Abstract Various optimization algorithms are used in many engineering problems to guide the design in the preliminary design phase. Lightening the weight of turbomachines, increasing their efficiency, and having faster-rotating rotor structures constitute the basis of the engineering problems in the turbomachinery field. According to the No Free Lunch Theorem, appropriate optimization algorithms must be used for appropriate optimization problems. When evaluated in terms of the rotordynamics of a turbomachine, although there are rotordynamics optimization studies based on the Finite Element Method (FEM) in the literature, the selection of the appropriate algorithm that provides better solutions for the optimization problems encountered has not been entirely determined. In this study, the Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) methods, which are population-based modern optimization algorithms, were evaluated for the multi-objective rotordynamics optimization problem and compared with the non-population-based Pattern Search (PS) method. In order to speed up the solution time, the JoeRot rotordynamics solver based on the Complex Transfer Matrix Method (CTMM), which gives a natural frequency response with an average of 0.0345 seconds even if the number of elements increases, was used to create the objective function. Each algorithm was compared 50 times for 24 particles and 1000 iterations, ensuring fast solutions. As a result, 33.82% natural frequency and 23.56% mass improvement were achieved for the initial rotor model. The PSO algorithm gave the best result. Although PS gave results with lower standard deviation than other methods, it did not converge as well as PSO.
Niş et al. (Mon,) studied this question.