Local and International students are welcome in our Department of Machine Design and Production. Subjects in Reliability and Maintenance require skills in statistical analysis and numerical simulation.
Two families of topics can be addressed:
- Predictive Maintenance Optimization. This includes statistical reliability models, physics-of-failure models, degradation models to estimate the residual lifetimes and to optimize maintenance policies. Relations between product quality, process performance and reliability. (Note: internships in English-speaking industrial companies can be arranged)
- Risk-based Maintenance. Relations between maintenance (including inspection) and major risk hazards. Risk aversion in maintenance. Dynamic Fault Trees.
To be concrete, here are some proposals of subjects in Reliability and Maintenance Engineering
Identification and interpretation of Weibull reliability laws
The two-parameter Weibull distribution is widely used for modelling the reliability of components. There are well established methods to identify the parameters from failure times (even censured) but more attntion should be paid for this when dailure times are biaised by preventive maintenance actions. Some outputs of the Weibull models like time to failure, age at replacement based on risk and residual lifetime merti further discussion as suggested in the article of Jiang and Murthy published in 2011 in Reliability Engineering ad System Safety.
Improvement of statistical reliability models through appropriate sets of failure data
With the improvement of the reliabilty of indistrial equipment, failure occur less and less often: this makes difficult the estimation of statistical reliability models. To which extent is it useful to define broad failure sets? We could define a set of failure times associated to the pitting of the inner races of ball bearings, a larger set with failure of ball bearings, a even larger set with failure associated to surface degradation, and so forth. What is the best set for reliability-based maintenance optimization?
Opportunistic maintenance policy with resource constraints
Opportunistic maintenance takes into account the unavoidable interruptions of a production process to perform preventive maintenance actions without productivity penalty. This work will investigate how a limitation of means (human power, tools, ...) could affect such a maintenance policy with respect to availability and cost.
Reduction of Major Risk Hazard as a Goal for Preventive Maintenance Policies
The Fault Tree Analysis defines a model for the occurence of a major risk (the explosion of a tank, for example). A software tool providing a sensitivity analysis to the probability of occurence of the top accident with respect to maintenance time interval is a deliverable of this work
Risk-Based Maintenance Strategy: how maintenance induces additional risk
There is a bilateral relation between risk and maintenance: on the one hand, a maintained industrial system will be preserved from degradation and will be safer. On the other hand, the maintenance activity exposes people to additional risks. This factor is included in the original approach of Maintenance Impact on Safety developed by Ms. Soumia Hadni in her PhD thesis.
Decision Aid Analysis of a Maintenance Policy through Electre method
The appropriate choice for a maintenance policy (corrective, preventive, opprotunistic, predictive, ...) of an industrial equipment is determined from a list of influence factors. This list aims to be as comprehensive as possible, but is by no means exhaustive or all-inclusive. The aim of this work is to give a recommandation is to convert the evaluations of these factors into a recommandation for a maintenance policy. Investigations around Electre methods are suggested but other methodologies could be used.
Selection of Criteria for a Decision Aid Analysis of a Maintenance Policy
A list of factors influencing the choice for a given maintenance policy (corrective, preventive, opprotunistic, predictive, ...) of an industrial equipment is currently developed in the frame of the PhD thesis of M. Nasser Mahamoud. This list aims to be as comprehensive as possible, but is by no means exhaustive or all-inclusive. The number of criteria used must be kept reasonable in regard of the available time for the collection of preferences. This work focusses on the selection of criteria, not on maintenance policies.
Risk aversion in maintenance activities
The aversion to risk may be modeled by a so-called utility function. This master's thesis will study how maintenance can be modeled as a lottery: maintain ou not maintain? That is the question inducing different cost / benefit expectations.
Analysis of degradation signals for predictive health monitoring
This work will investigate signal processing techniques to derive degradation-sensitive indicators from physical measurements (vibration signals, for example) and data mining to provide an estimation of residual lifetimes. Data from IEEE PHM Challenges will be used. Further details can be found on the home page of the challenge (http://eng.fclab.fr/ieee-phm-2014-data-challenge/).
Repair or Replace?
With this short question, we address the complex decision between the maintenance of an industrial equipment and a new investment taking into account a risk of obsolescence or a possible increase of performance. During the academic year 2015-2016, Clément Dutoit, a student in Mechanical Engineering performed a nice work on the subject that can be extended further to more complex systems..
Data Mining Techniques for Root Cause Analysis
The objective of root cause analysis is to identify the factors that resulted in the defects or failures and to determine what conditions need to be changed to prevent recurrence of similar harmful outcomes. Data mining techniques could indentify hidden causal relations. The work will consist in the definition of the data to be collected in maintenance reports and to evaluate the performance of different approaches.
Bayesian networks for reliability analysis
A Bayesian Network (BN) is a probabilistic graphical model (a type of statistical model) that represents a set of random variables and their conditional dependencies. For example, a Bayesian network could represent the probabilistic relationships between diseases and symptoms. The application of BN in maintenance engineerign rage from reliability analysis to diagnosis. This firt final year's thesis on the subject has to evaluate BN with respect to other techniques.
Last update: May 22, 2017 for the academic year 2017-2018. Feel free to contact me for more information and awarding of the subjects.