有机过氧化物反应热的危险性预测研究

发布时间：2018-01-19 18:52

本文关键词： 有机过氧化物热危险性多元线性回归偏最小二乘法支持向量机　出处：《中国安全生产科学技术》2016年11期 　论文类型：期刊论文

【摘要】：为了采用非实验的方法对安全物质学的研究内容及研究方法进行初探,基于定量结构-性质关系法,选择13种与有机过氧化物热危险性的影响因子密切相关的描述符,分别对起始分解温度T0和分解热△H的实验数据进行多元线性回归、偏最小二乘和支持向量机回归分析,从而获得3种相应的预测模型。对比T0与△H的实验值和预测值,结果发现:SVM预测模型的精度高于PLS预测模型,MLR预测模型的精度最低;同种预测模型对分解热的预测结果均优于起始分解温度。此外,分析各预测模型的稳定性数据发现:MLR模型的预测过程发生了过拟合现象,不具备预测能力;PLS模型的交互验证系数均大于0.5,具备较稳定的预测能力;SVM模型的交互验证系数均大于0.9,具备非常稳定的预测能力。
[Abstract]:In order to explore the research contents and methods of safety substances by non-experimental methods, the quantitative structure-property relationship method was used. Thirteen descriptors closely related to the thermal hazard of organic peroxides were selected and the experimental data of initial decomposition temperature T0 and decomposition heat H were analyzed by multivariate linear regression. By regression analysis of partial least square and support vector machine, three corresponding prediction models are obtained, and the experimental and predictive values of T _ 0 and H are compared. The results show that the accuracy of PLS prediction model is the lowest than that of PLS model. The prediction results of the same prediction model for decomposition heat are better than the initial decomposition temperature. In addition, the stability data of each prediction model are analyzed and it is found that the prediction process of the 1: MLR model is overfitted. Lack of predictive capacity; The interactive verification coefficient of PLS model is more than 0.5, which has stable prediction ability. The interactive verification coefficients of SVM model are all greater than 0.9, and have very stable prediction ability.
【作者单位】：湖南科技大学资源环境与安全工程学院;煤矿安全开采技术湖南省重点实验室;中国矿业大学安全工程学院;
【基金】：国家自然科学基金项目(51274100,51474105) 湖南省教育厅科研一般项目(15C0553)
【分类号】：O622.31
【正文快照】： 0引言2009年,国家发布了新的学科分类与代码标准,安全科学技术一级学科下面新增加了5个二级学科,其中包括安全物质学[1]。从安全物质学的学科内容可知,其研究仍处于初步探索阶段。中南大学的吴超等[2-3]从安全学科的角度出发,将安全物质学定义为:以人的安全和健康为着眼点,研

【相似文献】