Pour savoir comment effectuer et gérer un dépôt de document, consultez le « Guide abrégé – Dépôt de documents » sur le site Web de la Bibliothèque. Pour toute question, écrivez à corpus@ulaval.ca.
 

Personne :
Morin, Michael

En cours de chargement...
Photo de profil

Adresse électronique

Date de naissance

Projets de recherche

Structures organisationnelles

Fonction

Nom de famille

Morin

Prénom

Michael

Affiliation

Université Laval. Département d'opérations et systèmes de décision

ISNI

ORCID

Identifiant Canadiana

ncf11859154

person.page.name

Résultats de recherche

Voici les éléments 1 - 7 sur 7
  • PublicationAccès libre
    Vine copula-based data generation for machine learning with an application to industrial processes
    (2022-12-02) Sexton, Jean-Thomas; Morin, Michael; Gaudreault, Jonathan
    Synthetic data generation of industrial processes exhibiting non-stationarity and complex, non-linear dependencies between their inputs and outputs is a challenging task. We argue that vine copula models are particularly well suited for this problem and present a method combining limited available data and expert knowledge in order to generate synthetic data by conditionally sampling from a C-Vine, a type of vine copula. We demonstrate our approach by generating synthetic data for a high-speed, sophisticated lumber finishing machine called a wood planer.
  • PublicationAccès libre
    Quality of sawmilling output predictions according to the size of the lot - The size matters!
    (2021-05-07) Morin, Michael; Gaudreault, Jonathan; Vallerand, Steve; Martineau, Vincent
    Lors de l'évaluation de modèles d'apprentissage automatique supervisé, on considère généralement le rendement de prédiction moyen obtenu sur les tests individuels comme mesure de choix. Toutefois, lorsque le modèle est destiné à prédire quels produits du bois seront obtenus lors du sciage de certains billots, c'est généralement la performance pour un lot complet qui importe. Dans cet article, nous montrons l'impact de cette nuance en termes d'évaluation du modèle. En fait, la qualité d'une prédiction (globale) s'améliore considérablement lorsque l'on augmente la taille des lots, ce qui offre un solide soutien à l'utilisation de ces modèles en pratique.
  • PublicationAccès libre
    Neural network architectures and feature extraction for lumber production prediction
    (Canadian Artificial Intelligence Association, 2021-06-08) Martineau, Vincent; Morin, Michael; Gaudreault, Jonathan; Thomas, Philippe; El-Haouzid, Hind Bril; Antonie, Luiza; Moradian Zadeh, Pooya
    We tackle the problem of predicting the lumber products resulting from the break down of the logs at a given sawmill. Although previous studies have shown that supervised learning is well suited for that prediction problem, to our knowledge, there exists only one approach using the 3D log scans as inputs and it is based on the iterative closest-point algorithm. In this paper, we evaluate the combination of neural network architectures (multilayer perceptron, residual network and PointNet) and log representation as input (industry know-how-based features, 2D projections, and 3D point clouds) in the context of lumber production prediction. Our study not only shows that it is possible to predict the output of a sawmill using neural networks, but also that there is value in combining industry know-how-based features and 3D point clouds in various network architectures.
  • PublicationAccès libre
    Explaining the results of an optimization-based decision support system : a machine learning approach
    (EDP Sciences, 2017-11-08) Morin, Michael; Thomopoulos, Rallou; Abi-Zeid, Irène; Léger, Maxime; Grondin, François; Pleau, Martin
    In this paper, we present work conducted in order to explain the results of a commercial software used for real-time decision support for the flow management of a combined wastewater network. This tool is deployed in many major cities and is used on a daily basis. We apply decision trees to build rules for classifying and interpreting the solutions of the optimization model. Our main goal is to build a classifier that would help a user understand why a proposed solution is good and why other solutions are worse. We demonstrate the feasibility of the approach to our industrial application by generating a large dataset of feasible solutions and classifying them as satisfactory or unsatisfactory based on whether the objective function is a certain percentage higher than the optimal (minimum) objective. We evaluate the performance of the learned classifier on unseen examples. Our results show that our approach is very promising according to reactions from analysts and potential users.
  • PublicationAccès libre
    A kNN approach based on ICP metrics for 3D scans matching : an application to the sawing process
    (Elsevier, 2021-11-09) Chabanet, Sylvain; Thomas, Philippe; El-Haouzi, Hind Bril; Morin, Michael; Gaudreault, Jonathan
    The Canadian wood industry use sawing simulators to digitally break a log into a basket of lumbers. However, those simulators tend to be computationally intensive. In some cases, this renders them impractical as decision support tools. Such a use case is the problem of dispatching large volume of wood to several sawmills in order to maximise total yield in dollars. Fast machine learning metamodels were recently proposed to address this issue. However, the approach needs a feature extraction step which could result in a loss of information. Conversely, it was proposed to directly make use of the raw information, available in the 3D scans of the logs typically used by a recent sawmill simulator, in order to retain that information. Here, we improve upon that method by reducing the computational cost incidental with the processing of those raw scans.
  • PublicationAccès libre
    Machine learning-based models of sawmills for better wood allocation planning
    (Elsevier, 2020-03-11) Morin, Michael; Gaudreault, Jonathan; Brotherton, Edith; Paradis, Frédérik; Rolland, Amélie; Wéry, Jean; Laviolette, François
    The forest-products supply chain gives rise to a variety of interconnected problems. Addressing these problems is challenging, but could be simplified by rigorous data analysis through a machine learning approach. A large amount of data links these problems at various hierarchical levels (e.g., strategic, tactical, operational, online) which complicates the data computation phase required to model and solve industrial problem instances. In this study, we propose to use machine learning to generate models of the sawmills (converting logs into lumber) to simplify the data computation phase for solving optimization problems. Specifically, we show how to use these models to provide a recommendation for the allocation of cutblocks to sawmills for a wood allocation planning problem without needing extensive sawing simulations. Our experimental results on an industrial problem instance demonstrate that the generated models can be used to provide high-quality recommendations (sending the right wood to the right mill). Machine learning models of the sawmill transformation process from logs to lumber allows a better allocation exploiting the strengths of the mills to process the logs in our industrial case.
  • PublicationRestriction temporaire
    An image is worth 10,000 points : neural network architectures and alternative log representations for lumber production prediction
    (Elsevier BV, 2023-06-15) Martineau, Vincent; Morin, Michael; Gaudreault, Jonathan; Thomas, Philippe; El-Haouzi, Hind Bril; Khachan, Mohammed
    Predicting the lumber products that can be obtained from a log allows for better allocation of resources and improves operations planning. Although sawing simulators make it possible to anticipate the production associated with a log, they do not allow processing many logs quickly. It was shown that machine learning can be used in place of a simulator. However, prediction quality is still lacking and information rich log representations are seldomly used in the literature for machine learning purposes. We compare several log representations that can be used (industry know-how-based features, 2D projections, and 3D point clouds) and several neural network architectures able to process these log representations (multilayer perceptron, residual network and PointNet). We also propose a new way to implement a loss function that improves prediction of sparse object count in regression. This new approach achieves a 15 improvement of F1 score compared to previous approaches.