Abstract
The operation condition of aluminum electrolytic cells is critical to the stability of the aluminum electrolysis process. Under the tough working environment, there will be many abnormal states in the cells, detrimental to aluminum quality, aluminum electrolysis efficiency and electrolytic cells life. Therefore, it's of great significance to develop an effective process monitoring and multi-fault diagnosis method. The diagnosis accuracy and the speed of traditional multi-classification methods are limited. To solve these problems, a fault detection and multi-fault diagnosis framework based on wavelet packet decomposition (WPD) and directed acyclic graph support vector machine (DAG-SVM) is proposed. First, features are extracted from aluminum electrolytic cell voltage signals. Then, DAG-SVM is used to detect and diagnose abnormal states. The result shows that the proposed method can detect and diagnose three abnormal states, anode effect state, voltage instability and liquid aluminum fluctuation with a high precision. The accuracy can reach about 86.8%.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have