Description

This paper presents the results of research into the problem of optimizing large-scale engineering problems. Such problems can often be simulated extensively by computer beforehand to aid in the design process. Optimization programs can be very useful at this point to help the designer minimize or maximize certain aspects of the design. A drawback is that the computer programs used are often very large, and computing costs can become prohibitively expensive as the number of simulations increase. It then becomes necessary to turn to an optimization routine that not only finds a good solution but can do so efficiently with respect to the number of objective function evaluations, i.e., simulation runs.

A quadratic search algorithm is proposed as being an efficient strategy for some problems and is described. Its application to the optimization of a central chilled water distribution system is presented. For this problem, the results of the quadratic search algorithm are shown to reach comparable or better optimum points in fewer Function evaluations than two other popular search techniques, the Generalized Reduced Gradient technique and the Nelder and Mead method.

Citation: Symposium, ASHRAE Transactions, 1987, vol. 93, pt. 2, Nashville, TN

Product Details

Published:

1987

Number of Pages:

14

File Size:

1 file , 1000 KB

Product Code(s):

D-NT-87-24-1