Simpler is Sometimes Better: A Dynamic Aero-Engine Calibration Study

Many real-world optimisation problems are in dynamic environments such that the search space and the optimum usually change over time. Various algorithms have been proposed in the literature to deal with such dynamic optimisation problems. In this paper, we focus on the dynamic aero-engine calibration, which is the process of optimising a group of parameters to ensure the performance of an aero-engine under an increasing number of different operation conditions. A real aero-engine is considered in this work. Three different types of strategies for tackling dynamic optimisation problems are compared in our empirical studies. The simplest strategy shows the superior performance which provide an interesting conclusion: Given a new dynamic optimisation problem, the algorithm with complex strategies and having excellent performance on benchmark problems is likely to be applied due to the lack of prior knowledge, however, the simplest restart strategy is sometimes well enough to solve real-world complex dynamic optimisation problems.