CEDoc-UM6P-AI MOV : Optimization of Maintenance Planning based on Full-time

Mohammed VI Polytechnic University is an institution dedicated to research and innovation in Africa and aims to position itself among world-renowned universities in its fields

The University is engaged in economic and human development and puts research and innovation at the forefront of African development. A mechanism that enables it to consolidate Morocco’s frontline position in these fields, in a unique partnership-based approach and boosting skills training relevant for the future of Africa.

Located in the municipality of Benguerir, in the very heart of the Green City, Mohammed VI Polytechnic University aspires to leave its mark nationally, continentally, and globally.

1 Context

Maintenance is essential for slowing down component degradation in complex systems, hence reducing the risk of unexpected failures and minimizing costly downtimes. The maintenance can mainly be achieved by anticipating failures (e.g. predictive maintenance). In the context of industry 5.0, companies are not only adapting to new methods of producing goods and services (e.g., more flexible/reconfigurable), but evolving their maintenance activities, methods, and tools to promote the agility and resilience necessary for success. Particularly, with the advancement of the internet of things, opening up new possibilities to incorporate Artificial Intelligence techniques in PHM (Prognostics and Health Management). The PHM aims to monitor, diagnose, and prognostic the health status of industrial equipment. The availability of sensor measurements provides the opportunity to analyze degradation trends over time such as Remaining Useful Life (RUL), and to plan maintenance accordingly while continuously receiving real-time data. Predictive maintenance is based on predictive models that are developed using mainly vibration data to estimate the degradation level, and eventually the failure time. Basically, RUL is a reliable indicator to evaluate state of health of an equipment [1]. However, decisions related to maintenance operation dates are still not fully optimized and are made independently from prognostic results. This means that decision-makers must plan maintenance interventions subject to the true operational state and degradation of the system. Indeed, optimization maintenance planning is basically based on set of constraints, containing among others, resource limitation existing and key indicators such as RUL to evaluate the degradation level of a given physical asset or complex system. The objectives of the optimization models are: to minimize maintenance costs, maximize the lifetime of components, and limit the probability of a failure based on RUL prognostics to guide the scheduling of maintenance tasks [2]. One of the main challenges in predictive maintenance is to obtain reliable RUL prognostics and to integrate them into maintenance planning. In the literature, several decision support models including maintenance strategies and optimization algorithms have been developed. However, the development and implementation of adequate solutions on an industrial scale still lack foundations, methods, and tools [3].

2 Research Objectives

The first objective of this thesis is to develop prognostic approaches to predict the RUL of critical components using machine learning algorithms and/or stochastic processes. The second objective of this thesis is to investigate embedding Prognostic in maintenance planning. In recent years, researchers have begun to focus on predicting and then optimizing to enhance traditional optimization methods [4, 5]. The idea is to use real-time collected data from sensors to provide more accurate health state predictions using modified machine learning algorithms with a loss function correlated to the output of the optimization engine. Indeed, the model developed consists of using the prognosis results obtained from the previous step while respecting the properties of the system including equipment/machine, and constraints related to operation and logistical support (e.g., availability of spare parts, maintenance budget, limited maintenance duration). A multi-criteria optimization (e.g., minimizing maintenance costs and/or maximizing lifetime of component) with constraints will be developed using conventional optimization algorithms to find the optimal maintenance schedule.

3 Admission Criteria

The PhD position is available at Ai movement, the International Center for Artificial Intelligence of Morocco of UM6P. Applicants with excellent academic credentials must be holders of a Master’s, an engineering or an equivalent recognized degree with good skills in applied mathematics, in relation to optimization, operations research, and machine learning. The candidate should also be excellent in programming in (Python, Java or C++), should have soft skills, and be fluent in English and French languages. Letters of recommendation are welcome.

References

[1] Yigit A Yucesan, Arinan Dourado, and Felipe AC Viana. A survey of modeling for prognosis and health management of industrial equipment. Advanced Engineering Informatics, 50:101404, 2021.

[2] Rim Louhichi, Mohamed Sallak, and Jacques Pelletan. Avenues for future research on predictive maintenance purposes in terms of risk minimization. In 30th European Safety and Reliability Conference (ESREL 2020), pages 3461-3468, 2020.

[3] Yang Hu, Xuewen Miao, Yong Si, Ershun Pan, and Enrico Zio. Prognostics and health management: A review from the perspectives of design, development and decision. Reliability Engineering & System Safety, 217:108063, 2022.

[4] Adam N Elmachtoub and Paul Grigas. Smart “predict, then optimize”. Management Science, 68(1):9-26, 2022.

[5] Toon Vanderschueren, Tim Verdonck, Bart Baesens, and Wouter Verbeke. Predict-then-optimize or predict-and-optimize? an empirical evaluation of cost-sensitive learning strategies. Information Sciences, 594:400-415, 2022.

UM6P.