Démarche outillée pour une modélisation et un diagnostic rapides des processus de patients

par Sina Namakiaraghi

Projet de thèse en Génie Industriel et Informatique

Sous la direction de Frédérick Benaben et de Franck Fontanili.

Thèses en préparation à l'Ecole nationale des Mines d'Albi-Carmaux , dans le cadre de SYSTEMES , en partenariat avec CGI - Centre de Génie Industriel (laboratoire) depuis le 01-10-2016 .


  • Résumé

    Cette thèse vise le développement d'une démarche outillée permettant l'obtention rapide d'un modèle du processus métier (BPM) puis d'un diagnostic guidé de sa performance en faisant appel à des outils tels que le RTLS, le Process Mining, Contrôle statistique de la qualité, et la simulation à événement discret. La modélisation fait l'objet de nombreux travaux de recherche pour ce qui concerne les langages et les outils de modélisation. Par contre, les travaux portant sur la démarche permettant de passer d'un processus réel à un modèle métier le plus fidèle possible sont moins nombreux. Aussi cette thèse vise le développement d'une démarche outillée qui accélèrera très nettement l'obtention et l'objectivité d'un tel modèle. La démarche visée repose sur trois fonctions principales qui constituent les axes des travaux de recherche : tracer, cartographier et évaluer. L'état de l'art met en évidence qu'une telle chaîne de valeur n'a jamais été couverte de manière totale par couplage de ces outils. Le premier obstacle à vaincre dans ce projet de recherche consistera donc à adapter et éprouver cette technologie afin de générer automatiquement un fichier log (ou fichier de traces) contenant tous les événements permettant de cartographier les parcours suivis par les entités traitées par le processus. Pour cette thèse, plusieurs terrains d'application sont envisagés, à la fois dans le domaine de la production de biens (logistique de production et supply chain dans tous les secteurs manufacturiers) et aussi dans le domaine de la production de soins (parcours patients en hôpital).

  • Titre traduit

    A Tool-Based Approach for modeling, analyzing, and diagnosing Patients' Processes.


  • Résumé

    This thesis aims at developing a tool-based approach enabling the user to obtain a fast business process modeling (BPM)technique which helps to discover patients' processes from location data. Later on, by applying the developed approach (DIAG) users are able to acquire a guided diagnosis of the process. The tools and methods that will be used are RTLS (Real Time Location System), Process Mining, Statistical Process Control, and Discrete Event Simulation. Having a realistic and relevant modeling of a real world process is an essential prerequisite in any continuous improvement approach. Nowadays, to reach that, it is necessary to proceed with data extraction from the information system. But this approach rarely gathers all the data which could be useful to the building of a process model. Some activities, particularly those which are not automated or instrumented, such as manual tasks, are not always tracked or recorded. For instance, in the case of a hospital, except recording patients on their arrival at the registration desk and then for discharging, the hospital information system does not gather any more data enabling to model the different pathways followed. It is the same problem in a manufacturing workshop or, on a larger scale, in a supply chain, where the different steps of operation, transfer, and inspection of products are not systematically recorded. When the useful data for modeling are not sufficient or even nonexistent, it is necessary to perform field observations, interviews of staff and data collection has done manually. That may be out of phase with the real world and moreover, involves a very high workload. Furthermore, if these observations are collected, they may be misinterpreted due to the limitation of a sample which does not ensure a true representation of the entire population. Consequently, the modeling will be inconsistent compared to the real processes, following a risk of a wrong diagnosis and then unsuitable and unsuccessful improvement solutions. Modeling is the topic of many research projects concerning modeling languages and tools. On the other hand, projects about the approach enabling to go from a real process to a business model as accurately as possible are more unusual. It is the reason why this Ph.D. thesis aims at developing a tool-based approach which will significantly accelerate the obtention of such a model. The developed methodology in this thesis (DIAG) is based on six main functions which are running within four different levels known as DATA, INFORMATION, AWARENESS, and GOVERNANCE. The function are Gathering, Data Interpreting, Modeling, Analyzing, Diagnosing, and prognosing. They are complementary, naturally linked and will be run in an integrated way using tools chosen for their relevance to the question: indoor positioning (called RTLS for Real Time Location System), Process Mining and Discrete Event Simulation. This tool-based approach aims at defining components to automate, accelerate and improve the design of a "digital clone", which must be a virtual and dynamic copy of the real system. Thanks to this digital clone, improvement solutions will be assessed before being implemented on the real system. It will also provide interesting opportunities in terms of agile and predictive control in the near future. This topic is close to ongoing research work on Industry 4.0. Several application areas are considered for this research project, in the industry (manufacturing process and supply chain of products or parts) and in healthcare (patient pathways in a hospital).