Embedding Vectorial Learning Method for Multi-Objective Reliability Analysis of Nose Wheel Steering Rack and Pinion Mechanism

Abstract

To improve the efficiency and accuracy of multiple objectives reliability analysis, an embedding vectorial learning (EVL) method is proposed by integrating embedding learning with vectorial modeling method. The embedding learning is designed to establish optimal limit state functions of multi-objective through structured combination generalization and error correction of modification generalization. The vectorial modeling method is developed to transform the repeated modeling of a single objective to synchronous modeling of multi-objective through matrix theory. The components of nose wheel steering rack and pinion mechanism are adopted to highlight the abilities of the proposed EVL in accuracy, wall-clock runtime and design conformance. It is illustrated that (i) the best performance of modeling phase is held by the EVL method, respectively 0.8872, 4.9982 and 3.9445s in MAE, R2 and calculational time; (ii) the EVL method has advantages in simulation performance, severally at 0.9745, 0.9988 and 0.3373s in reliability, closeness and simulation time; (iii) the joint failure probability by considering the correlation of multi-failure modes is 2.5488% that is lower than the failure probability 2.6762% depending on the mutual independence assumption. The effort of this work provides an effective analytical tool for complex mechanism with multi-objective and failure correlation.

Keywords: multiple objectives, embedding learning, reliability analysis, vectorial modeling, nose landing gear