolate exactly all kinds of failure modes, which indi- cates the validity and feasibility of this method, and achieve the goal of design for testability.
出处
《计算机测量与控制》
北大核心
2013年第3期577-579,共3页
Computer Measurement &Control
关键词
调压器
相关性模型
神经网络
测试性
voltage regulator
dependency matrix
neural network
testability
