Sensor Fault Signal Classification Based on Visual Information Enhancement for In-service Gas Turbine Control Systems

One-dimensional time series data was converted into a two-dimensional image through the Graham angle field (GAF), and image feature extraction was completed by using the histogram of the directional gradient, then fault signal classification was completed by using multi-classification support vector machine. Based on the actual operating data of a 9F gas turbine in power plant, five typical fault signals with noise were superimposed on the normal data to verify the effectiveness of the proposed method, and the confusion matrix was used to further visualize the fault classification results. Results show that the proposed method can effectively identify weak and gradient fault signals. The accuracy rate of fault classification is 96.6%, and both the GAF and the Markov transition field (MTF) have high fault classification accuracy.