An ensemble SVR based on modified Adaboost.RT algorithm for predicting the degradation of a gas turbine compressor
| Autor: | Dongfeng Liu, Zhicai Zhou |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
Imagination
Gas turbines 0209 industrial biotechnology Engineering Support vector regression machine business.industry media_common.quotation_subject 02 engineering and technology Support vector machine Search engine ComputingMethodologies_PATTERNRECOGNITION 020901 industrial engineering & automation Robustness (computer science) 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing AdaBoost business Gas compressor Algorithm media_common |
| Zdroj: | 2016 Prognostics and System Health Management Conference (PHM-Chengdu). |
| DOI: | 10.1109/phm.2016.7819927 |
| Popis: | The original AdaBoost.RT algorithm determines the fixed weights of combined machines by learning the errors of each training sample, but the difference in learning ability between the new sample and the training samples is neglected. To address this problem, this study propose an ensemble support vector regression machine (SVR) based on modified AdaBoost.RT algorithm. Firstly, an SVR algorithm is selected as the ensemble predictor due to its strong performance and robustness. Secondly, a modified AdaBoost.RT combined k-nearest neighbor algorithm (kNN) is proposed which overcomes the limitation of the original AdaBoost.RT by dynamically adjusting the weights of the combined machines. Finally, an ensemble SVR based modified AdaBoost.RT is presented and examined by predicting the gas turbine compressor degradation, and compared with models that are established using a single SVR and original AdaBoost.RT. The experiments have demonstrated that the proposed method can improve the general performance and effectively boost the accuracy, and the method is satisfactory to predict the process of compressor degradation in real time. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|