基于思维进化神经网络的真空绝热板真空度测量精度改进方法研究  

Improved Pressure Measurement Precision for Vacuum Insulation Panel with Mind Evolutionary Algorithm

作　　者：黄俊[1] 冯勇建[1] 

机构地区：[1]厦门大学机电工程系,厦门361005

出　　处：《真空科学与技术学报》2015年第5期528-532,共5页Chinese Journal of Vacuum Science and Technology

基　　金：福建省重大科技计划项目(2010H6025)

摘　　要：以基于红外热成像机理测量真空绝热板(VIP)的真空度为研究对象。鉴于目前此方案测量真空度存在精度低的问题,本文提出一种基于BP神经网络的真空度测量精度改进方法。针对BP神经网络存在易陷入局部最优、收敛速度慢等缺陷,本文巧妙利用思维进化算法优化BP神经网络的初始权值和阈值,从而弥补以上缺陷。最终实验结果表明,利用思维进化算法优化BP神经网络创建的数学模型大大提高了真空绝热板真空度的测量精度,其实际测量精度优于2%。因此,本文所提方法具有广泛推广的应用价值。A novel technique was developed to improve the measurement precision of the initial pressure in vacuum insulation panels （VIPs） by infrared thermal imaging via the embedded sensors. The modified technique mainly involves back propagation （BP） neural networks to increase the measurement accuracy. To effectively tackle the problems of BP neural networks,including but not limited to the local optimum and slow convergence, the newly-developed solution in- cludes application of the mind evolutionary algorithm with optimized initial weights and thresholds of BP neural networks. The novel algorithm was tested in measuring the pressure in VIPs. The test results show that the novel technique outperforms the conventional one with a precision higher than 2%. We suggest that the new method,based on mind evolutionary algorithm,be of much technological interest in the pressure evaluation of VIPs.

关 键 词：真空绝热板 真空度 MATLAB BP神经网络 思维进化算法 

分 类 号：TP216[自动化与计算机技术—检测技术与自动化装置]