Kriging Modeling for Engine Mount Optimization in Motorcycles
This paper addresses two critical aspects associated with the successful use of a Kriging model for solving the engine mount optimization problem. The two aspects are the selection of an appropriate correlation function and the use of a suitable governing design for sampling within the design space. The selection of a correlation function is critical in building a Kriging model since the function should accurately represent the behavior of the response over the entire design space. Whereas the Gaussian correlation function is most commonly used for building Kriging models, it is generally suitable for only those processes or systems which have a relatively smooth response within the entire design space. The correlation functions that have been evaluated in this paper for building the Kriging models for solving the engine mount optimization problem are as follows: Exponential, Linear Spline, Matern’s 3/2, Matern’s 5/2 and Gaussian. Three types of experimental designs – Fractional Factorial, D-optimal and Latin Hypercube, have been used to select the sampling points for making simulation runs in order to build the Kriging models. A theoretical model that represents the dynamics of the engine mount system in a motorcycle application has been used to build all the surrogate models. The Kriging models are then used to solve the engine mount optimization problem for enhanced vibration isolation with mount stiffness, mount orientation and mount location as the design variables. The optimization results of the Kriging models are compared to the results of the theoretical model. It is found that the D-optimal design in conjunction with Matern’s 3/2 correlation function provides the best results. This can be attributed to the high irregularity of the response function in the design space, especially due to the influence of orientation variables. The use of the surrogate Kriging model simplifies the governing model and leads to a substantial reduction in computational effort for solving the optimization problem. Based on the results, it can be concluded that the Kriging modeling technique can be successfully used to build surrogate models for the engine mount problem for design iterations as well as for design optimization if the correlation function and the governing design are judiciously chosen.