Applicability of correlational data-mining to small-scale turbojet performance map generation

Abstract

Abstract Turbojet applications benefit from accurate performance models. The aim of this study is to explore the applicability of data-mining algorithms to determine relationships between the generated thrust, the environmental conditions (free stream air-speed, inlet temperature and pressure) and the operating conditions (input fuel flow and shaft angular speed). For this purpose, experimental tests were carried out within wind-tunnel facilities using an experimental single-spool turbojet test bench. It is well-known that a large set of data-mining approaches relies on establishing linear correlations among input and output variables. The scope of this article is to assess the applicability of correlational data-mining approaches by i) an exploratory data analysis to find underlying data patterns and ii) principal component regressions to obtain a suitable predictive model for the generated thrust. Validation experiments demonstrated that the data-based model allows capturing the effects of the environmental and operating conditions with good accuracy (Root Mean Squared Error RMSE = 3.5100%), while maintaining a low complexity in the resulting structure. These results show that it is possible to generate turbojet experimental performance maps through data-mining algorithms with a correlational approach.