Intelligent Fault Classification and Identification of Heat Exchange Station Based on Time-Series Analysis

In this paper, an online fault diagnosis strategy based on time-series analysis is proposed, by combining the problem of inaccurate fault identification and fault classification during operation of the automatic control system in heat exchange station. First of all, this paper simulates common operating fault of heat exchange station, builds the fault database by selecting feature data, at the same time introduces Support Vector Data Description (SVDD) method to set the judgment threshold for the identification of unknown faults. Next, in order to solve the information transfer relationship between multiple operating conditions and various variables, transfer entropy and data sliding window technology are used to analyze the time series. Finally, introduces Long Short-Term Memory (LSTM) to classify of status and determine of fault type for different running data sequences. This method can identify known types of faults online, and can intelligently increase the abnormal detection of unknown faults verified by physical simulation platform, and it is helpful to guarantee the operation safety of the heat exchange station. This method can also be used as a reference for fault detection and dynamic and accurate classification in this kind of production process.