Design of a decision support system for buried pipeline corrosion assessment

Laurence Boudet (1) , Jean-Philippe Poli (1) , Alicia Bel (2) , François Castillon (3) , Frédéric Gaigne (3) , Olivier Casula (2)

1 LI3A (CEA, LIST) - Intelligence Artificielle et Apprentissage Automatique (CEA, LIST)
2 DNAQ - CEA Tech Nouvelle-Aquitaine
3 TIGF

Maintaining the level of integrity of pipeline networks to guarantee at least a reliable and safe service is a challenge operators of such networks are facing everyday. TIGF is one of the French operator which manages 5000 km of pipelines in the south-west quarter of France. This paper presents a decision-making tool which automatically ranks the pipeline sections regarding the risk of deterioration (damages and corrosion) and the gravity of the consequences, indicating which pipeline sections should be excavated. The tool relies on a fuzzy expert system which gathers 26 input variables, processes more than 300 rules, classifies the risk of deterioration into 7 classes and estimates the gravity. The rules are a formalization of human expertise: the fuzzy logic helps to tackle the vagueness of their knowledge and the measurement inaccuracy of some of the 26 input variables. The method has been tested on past excavations to assess its performances.

Mots clés: artificial intelligence; fuzzy logic; machine learning; online learning; decision-making tool; pipeline integrity; classification; signal processing; risk analysis; Trustworthy Artificial intelligence

Domaines: Intelligence artificielle [cs.AI]; Machine Learning [stat.ML]

Laurence Boudet, Jean-Philippe Poli, Alicia Bel, François Castillon, Frédéric Gaigne, et al.. Design of a decision support system for buried pipeline corrosion assessment. 17th International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems. Applications, Jun 2018, Cadiz, Spain. pp.74-85, ⟨10.1007/978-3-319-91479-4_7⟩. ⟨cea-01838410⟩