摘要
为了有效解决飞机突发故障诊断的时效性和准确性问题,提出了一种基于自适应神经网络的诊断方案。根据神经网络的适应度决定引入遗传算法优化网络的时机,提取出网络训练陷入饱和状态无法继续收敛时的权值与阈值,编码为染色体加入到种群中,用遗传寻优操作优化网络,达到快速准确诊断故障的目的。建立了提出的方案模型、BP神经网络模型和遗传算法优化神经网络常用模型,用MATLAB语言结合飞机突发故障的航班数据对3种模型进行了对比实验。实验结果表明,该方案较之常用方法具有用时较少且故障类型识别率较高的优点,能够满足诊断的时效性和准确性要求。
To effectively solve the aircraft burst fault diagnosis timeliness and accuracy issues, a new adaptive neural network- based diagnostic scheme is proposed. The opportunity to introduce genetic algorithm to optimize the network decided hy the fit- ness of neural network, the weights and threshold of network when training can no longer continue to converge into a saturated state, coding the weights and threshold added to the population of chromosomes, genetic optimization operations care used to op- timize network and to reach fast and accurate fault diagnosis purposes. A new program model, common neural network model and genetic algorithm optimization neural network commonly used model is established Comparison test on three models by using MATLAB language and typical sample database of aircraft burst failure. The results show that the scheme compared with the commonly used method has less time cost and higher fault type identification rate, which can meet the timeliness and accuracy of diagnostic requirements.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第3期1010-1015,共6页
Computer Engineering and Design
基金
国家自然科学基金与中国民航联合基金项目(61179063)
