Innovative Projects Realized

Transcritical CO2 Pulverization

Crushing and grinding rock is the largest consumer of energy at a mining operation. Breaking rock by expansion within is a lower energy process than crushing or impact from outside, since rock tensile strength is significantly lower than rock compressive strength. This research intends to explore and develop a novel method for rock comminution using a new form of explosive pulverization or shattering based on transcritical CO2 cycling rather than traditional compressive approaches. The intent of this project is to both reduce comminution energy as well as reduce equipment wear. Both result in reduced energy consumption with the associated social and environmental benefits that come from reducing energy usage.

Employing Raman Spectroscopy for monitoring a vaccine manufacturing process

Vaccine antigen production is a process that entails numerous variables. In order to have a consistent and robust process, monitoring of process parameters and controlling output variables within a certain range is the best practice. To accomplish this objective, analytical tools are used, on-line, off-line, at line. Real time monitoring of the processes is advantageous as operating parameters can always be adjusted to keep the process in check. For the sake of efficiency, it would be reasonable to determine those parameters that are correlated to the productivity obtained at the end of the process. On-line measurements would facilitate calculating and defining these parameters. There are non-invasive tools that can collect continuous measurements without interfering with the integrity of the process. Raman probes and NIR probes are examples to this type of approach. In this study the intern will be tasked to use these tools to identify any patterns which can be linked to the final productivity. The instrument will be tested in all unit operation steps to map the evolution of measured metabolites/substrates upstream and downstream of the process. Mathematical modelling, such as multivariate analysis, Principal Component Analysis, Partial least Squares will be used in synthesizing and interpreting the data.

Using document embedding for code completion

Dans les recherches récentes en génie logiciel et plus particulièrement en compréhension du logiciel, une grande partie des approches utilisent des techniques issues de la communauté du traitement automatique des langues naturelles. L’utilisation de ces techniques de traitement des langues au logiciel s’est avérée être efficace pour capturer des informations textuelles et sémantiques dans des programmes informatiques. Des travaux précédents ont montré que l’utilisation de ces informations sémantiques extraites automatiquement sur de larges ensembles de données ont des impacts bénéfiques dans la production d’outils d’aide au développement logiciel.
L’objectif de ce projet de recherche est d’appliquer des techniques de traitement des langues et d’apprentissage automatique au logiciel afin d’automatiser des tâches du développement logiciel (ex: suggérer au programmeur le code qu’il devrait écrire étant donné ce qu’il a écrit précédemment)

Optimization of Auxiliary Electrical Operation of a Handy-Size Bulk Carrier

Environmental regulations that aim to reduce greenhouse gas (GHG) emissions from new ships will become effective by 2025 [1]. A lot of solutions are evaluated by the industry to reduce their emissions including shore power, a process that aims to reduce the ships emissions in ports by turning ship engines off and supplying the ship by the energy available on shore. A lot of issues reside within this solution: the high-power demand on the microgrid during loading and unloading process, the lack of standardization for the electrical connection, the capital expenditure (CAPEX) needed to implement a shore power system and more. Therefore, the project is about detailing the energy demand of a handy-size FEDNAV [2] bulk carrier multi-sources multi-generators of 35,000 Dead Weight Tonnage (DWT) with one to four deck cranes while at the port, understanding and mathematically modelling the different energy sources that can interact with the ship, providing a microgrid solution for the port and producing a technical-economic study on the viability of the best solution.

Operationalizing quality evaluation for heterogeneous legacy systems

Most large companies use software systems that were developed over decades. They comprise

systems that are in different programing languages and must all communicate with one another to solve

bUSiness problems. As these systems become increasingly large and complex, there is a need to develop

tools and techniques to ensure their quality. Current state-of-the-art techniques deal with systems individually,

yet many quality Issues arise due to the interactions between different systems. The goal of this project is to

develop techniques to analyse these complex heterogeneous systems and evaluate their quality. The two

first interns will to provide ways to recover the interactions between systems. The second two will find and

categorize good and bad patterns corresponding to these interactions. The final pair will assess the impact of

these patterns on quality.

Exploring and modeling human cognition through deep learning

Some of us recall our dreams very often, while other hardly ever remember them. The neuroscience of sleep and dream research is a thriving field with still many unanswered questions. Differences in brain network dynamics between individuals with high versus low dream recall rates, are still poorly understood. In this internship, we will address this question using state-of-the-art machine learning tools. In particular, we will frame it as a classification problem where we apply deep convolutional neural networks (CNN) to sleep EEG recordings in order to predict whether subjects belong to a high or low dream recall group (HDR and LDR resp.). In addition, to explore the neural properties that the AI approach used for successful discrimination, we will test several techniques for the visualization of the feature space learned by the network. This project lies at the intersection between AI and Neuroscience and carries the potential to advance our understanding of dream-related cognitive processes y combining EEG and data-driven machine learning tools.

