免疫BP网络的机载嵌入式训练系统效能评估  

Effectiveness Evaluation of Airborne Embedded Training System of Immune BP Networks

作　　者：邓晓政 叶冰[1] DENG Xiao-zheng;YE Bing(Chinese Flight Test Establishment,Xi’an 710089,China)

机构地区：[1]中国飞行试验研究院

出　　处：《计算机技术与发展》2019年第12期173-177,共5页Computer Technology and Development

基　　金：中国航空工业联合基金项目(6141B05110101a)

摘　　要：针对如何定量、精准地对机载嵌入式训练系统进行效能评估的工程问题,提出了一种基于免疫BP神经网络的效能评估方法。首先将效能评估工程问题建模为一个非线性回归的数学问题;其次,根据机载嵌入式训练系统的组成结构和特点,设计了简洁、实用的效能评估指标体系;再次,综合利用了免疫克隆选择优化算法全局搜索能力强的优势以及BP神经网络算法局部搜索能力强的优点,从而快速有效地求解神经网络的突触权值,进而得到训练好的神经网络。最后在算法验证部分,通过四组仿真数据实验,并对比经典的BP神经网络算法、基于进化计算的BP神经网络算法,结果表明该效能评估方法在评估精度和评估稳定性方面都是较优的。Aiming at the engineering problem of how to quantitatively and accurately evaluate the effectiveness of the airborne embedded training system,we propose a novel effectiveness evaluation method based on immune BP neural networks.Firstly,the problem of effectiveness evaluation engineering is modeled as a nonlinear regression problem.Secondly,according to the composition and characteristics of the airborne embedded training system,a concise and practical effectiveness evaluation index system is designed.Thirdly,the advantages of immune clonal selection optimization algorithm in global search and BP neural network algorithm in local search are comprehensively utilized,so that the weights of the neural network can be solved quickly and effectively,and the trained neural network can be obtained.Finally in the algorithm verification,through four groups of simulation experiments and the comparison with classic BP neural network and BP neural network based on evolutionary computation,it shows that this method is superior in evaluation accuracy and stability.

关 键 词：嵌入式训练系统 效能评估 BP神经网络 免疫克隆选择 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]