Age Based Clustering Prognostics of Gas Turbines and Evaluation of The proposed Method Robustness in Data Deficient Conditions

The acceptable preformance of the data driven prognostics methods usually requires large amount of data, therefore the performance usually is not desirable for small amount of data conditions. Age clustering method multiplies the volume of the train data through observing data at multiple points. The advantage of the method is that it can be used for learning from a small set of data. The proposed approach is integratable with existing prediction methods and improves their results accuracy significantly. In this article, the ABC prognosis framework is described, its effectiveness for prognosis in normal conditions is illustrated in a case study on turbofan engines and a comparison with existing results on the same data is made. The paper continues with a study on the robustness of the proposed method under limited data conditions. The prognosis accuracy is compared for the case study in various conditions of available train data. The results emphasize (1) the efficiency of the method compared to other existing approaches in normally rich data condition and (2) the robustness of the results under limited data condition.