Start-up Fault Detection of Gas Turbine Based on Dynamic Time Warping

The start-up system is an important auxiliary device of the gas turbine, and the start-up sequence of the gas turbine is a time series process. The performance of the start-up will directly affect whether the gas turbine operates or not. This paper takes a three-shaft gas turbine generator set as the research object and makes a detailed analysis of the common faults in the start-up process, and the fault tree of the start-up process is established. A fault detection method based on the dynamic time warping algorithm for the start-up process is proposed. The Euclidean distance matrix of the starting up time series is calculated. Then the dynamic time warping paths of the starting up time series data are attained by using reverse search. Finally, the start fault is detected by the constraint boundary conditions. The experimental results show that the presented method can accurately and effectively detect typical faults in the start-up process.