polito.it
Politecnico di Torino (logo)

Data-driven Overall Equipment Effectiveness modelling and optimal scheduling in GSK Oak Hill (NY, US) Production Site

Angelo Cinquemani

Data-driven Overall Equipment Effectiveness modelling and optimal scheduling in GSK Oak Hill (NY, US) Production Site.

Rel. Roberto Fontana, Gueorgui Mihaylov. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Gestionale (Engineering And Management), 2022

[img]
Preview
PDF (Tesi_di_laurea) - Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives.

Download (6MB) | Preview
Abstract:

Overall Equipment Effectiveness (OEE) is a core Key Performance Indicator (KPI) in industrial manufacturing. The downtime experienced by production resources negatively affects it. Root cause analysis from data retrieved highlights that combinations of products and packaging materials impact unplanned maintenance time of manufacturing and packaging lines. The main idea of the project is to create an optimal scheduling model whose aim is to maximise the resource utilisation recommending combinations of products, mixers, packaging materials and packaging lines. It has been developed with the constant support of GlaxoSmithKline/Haleon R&D Team. The development process starts collecting and analysing real datasets about GSK Oak Hill (NY, US) Production Site. Process Mining is exploited in order to discover the end-to-end production flow. Then, information about hard and soft constraints, which the model enforces, are retrieved. They are related to production, maintenance, changeovers, shifts and breaks, inventory management and spare capacity. The objective function minimizes the machine downtime. The model is designed and developed using Python as programming language and exploiting the library "ortools.sat.python" as background optimization environment. The solution space is explored using Greedy Insertion. The heuristic allocates production blocks minimizing their impact on the objective function. Blocks are scheduled starting from packaging lines and going backward. Relevant insights emphasize that metaheuristics such as Large Neighborhood Search (LNS) can lead to outstanding results in improving the solution quality [1, Pisinger et al., 2010]. LNS works implementing a destroy and repair method. As a result, some blocks of the current solution are removed and located in new positions [1, Pisinger et al., 2010]. Its application often allows to escape from local optima and hopefully finding a global optima solution. Furthermore, a practical implementation of Agglomerative Hierarchical Clustering is developed. It aims at identifying product families based on operational flows. This would make the procedure of inserting meta-data more robust, leading to higher Data Integrity which is a relevant concern in the pharmaceutical industry.

Relators: Roberto Fontana, Gueorgui Mihaylov
Academic year: 2022/23
Publication type: Electronic
Number of Pages: 170
Subjects:
Corso di laurea: Corso di laurea magistrale in Ingegneria Gestionale (Engineering And Management)
Classe di laurea: New organization > Master science > LM-31 - MANAGEMENT ENGINEERING
Ente in cotutela: GlaxoSmithKline Consumer Healthcare (REGNO UNITO)
Aziende collaboratrici: GLAXOSMITHKLINE CONSUMER HEALTHCARE srl
URI: http://webthesis.biblio.polito.it/id/eprint/25398
Modify record (reserved for operators) Modify record (reserved for operators)