Efficient Pointing Accuracy Analysis Using Support Vector Machine for Highly Precise Space Truss Structure Due to Clearance Uncertainty

Abstract

This research proposes an efficient pointing accuracy analysis method applying a support vector machine (SVM), one of machine learning methods, to reduce computational cost sufficiently for a highly precise space truss consisting of multiple-stage structures with uncertainties of member lengths and the node clearances. The SVM is applied to classify the nodal position definiteness of the truss structure in terms of the uncertain variables to avoid useless pointing accuracy analysis for the instable structures. For an efficient classification, the nodal position definiteness for a larger-stage truss is estimated from learning results for a fewer-stage truss. Through numerical examples, the nodal position definiteness for 20- or 30-stage truss is demonstrated to be efficiently classified by the learning results for only 3-stage truss structure. Then, the pointing accuracy evaluated as the direction angle distribution is confirmed to be identical to the result by Monte Carlo simulation.