摘要
针对喷嘴工程检测中存在的问题:同一喷嘴在两台设备中性能检测结果不一致,分析两台设备的工作原理、管道结构,建立燃油管道模型,找出检测结果差异性存在的原因。在此基础上,选用合适的神经网络结构,利用模拟退火遗传算法设定网络初始阈值和权值,利用贝叶斯正则化算法训练网络,得到不同检测结果的映射关系,通过映射使两者得到统一。仿真实验证明,此方法既将检测结果差异控制在了允许范围之内,又简单易行,提高了检测效率。
According to the question in the nozzle detection that the detection result by two equipments to the same nozzle is different,working principle and pipe structure of the two detecting equipments were analyzed.The pipe model was built.On the base of these,it was proposed that finding the difficult function between the different detection results by suitable neural network.The initialized weights and thresholds were optimized with genetic and simulated annealing algorithm,and the network was trained by Bayesian regularization algorithm.The two were unified by the function.Simulation result shows that the method can keep the difference in allowable scope and improve the detecting efficiency,furthermore,it is easy and feasible.
出处
《计算机技术与发展》
2011年第7期194-198,共5页
Computer Technology and Development
基金
西安市科技计划项目(CXY09012-1)
关键词
燃油喷嘴
性能检测
管道模型
神经网络
fuel nozzle
performance detection
pipe model
neural network
