AText Mining based Reliability Analysis Method in Design Failure Mode and Effect Analysis

Abstract

Failure Mode and Effects Analysis (FMEA) is one of the basic and the most commonly used techniques in reliability analysis and management. Design FMEA is the application of the FMEA method in product development phase and have shown remarkable results in various engineering design fields. DFMEA reports usually contain a lot of text information about quality and reliability improvement opportunities for new product development. However, it can be very difficult and non-intuitive to fully understand these text information for design improvements. To address this challenge, we propose to apply text mining methods in DFMEA reports to discover the hidden reliability information. Specifically, three types of hidden information are investigated, which include the correlation of the failure related features in the potential failure mode or failure cause, the classification of failure modes and causes, and the optimal set of improvement activities according to the effect and cost metric proposed in this paper. Among them, one feature represents a failure characteristic or a location of the potential failure mode or failure cause (e.g. wear, break, bend, surface, insert, stem). These three pieces of information can be further integrated to product verification and validation process for effective risk identification and mitigation in a more economic and effective way. Finally, we illustrate the application of the text mining methods in DFMEA report for a subsystem in a power generation system within a diesel engine.