HDD Slider Air Bearing Design Optimization Using a Surrogate Model

This report addresses a new optimization method in which the DIRECT algorithm is used in conjunction with a surrogate model. The DIRECT algorithm itself can find the global optimum with a high convergence rate. However the convergence rate can be much improved by coupling DIRECT with a surrogate model. The surrogate model known as the Kriging model is used in this research. It is determined by using sampling points generated by the DIRECT algorithm. This model expresses the shape of a hyper surface approximation of the cost function over the entire search space. Finding the optimum point on this hyper surface is very fast because it is not necessary to solve the time consuming air bearing equations. By using this optimum candidate as one of the DIRECT sampling points, we can eliminate many cost function evaluations. To illustrate the power of this approach we first present some simple optimization examples using known difficult functions. Then we determine the optimum design of a slider with 5nm flying height (FH) starting with a design that has a 7nm FH.