Examining improved experimental designs for wind tunnel testing using Monte Carlo sampling methods

Abstract

Wind tunnels are used in the design and testing of a wide variety of systems and products. Wind tunnel test campaigns involve a large number of experimental data points, can take a long time to accomplish, and can consume tremendous resources. Design of Experiments is a systematic, statistically based approach to experimental design and analysis that has the potential to improve the efficiency and effectiveness of wind tunnel testing. In Defense Acquisition, wind tunnel testing of aircraft systems may require years of effort to fully characterize the system of interest. We employ data from a fairly large legacy wind tunnel test campaign and compare that data's corresponding response surface to the response surfaces derived from data generated using smaller, statistically motivated experimental design strategies. The comparison is accomplished using a Monte Carlo sampling methodology coupled with a statistical comparison of the system's estimated response surfaces. Initial results suggest a tremendous opportunity to reduce wind tunnel test efforts without losing test information. Published in

Examining potential reductions in wind tunnel testing requirements

“…Additional details on the phase I aspect of this study are found in Hill et al 12 and research awards, including the Outstanding…”

Nonlinear screening designs for defense testing: an overview and case study

“…The partitioning of the data into transonic and supersonic subspace design regions enabled the fitting of low-order polynomial models. As such, Hill et al 4 limited the models to full quadratic models and pure cubic terms.…”

“…Hill et al 4 proceeded to generate response surface models for both the transonic and supersonic design regions using approximately 900 experimental design points of two alternative designs (Nested Face-Centered Design (NFCD) and an I -optimal computer-generated design) by sampling from the “ground truth” model with an N ( 0 , 0 . 0125 ) error added. They then compared the corresponding surfaces using a Monte Carlo sampling methodology coupled with a statistical comparison to determine the functional equivalency of the surfaces.…”

“…The major disadvantage of the OFAT strategy is that it does not consider interactions between factors. In addition, as Hill et al 4 point out, the OFAT strategy is not cost effective nor does it control experimental uncertainty or produce minimum variance predictions. Variations on OFAT, such as those found in traditional wind tunnel testing, can consider factor interactions but remain inefficient with respect to modern experimental designs.…”

“…Despite the disadvantages and deficiencies of implementing traditional experimentation strategies, conventional wind tunnel tests, which are a critical component in the Developmental Test and Evaluation (DT&E) of aeronautical systems, are usually conducted in this manner. Hill et al 4 point out that current high-performance military aircraft development programs may require up to 3700 wind tunnel test hours in the conceptual design phase and up to 18,500 hours in the development/validation phase. Such requirements become tenuous during a time when the United States Air Force (USAF) and DoD are seeking reductions in developmental schedules and budgetary requirements for system development and test.…”