机载武器中光电目标探测器故障诊断研究  

Fault Diagnosis of Optical-Electronic Target Detector in Airborne Weapons

作　　者：邓俊[1] 周越文[1] 宋海方[1] 葛亚维 

机构地区：[1]空军工程大学航空航天工程学院,陕西西安710038

出　　处：《计算机仿真》2015年第11期445-449,共5页Computer Simulation

摘　　要：在机载武器中,光电目标探测器的故障诊断对机载武器系统作战效能的提高尤为重要。由于提供的测试信号有限,同时信号之间存在非线性关系,导致确定故障的位置和原因非常困难,诊断准确率不高。而传统故障诊断方法存在局限性,诊断时间长。针对以上问题,提出了一种结构优化模拟退火粒子群神经网络方法。首先,将光电目标探测器的采集数据预处理后训练网络,利用模拟退火粒子群算法调整权值。然后,根据神经元间的相对重要度和等效连接关系,优化连接结构,再引入神经元增益建立准则优化隐节点数。最后,利用训练后的神经网络完成光电目标探测器的故障诊断。实验结果表明,新方法能够有效地提高诊断准确率。It＇ s very important for the improvement of airborne weapons systems operational effectiveness. Due to the limitation of provided test signals and the nonlinear relationship among the signals, it＇ s very difficult to confirm fault positions and reasons, which leads low diagnosis accuracy. Traditional methods have some limitations, which leads long diagnosis time. Aiming at the above problems, a structure optimization simulated annealing particle swarm optimization neural network algorithm is proposed. After preprocessed, the sample data from optical-electronic detection equipment test are applied to training the network. Firstly, network weights are adjusted by simulated annealing particle swarm optimization. Then, according to the relative importance and equivalent relation among the neurons, networks＇ connection structure is optimized, and the hidden layer nodes are optimized by introducing neural gain to build the rule. Finally, the trained network is applied in optical-electronic detection fault diagnosis. The experiment results show that the new method can effectively improve fault diagnosis accuracy.

关 键 词：结构优化 模拟退火粒子群算法 神经网络 光电目标探测 故障诊断 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]