基于遗传BP神经网络的AlFe2O3铝热剂热力学性能预测方法研究
基金项目:
陕西省自然科学基金面上项目（2024JC-YBMS 414）。
Research on the Prediction Method of Thermodynamic Properties of Al/Fe2O3 Aluminum Thermite Based on Genetic BP Neural Network
- 摘要
|
- 图/表
|
- 访问统计
|
- 参考文献
|
- 相似文献
|
- 引证文献
|
- 资源附件
|
- 文章评论
摘要:为了探索Al/Fe2O3铝热剂热力学性能的预测方法，采用遗传算法对BP神经网络的初始权值和阈值进行优化，利用HSC Chemistry软件计算了不同温度及配比下的Al/Fe2O3吉布斯自由能及反应焓变，以此作为基础数据，建立了基于遗传BP神经网络的Al/Fe2O3铝热剂燃烧热力学性能预测模型，分别得到了124组Al/Fe2O3吉布斯自由能和化学反应焓变的训练集样本数据及31组预测集样本数据，并计算了模型误差。结果表明：该模型预测的Al/Fe2O3吉布斯自由能及反应焓变的测试集均方根误差（RMSE）分别为0.499 1和0.702 7，平均绝对误差（MAE）分别为0.533 2和0.441 1，决定性系数R2分别为0.982 7和0.988 5；遗传BP神经网络能够用于Al/Fe2O3铝热剂热力学性能的预测。
Abstract:In order to explore the prediction method for the thermodynamic properties of Al/Fe2O3 aluminum thermite, genetic algorithm was used to optimize the initial weights and thresholds of BP neural network. HSC Chemistry software was used to calculate the Gibbs free energy and reaction enthalpy change of Al/Fe2O3 at different temperatures and ratios. Based on this, a prediction model for the combustion thermodynamic properties of Al/Fe2O3 aluminum thermite was established using genetic BP neural network. 124 sets of training set sample data and 31 sets of prediction set sample data for Gibbs free energy and chemical reaction enthalpy change of Al/Fe2O3 were obtained, and the model error was calculated. The results show that the root mean square errors (RMSE) of the Gibbs free energy and reaction enthalpy change predicted by the model for Al/Fe2O3 are 0.499 1 and 0.702 7, respectively, with mean absolute errors (MAE) of 0.533 2 and 0.441 1, and determination coefficients R2 of 0.982 7 and 0.988 5, respectively; Genetic BP neural network can be used for predicting the thermodynamic properties of Al/Fe2O3 aluminum thermite.
var swiper = new Swiper('.swiper_xq', {
navigation: {
nextEl: '.swiper-button-next',
prevEl: '.swiper-button-prev',
},
});
$(function(){
$('img').bigic();
});
jQuery(".slideTxtBox").slide({ trigger: "click" });
