Skip to Main content Skip to Navigation

Event-based probabilistic approach for proactive maintenance to improve production capacities in HMLV industries

Abstract : INTRODUCTIONThis thesis takes place in the framework of the European project INTEGRATE.Given the highly competitive economic environment in many industry, increasing demand volumes and product functional diversities have led the emergence of automated and hi-mix low-volume production environment. The production is characterized by its complexities and uncertainties. The production equipments represent very significant capital costs and are actually operated at the limit of their capabilities. These are challenged by equipment breakdowns with unknown failures that significantly disrupt the production capacities. In this environment, improved equipment availability becomes more than ever a key factor.MOTIVATIONTo improve the esciency of these production equipments, it is necessary to enhance the availability of equipment by intervening at the right time, on the right equipment or piece of equipment.-what is expected (Predictive methodology to anticipate failureoccurrence-why do we need it (Anticipation of failure may become a criteriaof decision to proactive decision and actions aiming the avoidanceof failure or minimization of its effects)-where does it takes place (Physical level of analysis on productionequipment)-when does it takes place (On-line, before functional failures)-who is involved (Experts on the input to the methodology, andmaintenance personnels for the expected output criteria)-how it is done (Identification of rules/patterns to failure).The main objective of the phD thesis is to study equipment failureprediction techniques and approaches to make such decisions.CONTRIBUTIONThis dissertation presents methodology to extract and validate rules and/orpatterns as failure signatures for real time failure predictions.After a detailed study of the failure prediction approaches, we proposean integral methodology respecting following properties,1. Aiming on-line maintenance strategies generation2. Using event based contextual information collected from product, process,equipment and maintenance data sources due to the facts thatequipment is not always the source of product quality drifts and thatsensors reliability issues could result in under or over engineering3. Working at module level of production equipment (piece sharing commonfunctions)4. Using Bayesian Network to model current state of equipment (representingfailure-cause relationship and taking into account uncertainties).It serves as the basis to extract the rules and patterns to failures.5. Flexibility with consideration of experts provided criteria, specificallylead and warning time for maintenance decisions and actions.The solutions that will be proposed is expected to enable improved: availabilityof equipment, resource management, cycle time and overall performanceof production.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Thursday, September 2, 2021 - 4:42:11 PM
Last modification on : Friday, March 25, 2022 - 9:43:13 AM
Long-term archiving on: : Friday, December 3, 2021 - 8:51:58 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03332417, version 1



Asma Abu Samah. Event-based probabilistic approach for proactive maintenance to improve production capacities in HMLV industries. Automatic. Université Grenoble Alpes, 2016. English. ⟨NNT : 2016GREAT122⟩. ⟨tel-03332417⟩



Record views


Files downloads