Numerical hybrid Taguchi-random coordinate search algorithm for path synthesis

Abstract

We present a new algorithm for application to path synthesis of a four-bar linkage with continuous variables called the Hybrid Taguchi random coordinate search algorithm (HTRCA). The HTRCA combines two algorithms of the Taguchi method (TM) and the random coordinate search algorithm (RCA). The RCA adjusts one variable simultaneously with two directions and various step sizes in random order to escape a local minimum. Since the RCA modifies one variable at a time according to the random order, the result of the RCA is sensitive to an initial condition of variables. The TM was adopted to generate nearly optimal initial conditions. Using the TM, the approximate optimal value can be found in even multi-modal functions. Also, the TM is used with different setting values to escape a local minimum point by changing two or more variables in one step. By combining the two methods, a global optimal value can be found efficiently. Seven test functions were optimized and the robust and efficient performance was verified by comparison with other hybrid optimization algorithms. Finally, path synthesis of four-bar linkage is done by the developed optimization algorithm, and the result is compared to previous works on the same problem with evolutionary algorithm.