Conditional Generative Adversarial Networks (CGAN) for Abnormal Vibration of Aero Engine Analysis

The purpose of this paper is to present an evaluation method for abnormal vibration of aero-engine analysis based on flight data through the Conditional Generative Adversarial Networks (CGAN). The flight data are extracted from quick access recorder (QAR) designed to provide quick and easy access to raw flight data, including thousands of parameters. Through Conditional Generative Adversarial Networks for vibration of aero-engine analysis, we observe that distribution of aero-engine vibration parameters, which is based on the QAR data of the civil aircraft. Furthermore, the distribution of engine abnormal vibration can be observed based on few abnormal vibration flights by CGAN. Ultimately, the aero-engine vibration status can be obtained by the SVM (Support Vector Machine) classifier, relying on data augmentation for enhancement of abnormal vibration by CGAN. The results show that CGAN is an effective tool to solve the problem of insufficient samples. It can provide sufficiently accurate data to support the training of the analysis model, and can significantly improve the accuracy of the SVM classifier in identifying aero-engine abnormal vibration.