A Tutorial on Support Vector Machines for Pattern Recognition
Data Mining and Knowledge Discovery
Structural and Multidisciplinary Optimization
| Hi-index | 0.00 |
The paper proposes a novel approach to identify the feasible region for a constrained optimisation problem. In engineering applications the search for the feasible region turns out to be extremely useful in the understanding of the problem as the feasible region defines the portion of the domain where design parameters can be ranged to fulfil the constraints imposed on performances, manufacturing and regulations. The search for the feasible region is not a trivial task as non-convex, irregular and disjointed shapes can be found. The algorithm presented in this paper moves from the above considerations and proposes a recursive feasible-infeasible segment bisection algorithm combined with Support Vector Machine (SVM) techniques to reduce the overall computational effort. The method is discussed and then illustrated by means of three simple analytical test cases in the first part of the paper. A real-world application is finally presented: the search for the survivability zone of a crashworthy helicopter seat under different crash conditions. A finite element model, including an anthropomorphic dummy, is adopted to simulate impacts that are characterised by different deceleration pulses and the proposed algorithm is used to investigate the influence of pulse shape on impact survivability.