Aero-engine Working Condition Recognition Based on Light Gradient Boosting Machine Optimized by Improved Bat Algorithm
Working condition recognition is an important part of military aircraft aero-engine state monitoring. In order to improve the accuracy and efficiency of aero-engine working condition recognition, and avoid misjudgment during manual recognition and time-consuming and laborious problems in manual recognition. A rapid improved Light Gradient Boosting Machine (LightGBM) intelligent recognition method optimized by mutation chaotic bat algorithm (MCBA) is proposed in this paper. First, the weight velocity, the uniform transformation and the Gaussian transformation, the chaotic optimization are applied to optimize the bat algorithm, which constructs the MCBA and improves the convergence speed and accuracy of the bat algorithm. Second, the proposed MCBA is used to optimize the hyper parameter of LightGBM, Learning_rate and Max_depth. Extract features of flight data to select input parameters, and then the MCBA-LightGBM classifier is trained based on the selected feature parameters. Third, we test the performance of the classifier. And the working condition of two flight sorties of a certain aero-engine is identified. Experimental results show that the method can identify the engine working condition rapidly and accurately, and can be applied to the research and application of aero-engine state monitoring.