基于"物理模型+数据驱动"神经网络的纤维过滤效率预测研究
Research on the Prediction of Fiber Filtration Efficiency Based on"Physical Model+Data"Driven Neural Network
刘烈亮 1谭百万 1时运强 1朱辉2
作者信息
- 1. 东莞拓斯达智能环境技术有限公司绿能事业部
- 2. 桂林电子科技大学建筑与交通工程学院
- 折叠
摘要
大数据和人工智能的飞速发展为复杂工程问题的研究提供了一种全新的探索范式.粉尘颗粒在纤维过滤器中清除过程是一个复杂的非线性动力学过程,难以考虑多因素非线性耦合作用给出准确的纤维过滤效率预测模型,采用物理模型+数据驱动的人工神经网络技术构建纤维过滤效率预测模型.在简化纤维过滤模型基础上,采用Lagrangian颗粒动力学方法求解颗粒运动规律,获得在拦截和扩散耦合作用下的纤维过滤效率;以Pe数、拦截参数R和纤维填充率C作为因变量建立纤维过滤效率数据库,以此作为神经网络模型学习训练数据样本.结果表明,神经网络模型可以在合理的计算时间内进行训练,并且能够准确估计纤维过滤效率.
Abstract
The rapid advancement of big data and artificial intelligence has opened up new avenues for exploring complex engineering problems.The process of dust particle removal in fiber filters is a highly intricate nonlinear kinetic process,making it challenging to develop an accurate prediction model for fiber filtration efficiency that takes into account the multifactorial nonlinear coupling.In this study,an artificial neural network driven by both physical models and data was employed to construct a predictive model for fiber filtration efficiency.Based on a simplified fiber filtration model,the Lagrangian particle dynamics methodwas employed to solve particle movement with consideration of the capture mechanismsof interception and diffusion.A database of fiber filtration efficiency was established with dependent variables including Pe number,interception parameter R,and fiber filling rate C.This database serves as training data samples for the neural network model,and the results show that the neural network model can be trained within a reasonable calculation time frame while accurately estimating fiber filtration efficiency.
关键词
空气过滤/颗粒捕集/数值模拟/神经网络Key words
Air filtration/Particle capture/Numerical simulation/Neural network引用本文复制引用
出版年
2024
NETL
NSTL
万方数据