Pipeline leak diagnosis based on wavelet and statistical features using Dempster-Shafer classifier fusion technique

详细信息    查看全文

• 作者：Morteza Zadkarami ^{zadkarami@put.ac.ir} ; Mehdi Shahbazian ; ^{shahbazian@put.ac.ir} ; Karim Salahshoor ^{salahshoor@put.ac.ir}
• 关键词：Leak diagnosis ; Multi-Layer Perceptron Neural Network (MLPNN) classifier ; Wavelet transform ; Statistical features ; Dempster&ndash ; Shafer (D&ndash ; S) classifier fusion ; Correct Classification Rate (CCR) ; OLGA software
• 刊名：Process Safety and Environmental Protection
• 出版年：2017
• 出版时间：January 2017
• 年：2017
• 卷：105
• 期：Complete
• 页码：156-163
• 全文大小：1428 K
• 卷排序：105

文摘

The proposed leak diagnosis method not only detects the leak occurrence, but also identifies its location and size. Different feature extraction methods including statistical and wavelet-based techniques are applied. In this paper the leak diagnosis problem is viewed from a classification perspective. Dempster–Shafer classifier fusion approach is used to improve the performance of this method. This novel leak diagnosis approach only uses inlet pressure and outlet flow signals.