摘要
针对工况改变后数据分布差异造成模型失配的问题及球磨机运行时振动信号的不确定性,研究了半监督域适应模糊推理的球磨机负荷参数软测量方法。该方法首先集成多个约束条件寻找特征变换矩阵,将待测工况和历史工况数据投影到公共子空间;接着利用模糊C均值聚类对投影后的历史数据划分规则并建立模糊推理模型。仿真实验结果表明该方法具有较高的预测精度,可以有效削弱不确定因素的影响,解决工况改变后数据分布差异造成的模型失配问题。
Aimed at the model mismatch problem caused by difference of data distribution after the change of working conditions and the influence of uncertain vibration signal when ball mill is running, a semi-supervised domain adaptation fuzzy inference method was researched. The method firstly integrates multiple constraint conditions to find the feature transformation matrix and projects the tested and historical data to the public sub-space. Then, the projected historical data is divided using fuzzy C-means clustering to establish fuzzy inference prediction model. The experiment shows that the method had high prediction accuracy, and could effectively weaken the influence of uncertain factors and solve the problem of model mismatch which is caused by data distribution difference of working condition changes.
作者
李思思
杜永贵
闫飞
阎高伟
LI Sisi;DU Yonggui;YAN Fei;YAN Gaowei(College of Electric and Power Engineering,Taiyuan University of Technology,Taiyuan 030024,China)
出处
《太原理工大学学报》
CAS
北大核心
2019年第3期364-368,共5页
Journal of Taiyuan University of Technology
基金
国家自然科学基金资助项目(61450011)
山西省煤基重点科技攻关项目(MD 2014-07)
山西省自然科学基金资助项目(2015011052)
关键词
迁移学习
多工况
半监督域适应
模糊推理
球磨机
负荷参数
软测量
transfer learning
multi-mode
semi-supervised domain adaptation
fuzzy inference model
ball mill
load parameter
soft sensor