基于机器学习的烘干机逆向因果推断研究  

Reverse Causal Inference for Dryers Based on Machine Learning

作　　者：卢忠岩 张伟 LU Zhongyan;ZHANG Wei(Shanghai Wisdom Information Technology Co.,Ltd.,Shanghai 200052,China)

机构地区：[1]上海威士顿信息技术股份有限公司,上海200052

出　　处：《数字制造科学》2024年第4期316-318,共3页

摘　　要：针对烘干机控制参数和物料烘后含水率的关系进行因果推断时受限于数据、成本等因素导致常用的因果推断方法无法有效解决的问题,基于烘干机及其控制系统的一些特性,提出了在烘干尾部阶段进行少量实验,并借助机器学习生成这些实验的反事实,从而推断出部分控制参数和烘后含水率的因果关系的方法。经实验验证通过这种方法得到的因果关系可以有效改进烘干尾部阶段的控制效果。To improve product quality in the drying process,a causal inference method was developed to explore the relationship between dryer control parameters and the moisture content of post-drying materials.Traditional causal inference methods were ineffective due to limited data and high costs.However,leveraging specific characteristics of the dryer and its control system,a novel approach was devised.This method involves conducting a small number of experiments during the tail-end phase of the drying process and using machine learning to generate counterfactual scenarios for these experiments.Experimental validation confirms that the causal relationships inferred through this method can significantly enhance control effectiveness during the tail-end phase of the drying process.

关 键 词：机器学习 烘干机 因果推断 自动化控制 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]