AMMSI: Ageing and Maintenance in reliability: Modelling and Statistical Inference

ANR Project Blanc SIMI1 Mathématiques et interactions 2011 - ANR-2011-BS01-021

March 2012 - February 2016

The aim of project AMMSI is to provide innovative methods and mathematical tools for the management of the ageing of industrial systems. The project includes the proposition of new stochastic models of degradation, failure and maintenance of complex systems, taking into account recurrent events, competing risks, covariates... It also includes the design of new statistical methods for analyzing such models and data from operation feedback, parametric, semi- or non-parametric methods, including goodness-of-fit tests and the treatment of missing data. Finally, it provides tools for decision support and industrial implementation of these methods in order to extend the lifetime of industrial systems by redefining the maintenance policy, while respecting safety, regulation and operating performance, in an economically optimal approach. The industrial implementation will require the development of computer codes, which may take the form of R or Matlab libraries, or may rely on the MARS software, developed by LJK and EDF R&D.

The scientific issues treated in the project include:

• Theoretical results on statistical inference of imperfect maintenance models.
• Goodness-of-fit tests for these models.
• Including covariates in these models.
• Integration of imperfect maintenance in degradation models.
• Dynamic maintenance planning.
• Development of new maintenance models: multivariate, with random or non-symmetric effects...
• Development of new degradation models: multivariate, competing, perturbed, with changes of degradation rate...
• Exploitation of online monitoring data.
• Treatment of masked data.

The originality of our proposal lies in the following points:

• Our project lies at the interface between mathematics and industry. The type of research is both basic and industrial. The core of our programme is basic research on stochastic modelling and statistical inference in reliability. But there will also be a strong part of industrial research since the theoretical results will be directly used for the management of ageing and maintenance planning of industrial systems.
• Each part of our work will treat both probabilistic and statistical aspects.
• We aim to link both approaches of ageing, lifetime and degradation.
• We highlight multivariate aspects, both on maintenance and degradation models.

To meet these challenges, our strength is to bring together various and complementary partners: 4 university laboratories (LJK, LMAP, LMB, ICD) combining expertise in applied probability, statistics and dependability, and 2 companies (EDF and SNCF) for which ageing management is a major industrial issue and who will bring new problems and data.

Keywords: stochastic modelling, statistical inference, reliability, maintenance planning and monitoring, ageing, degradation, industrial systems