XGBoost based delay estimation and compensation control of engine guide vane angle servo system

To achieve effective compensation control of the engine guide vane angle servo system （GVASS） during delay changes， the characteristics of the GVASS are tested and sampled. The time delay estimation problem was treated as a classification problem， and the extreme gradient boosting （XGBoost） algorithm was proposed to classify the degree of the time delay. The 2D tensor after the command change was used as the input to construct the tree model and achieve accurate delay estimation with an accuracy exceeding 98.29% on both training and testing datasets. The extreme learning machine is adopted to establish the GVASS model and perform delay compensation control. Simulations of guide vane time delay changes were carried out， and the results demonstrate that compared with the compensation algorithm using constant time delay， the time delay estimation and compensation control based on XGBoost can effectively deal with the deterioration of the system response characteristics after the delay changes， and achieve no-overshoot fast tracking of the guide vane servo system under large time-delay variations.