Applying least squares support vector machines to the airframe wing-box structural design cost estimation

This research used the least squares support vector machines (LS-SVM) method to estimate the project design cost of an airframe wing-box structure. We also compared the estimation performance using back-propagation neural networks (BPN) and statistical response surface methodology (RSM). The solution mechanism of the LS-SVM involved a simultaneous searched for the maximal margin as the target, taking into account the error calculated during training phase to determine the estimation problem models. Two case studies involving the wing-box structure was investigated the separate structural parts case and the mixed structural parts case. The test results verified the feasibility of using the LS-SVM as well as its ability to make accurate estimations.