基于IPSO-BP模型的火灾气体传感器气压补偿算法

为完善飞机火灾检测系统,设计一套方案,模拟试验不同气压下CO、CO₂气体传感器采集气体的体积分数值,并与理论值比较,进而提出一种根据粒子适应度值动态调整学习因子的粒子群算法。采用改进的粒子群(IPSO)算法寻找反向传播(BP)神经网络的最优初始权值阈值,再利用寻优后的BP神经网络修正CO、CO₂气体传感器的检测结果,消除气压对于传感器采集数据的影响。结果表明:经过IPSO-BP算法补偿后,选取的27个气压点下气体体积分数值都接近拟合真实值,其中,CO₂气体传感器经过气压补偿后,测量最大误差从542×10^-4%下降到0.1×10^-4%,CO气体传感器测量最大误差从15.7×10^-4%下降到0.01×10^-4%。与BP神经网络气压补偿模型相比,IPSO-BP神经网络气压补偿模型的精度有明显的提高。

Pressure compensation algorithm of fire gas sensors based on IPSO-BP model

In order to improve fire detection system of aircraft,volume fraction values of CO and CO₂ gas sensors under different air pressure were simulated and compared with the theoretical value by designing a set of experimental scheme,and a particle swarm optimization algorithm was proposed to dynamically adjust learning factor according to particle fitness value.Then,IPSO algorithm was used to find optimal initial weight and threshold of BP neural network which,after optimization,was employed to modify detection results of CO and CO₂ gas sensors to eliminate influence of air pressure on sensor data acquisition.The results show that after compensation by IPSO-BP algorithm,gas volume fraction values at the selected 27 pressure points are close to the fitted true value.Among them,the maximum measurement error of CO₂ gas sensor drops from 542×10^-4%to 0.1×10^-4%after pressure compensation,and that of CO gas sensor decreases from 15.7×10^-4%to 0.01×10^-4%.Compared with BP neural network pressure compensation model,the accuracy of IPSO-BP neural network model is significantly improved.