A Clustering-Based Method for Health Conditions Evaluation of Aero-Engines

Due to the effect of performance degradation and various work loads of aero-engines, it is difficult to accurately evaluate and correctly classify the operational condition of engines. This paper proposes a condition evaluation method based on the k-means clustering method, random forests algorithm (RF) and principal component analysis (PCA) for aero-engines. Firstly, the RF and PCA are used to preprocess the health condition data of engines before clustering by k-means. Then, the k-means method is adopted to cluster the different health condition of engines according to the preprocessing. The feasibility and reasonability of the proposed method are verified by operation data of aero-engines monitored by sensors in different conditions finally. The results show that the proposed method has an excellent performance to distinguish the different latent conditions of the engine, indicating it can better meet the needs of engineering applications and make decisions via the analysis results reasonably. Meanwhile, the analysis results of this paper also proof the correctness and reasonableness of the predefined health conditions of aero-engines in previous researches.