Effective Tests for Discovering Unexpected Structural Failure Modes

Abstract

Many of airplane accidents are due to unexpected failure modes. In structural design, appropriate failure-criterion characterization including identification of potential failure modes is key to reducing such accidents. However, since analytical predictions often have substantial errors due to complexity of failure mechanism, this process tends to depend much on tests. In this study, we demonstrate failure-criterion characterization with simple structural elements in order to shed light on effective testing. We first show that ignoring the existence of critical failure modes could lead to a substantially overestimated probability of failure by two orders of magnitude for a composite laminate example. Tests suffer from noise and so they are often repeated to reduce that noise. Resource allocation between the number of different tests in search of unexpected modes and the number of repetitions of each test is investigated. It is shown that increasing the number of tests for exploration is vital to spot unexpected failure modes, at the same time, it is almost as effective as repetitions for reducing the effect of noise on accuracy. Furthermore, the importance of the number of different tests for exploration is amplified when the failure load surface becomes complicated.