Complex and computationally intensive modeling and simulation of real-world engineering systems can include a large number of design variables in the optimization of such systems. Consequently, it is desirable to conduct variable screening to identify significant or active variables so that a simpler, more efficient, yet accurate optimization process can be achieved. This paper proposes a variable screening method based on a Kriging model with Restricted Maximum Likelihood criterion. The Kriging metamodeling method is more reliable for highly nonlinear systems than the traditional response surface method, and the Restricted Maximum Likelihood criterion makes the variable screening process more efficient. In this work, three different sampling methods, namely, Latin Hypercube, Improved Distributed Hypercube and D-optimally selected Latin Hypercube sampling methods, are compared when used with the proposed variable screening method. The new variable screening method is evaluated using a standard benchmark nonlinear numerical example that employs 20 factors. The variable screening method then is applied to a gunner restrain system design problem. Starting with a set of eight of the most important gunner’s joints variables in a wide open property space, the Improved Hypercube space filling sampling technique is used to screen the joints variables and the variable screening method based on the Kriging method with the restricted maximum likelihood criterion is used to determine the most important joints properties for gunner’s performance.

This content is only available via PDF.
You do not currently have access to this content.