Standardized failure signature for a turbofan engine

The capacity to master engines behavior is fundamental for a manufacturer to prove its efficiency in conception and maintenance capability. This understanding goes through the capacity to acquire and treat data flows produced by sensors for monitoring purposes. One of many issues encountered by engine experts when reading measurements is that a given engine never works under exactly the same conditions. However, a practical abnormality detector must be able to check for failures in any possible context. Maintaining a monitoring application is expensive and one never wants to build specific computations for each flight phases and external conditions. The purpose of this article is to present a normalization pretreatment which leads to the conception of standardized signatures. Those signatures continue to carry a clear meaning for experts but are independent of flight conditions. With such a data pretreatment a classification of failures becomes possible and may be interpreted in physical terms by engineers. This will help to produce a degradation identification solution and adapt maintenance workflows. This work is mainly illustrated by the failure classification and critical component identification for the start system of a civil engine.