王昌健, 蒋国璋, 段现银.基于文本数据挖掘的复杂工况螺纹连接力矩衰减预测方法[J].河南理工大学学报（自然科学版）,2021,40(2):142-151.

WANG C J, JIANG G Z, DUAN X Y.Prediction method for attenuation decay of threaded connection incomplex working condition based on text data mining[J].Journal of Henan Polytechnic University(Natural Science) ,2021,40(2):142-151.

基于文本数据挖掘的复杂工况螺纹连接力矩衰减预测方法

王昌健1,2, 蒋国璋1,2, 段现银1,2

1.武汉科技大学 冶金装备及其控制教育部重点实验室，湖北 武汉 430081;2.武汉科技大学 机械传动和制造工程湖北省重点实验室， 湖北 武汉 430081

摘要:针对冲击、振动或变载荷环境等复杂工况下螺纹连接易失效，传统基于实验和仿真的方法无法高效、经济、准确地找出原因并对衰减进行预测的问题，本文利用人工智能技术在故障预测领域中的优势，提出一种基于数据挖掘的复杂工况螺纹连接力矩衰减预测方法。该方法从原始数据中挖掘其与螺纹连接故障的精确映射关系，准确找出螺纹连接发生失效的主要原因并对力矩衰减进行预测。首先，考虑导致文本极性变化的特殊语言结构，提出一种规则化故障量化方法；其次，结合螺纹连接力矩衰减专业领域词典，对螺纹连接力矩衰减进行基于文本描述的故障量化评级；再次，提出基于Logistic函数模型的特征构建方法并构建力矩衰减相关特征集；最后，建立基于随机森林和岭回归算法的Stacking集成学习预测模型。本文以重型卡车推力杆螺纹力矩衰减预测作为实际案例，验证了该方法的可行性和有效性，集成后的模型预测准确率较单一随机森林和岭回归算法模型平均提升了 53.39%。

关键词:文本数据挖掘;螺纹连接;力矩衰减;故障预测;Stacking集成学习

doi:10.16186/j.cnki.1673-9787.2019120093

基金项目:国家自然科学基金资助项目（51875379）；武汉科技大学国防预研基金资助项目（GF201901）

收稿日期:2019/12/27

修回日期:2020/04/13

出版日期:2021/03/15

Prediction method for attenuation decay of threaded connection incomplex working condition based on text data mining

WANG Changjian1,2, JIANG Guozhang1,2, DUAN Xianyin1,2

1.Key Laboratory of Metallurgical Equipment and Control of Ministry of Education， Wuhan University of Science and Technology， Wuhan  430081 ， Hubei， China;2.Hubei Key Laboratory of Mechanical Transmission and Manufacturing Engineering， Wuhan University of Science and Technology， Wuhan  430081 ， Hubei， China

Abstract:In view of the easy failure of threaded connections under complex working conditions such as shock， vibration， or variable load environments， the traditional methods based on experiments and simulations cannot efficiently，economically and accurately identify the main causes and predict the attenuation. In this paper， using the advantages of artificial intelligence technology in the field of fault prediction， a data mining-based method for predicting threaded connection torque attenuation under complex working conditions was proposed. This method could mine the accurate mapping relationship between the threaded connection failure and the original data， and accurately find out the main reason of the threaded connection failure and predict the torque attenuation. Firstty， considering the special language structure that causes the polarity of the text to change， a regularized fault quantification method was proposed； Secondly， combined with the threaded connection torque attenuation professional field dictionary， the threaded connection torque attenuation was fault quantitatively rated based on the text description ； Then， the feature construction method based on the Logistic function model is proposed and the torque attenuation related feature set was constructed； Finally， a stacking integrated learning prediction model based on random forest and ridge regression algorithm was established. The prediction of the torque attenuation of the thrust rod bolt of a heavy truck was used as a practical case to verify the feasibility and effectiveness of the method. The integrated model prediction accuracy was improved by an average of 53. 39% compared with a single random forest and ridge regression algorithm model.

Key words:text data mining;threaded connection;torque attenuation;fault prediction;Stacking ensemble learning

 基于文本数据挖掘的复杂工况螺纹连接力矩衰减预测方法_王昌健.pdf