基于SSA优化BP神经网络的易燃易爆气体分类算法研究  

作　　者：万成炜 李捷 

机构地区：[1]武汉轻工大学,武汉430023

出　　处：《科技创新与应用》2024年第4期28-32,38,共6页Technology Innovation and Application

基　　金：The Natural Science Foundation of Hubei Province under Grant(2022CFB941)。

摘　　要：由各种传感器检测爆炸物的技术已经在多种条件下被用于检测不同类型的爆炸物,但同类型的爆炸物识别分类技术几乎没有,且针对传统BP神经网络模型收敛速度慢、易陷入局部最优值等问题,该文采用PWCLM混沌映射和高斯突变算子来改进麻雀搜寻算法(SSA),优化模型的初始权值,并对6种易燃易爆气体进行识别分类。最后再将该模型与BP、SSA-BP和WOA-BP等模型进行对比,使用3种评价指标对3种模型进行评价。结果表明,ISSA-BP模型稳定性高,识别精度优于其他几种模型,最终ISSA-BP模型的分类准确率、召回率和F1指数分别为99.01%、99.12%和99.12%。The technology of detecting explosives by various sensors has been used to detect different types of explosives under various conditions,but the same type of explosives identification and classification technology is almost not available,and the traditional BP neural network model convergence speed is slow,easy to fall into the local optimal value and other problems.In this paper,PWCLM chaotic mapping and Gaussian mutation operator are used to improve the Sparrow search algorithm(SSA),optimize the initial weight of the model,and identify and classify six kinds of flammable and explosive gases.Finally,this model is compared with BP,SSA-BP,WOA-BP and other models,and three evaluation indexes are used to evaluate the three models.The results show that ISA-BP model has high stability and better recognition accuracy than other models.The classification accuracy of ISA-BP model,recall rate and F1 index were 99.01%,99.12%and 99.12%respectively.

关 键 词：麻雀搜寻算法 PWCLM混沌映射 高斯变异算子 BP神经网络 易燃易爆气体 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]