A regulated boosting technique for material fatigue property prognostics

详细信息    查看全文

• 作者：De Z. Li^a ; ^{Derek.Li@emechsys.com} ; Wilson Wang^b ; ^{wwang3@lakeheadu.ca} ; Fathy Ismail^c ; ^{fmismail@uwaterloo.ca}
• 关键词：Material fatigue property prediction ; Boosting technique ; Ensemble learning ; Overfitting ; Time series forecast
• 刊名：Engineering Applications of Artificial Intelligence
• 出版年：2017
• 出版时间：January 2017
• 年：2017
• 卷：57
• 期：Complete
• 页码：153-159
• 全文大小：650 K
• 卷排序：57

文摘

A new regulated boosting technique, rBoost, is proposed in this work for time series forecast and material fatigue property prognosis. The rBoost employs the principle of ensemble learning, associated with base predictors. Different from general boosting techniques that are prone to overfitting when using relatively strong base predictors such as autoregressive model and radial basis function, the proposed rBoost technique aims to improve error convergence and reduce the overfitting problem using the new sample weight regulator. The effectiveness of the developed rBoost predictor is firstly demonstrated by simulation tests, and then the rBoost is implemented for material fatigue property prognosis. Test results show that the proposed rBoost predictor is an effective forecast tool; it can capture system dynamics effectively and track system characteristics accurately.