No abstract

A statistical engineering approach featuring a formally designed experiment has recently been implemented for calibration of an internal, strain-gauged, wind-tunnel balance at Arnold Engineering Development Complex. It was necessary to consider a substantially wide variation in balance temperature for this particular application. The cost to perform a balance calibration with the balance equilibrium temperature as a completely randomized factor was therefore prohibitive. A design-of-experiments approach using a run schedule with restricted randomization, known commonly as a split-plot design, is being evaluated. For more than a decade, design of experiments has been implemented in wind-tunnel strain-gauge balance calibration processes. Calibration with design of experiments is a characterization process where data are collected and analyzed using statistical methods, allowing conclusions to be drawn with chosen levels of confidence and power. A sequential approach to new balance calibration using existing hardware was demonstrated using a crossed design. A modified, two-active-factor, Box-Behnken design was executed at three temperature levels, and a second-order regression model was constructed in all factors. When compared to the traditional approach, a minimum 50% reduction in calibration duration and a reduction in overall residual error were demonstrated. Nomenclature a = number of whole-plot factors b = number of subplot factors PM1 = pitching moment of forward bridge, in: · lb PM2 = pitching moment of aft bridge, in: · lb rAF = axial force response, mV rPM1 = pitching moment response of forward bridge, mV rPM2 = pitching moment response of aft bridge, mV rRM = rolling moment response, mV rYM1 = yawing moment response of forward bridge, mV rYM2 = yawing moment response of aft bridge, mV β 0 = intercept of regression model β i = regression coefficient δ = whole-plot error ε = subplot error

No abstract