Evaluation of the reinforcing capabilities of cellulose filaments in elastomeric matrices

Performance BioFilaments Inc. (PBI) is focused on the development of commercial applications for cellulose filaments (CF), one of the world’s most exciting and abundant biomaterials. Cellulose filaments have unique performance-enhancing properties with significant potential to improve a wide array of consumer and industrial products. Derived from wood fibre – a renewable natural resource – cellulose filaments optimize the strength, stability, flexibility, and longevity of a variety of materials. One of the potential applications for cellulose filaments is to use them as reinforcing agents in polymeric matrices. In partnership with PBI, the proposed research aims to evaluate the best incorporation strategy for cellulose filaments as a suitable reinforcing additive in elastomeric matrices such as styrene-butadiene and natural rubbers, used for transportation applications, and to evaluate potential additives to improve loss of elongation. Gaining such insights will enable PBI to develop practical solutions to improve the elongation behavior of cellulose filled rubbers.

Assurance qualité en continu pour la maintenance d’applications web

Les sites et les applications web contiennent souvent des bugs d’affichage : boutons mal placés ou non fonctionnels, éléments partiellement cachés par d’autres, etc. Ces bugs peut relever d’un simple problème esthétique, ou, de manière plus sévère, compromettre la fonctionnalité de l’application

Interactive Video Tutoring Module

The online video-based tutoring industry is growing fast but there is a lack of interaction between student and tutor. Some survey studies suggest that interactive video-based tutoring like giving bookmarks, hints, nudges, and quizzes in between video lectures helps the student in improving concentration and learning. We propose an interactive video tutoring module which given a student profile and past behavior (in different videos), predict the time points where a student would pause a video (or bookmark it), speed up or return to a video after exercise, or a combination of these. This can be formulated as a bookmark recommendation system and a standard approach to solve this problem is Collaborative Filtering. We will use a baseline based on matrix factorization, which is a class of collaborative filtering algorithm. Later, we aim to experiment with more advanced techniques that may use deep learning.

Stand-alone Low-cost Millimeter-Wave/Terahertz Security Inspection Imaging System

The exponential growth of e-commerce in conjunction with threats pertaining to the drug trade, the online pharmacy, and terrorism are creating a significant need for inspection sites (i.e. postal processing centers) to upgrade and expand their capabilities. Therefore new solutions are needed. In the frame work of this project, our goal is to develop a low-cost security imaging inspection system that can see not only through opaque postal packages but can also differentiate various objects and materials from each other within a reconstructed image of package interior without the need for opening them at the border or mail processing centers.

Création d’un indice de santé numérique pour les restaurateurs

Le projet a pour but la création d’un indice de la santé numérique des restaurateurs ayant recours au service de UEAT. Grâce aux données que génèrent les systèmes implantés chez les restaurateurs, il est maintenant possible d’obtenir un portrait relativement fiable de la santé numérique d’un client de UEAT. Le filon principal du projet est de fournir un rapport périodique aux restaurateurs indiquant l’état de leur face à leur compétiteur et leur propre filiale (si franchisée) en présentant différentes métriques de performance. Ces métriques seront une façon simple pour le client de savoir quels aspects de ses ventes numériques doit-il tenter d’améliorer. Puisque le modèle d’affaire de UEAT est basé sur une rémunération par transaction effectuée sur les différents sites web de leurs clients, une augmentation de revenu pour le restaurateur est directement reliée à la rentabilité de l’entreprise.

Selectomics to monitor and predict the emergence of resistance to new therapeutic approaches

Multiple antibiotic resistance has increased over the past decades, challenging our ability to treat

bacterial infections and thwarting our ability to develop new antimicrobial agents. Many resistance

genes have not evolved within the pathogenic isolates but were acquired by lateral transfer. We

recently showed that genes conferring glycopeptide resistance are highly prevalent in the human flora.

Some of these genes are present in novel commensal anaerobic species of the gut suggesting that these

bacteria may serve as a reservoir for resistance genes. We will use the human gut microbiome to

monitor the emergence of resistance genes that could have the potential to be transferred to pathogens

and address the question of how these reservoirs of resistance detenninants respond to antibiotic

pressure. The selectomics strategy generated by this project will constitute useful tools to assess the

potency of novel chemical compounds to select for resistance.