Proposal of a Validation Method of Failure Mode Analyses based on the Stress-Strength Model with a Support Vector Machine

Abstract This study aims at developing a validation method for the rational association between the design deviations, possible damage/fracture modes, and the eventual failure modes using SVM. Product failures due to damage/fracture modes of materials have recently increased, which suggests the importance of proactive prevention by failure modes analyses. Conventional Failure Modes and Effects Analyses (FMEA) lacks a specific process in determination of possible damage/fracture of component’ s materials. A modified Design Review Based on Failure Modes (DRBFM) or Design Deviation Method (DDM) were proposed to determine failure modes induced by damage/fracture modes of materials by possible associations from deviations in design/environmental factors to the deteriorations in stress-strength model. These methods still remain the possible errors in selection of rational damage/fracture modes to complicated patterns of design deviations, which suggests the importance of validation on the result of failure modes analyses. The procedure of DDM was formulated using sparse matrix and then analyzed using multi-class determination by Support Vector Machine(SVM). A case study of failure modes analyses to a laser-irradiation device using FMEA, DRBFM, the proposed DDM were conducted by individual three groups which had same number of 7 participants. The participants in the DDM group could predict more failure mechanisms by damage/fracture modes of materials than those by the DRBFM and FMEA groups. Furthermore, SVM showed higher precision rate from 77% to 100%, which evaluated the validity of specific failure modes. The SVM could determine the rational associations among the design deviations, the deviations in the SSM, and the corresponding damage/fracture modes.