摘要
为减少传统方法中利用单个参数对武器火控系统进行故障诊断的不确定性,采用粗糙集、BP神经网络和D-S证据理论相结合的多源信息融合方法 ,建立了武器火控系统故障诊断的模型。首先从原始数据中提取不同故障的特征信息,并将其作为诊断对象状态的证据;然后利用粗糙集对故障特征进行属性约简和值约简;建立BP神经网络,构造学习样本,通过离线学习训练得到各个证据体对于识别框架中各种故障的基本概率分配;最后利用D-S合成法则和诊断决策规则对融合信度区间进行决策分析,得到最终的诊断结果。该故障诊断模型的可靠性降额准确性远大于利用单个特征域进行诊断的局部诊断结果 ,提高了武器火控系统故障诊断的精度。
In order to reduce the traditional method of using a single parameter uncertainty of weapon fire control system fault diagnosis, the weapon fire control system fault diagnosis model is established by the multi-source information fusion method of combining the rough sets and BP neural network and D-S evidence theory. First of all, from the original data the characteristic information of the different fault is extracted as a diagnosis of the state of the object evidence; Then the fault feature is disposed by rough set attribute reduction and value reduction. The BP neural network is established, the learning samples are constructed, and the basic probability distribution of the evidence for identifying all kinds of faults in framework is obtained by offline training; The final diagnosis results are obtained by analyzing the convergence reliability interval in combination with D-S synthesis and diagnosis decision rules. The reliability of fault diagnosis model derating accuracy is higher than using a single feature domain localized diagnosis, the precision of weapon fire control system fault diagnosis is improved.
出处
《自动化技术与应用》
2016年第10期9-13,共5页
Techniques of Automation and Applications
