No description available.

Although the consistency, low cost, and high speed of machine vision systems make them suitable for many areas of manufacturing, there exist challenges for online inspection using machine vision systems due to the presence of variations in lighting, part position/orientation and part finish. In this thesis, a novel adaptive machine vision system based on pixel-by-pixel analysis and a neural network based vision system are presented to solve two challenging industrial inspection problems. Both of the vision systems stem from the idea of adaptive image processing to analyze an image with respect to its local properties. In the first part of the thesis, a pixel-by-pixel analysis based adaptive vision system is designed for an automotive water pump housing surface inspection problem. This vision system is used to inspect the machined surface of a die-cast part. The defects on this part may include pores, dents and scratches. This problem is challenging for several reasons. First, non-uniform surface finish on the parts produces large brightness variations that must be reduced using a controlled lighting work cell and adaptive camera control. Second, the defects can be subtle and less than lmm in diameter, while the surface is roughly 180mm by llOmm. Third, the surface often includes machining marks that appear similar to defects in the image, but are not considered defects. Finally, due to the manufacturing and fixturing variations, the size and location of the area to be inspected varies considerably, so that a simple fixed mask cannot be used to separate this area from the rest of the image. These challenges have been overcome by developing an adaptive machine vision system. This system includes custom-designed controlled lighting, and several software algorithms for adapting to the variations in surface quality and geometry. The system can detect defects as small as 0.15 mm. It has been tested with over 1,700 images that were collected at the factory. The majority of the defects were pores. These pores were correctly classified in 93% of the cases. In the second part of the thesis, a neural network-based vision system is developed for an automotive beam clip present/absent inspection problem. In this inspection problem, it is difficult to obtain the theoretical expression for the conditions of 'clip present', 'clip absent' and also for the clip orientation. Furthermore, there exist strong variations in this inspection problem, such as changing lighting conditions, environmental disturbances, clip locations and clip orientations. A CMAC neural network-based vision algorithm is developed to overcome these challenges. The CMAC neural network has the ability to learn fast and is suitable for real-time inspection applications. This vision system is demonstrated to correctly classify 100% of the cases, for the given automotive part images, after being trained for 151 seconds. / Thesis / Master of Applied Science (MASc)

The decreasing cost of chip manufacturing has greatly increased their distribution and availability such that sensors have become embedded in virtually all physical objects and are able to send and receive data -- giving rise to the Internet of Things (IoT). These embedded sensors are typically endowed with intelligent algorithms to transform information into real-time actionable insights. Recently, humans have taken on a larger role in the information-to-action path with the emergence of human-centric sensing. This has made it possible to observe various processes and infer information in complex personal and social spaces that may not be possible to obtain otherwise. However, a caveat of human-centric sensing is the high cost associated with high precision systems. In this dissertation, we present two low cost and high performing end-to-end solutions for human-centric sensing of physiological phenomena. Additionally, we present a post-hoc data-driven sensor synchronization framework that exploits independent, omni-present information in the data to synchronize multiple sensors. We first propose XTREMIS -- a low-cost and portable ECG/EMG/EEG platform with a small form factor that has a sample rate comparable to research-grade EMG machines. We evaluate XTREMIS on a signal level as well as utilize it in tandem with a Gaussian Mixture Hidden Markov Model to detect finger movements in a rapid, fine-grained activity -- typing on a keyboard. Experiments show that not only does XTREMIS functionally outperforms current wearable technologies, its signal quality is high enough to achieve classification accuracy similar to research-grade EMG machines, making it a suitable platform for further research. We then present SiCILIA -- a platform that extracts physical and personal variables of a user's thermal environment to infer their clothing insulation. An individual's thermal sensation is directly correlated with the amount of clothing they are wearing. Indeed, a person's thermal comfort is crucial to their productivity and physical wellness, and is directly correlated with morale. Therefore it becomes important to be aware of actions such as adding or removing clothing as they are indicators of current thermal sensation. The proposed inference algorithm builds upon theories of body heat transfer, and is corroborated by empirical data. SiCILIA was tested in a vehicle with a passenger-controlled HVAC system. Experimental results show that the algorithm is capable of accurately predicting an occupant's thermal insulation with a low mean prediction error. In the third part of the thesis we present CRONOS -- a sensor data synchronization framework that takes advantage of events observed by two or more sensors to synchronize their internal clocks using only their data streams. Experimental results on pairwise and multi-sensor synchronization show a significant drift improvement for total drift and a very low mean absolute synchronization error for multi-sensor synchronization. / Thesis / Doctor of Philosophy (PhD)

Functional data analysis is a statistical framework where data are assumed to follow some functional form. This method of analysis is commonly applied to time series data, where time, measured continuously or in discrete intervals, serves as the lo- cation for a function’s value. In this thesis Gaussian processes, a generalization of the multivariate normal distribution to function space, are used. When multiple processes are observed on a comparable interval, clustering them into sub-populations can provide significant insights. A modified EM algorithm is developed for cluster- ing processes. The model presented clusters processes based on how similar their underlying covariance kernel is. In other words, cluster formation arises from modelling correlation between inputs (as opposed to magnitude between process values). The method is applied to both simulated data and British Columbia coastal rainfall patterns. Results show clustering yearly processes can accurately classify extreme weather patterns. / Thesis / Master of Science (MSc)

Existing software engineering tools have proved useful in automating some aspects of the code review process, from uncovering defects to refactoring code. However, given that software teams still spend large amounts of time performing code reviews despite the use of such tools, much more research remains to be carried out in this area. This dissertation present two major contributions to this field. First, we perform a text classification experiment over thirty thousand GitHub review comments to understand what code reviewers typically discuss in reviews. Next, in an attempt to offer an innovative, data-driven approach to automating code reviews, we leverage probabilistic models of source code and graph embedding techniques to perform human-like code inspections. Our experimental results indicate that the proposed algorithm is able to emulate human-like code inspection behaviour in code reviews with a macro f1-score of 62%, representing an impressive contribution towards the relatively unexplored research domain of automated code reviewing tools. / Thesis / Master of Applied Science (MASc)

This thesis contains 13 blueprints that are not available in the digital version. The blueprints are very large and would require too much time to digitize. Please contact eScholarship@mcgill.ca to request more information.

Thesis advisor: Shakeeb Khan / Thesis advisor: Zhijie Xiao / This dissertation consists of three chapters demonstrating how the current econometric problems can be solved by using machine learning techniques. In the first chapter, I propose new approaches to estimating large dimensional monotone index models. This class of models has been popular in the applied and theoretical econometrics literatures as it includes discrete choice, nonparametric transformation, and duration models. A main advantage of my approach is computational. For instance, rank estimation procedures such as those proposed in Han (1987) and Cavanagh and Sherman (1998) that optimize a nonsmooth, non convex objective function are difficult to use with more than a few regressors and so limits their use in with economic data sets. For such monotone index models with increasing dimension, we propose to use a new class of estimators based on batched gradient descent (BGD) involving nonparametric methods such as kernel estimation or sieve estimation, and study their asymptotic properties. The BGD algorithm uses an iterative procedure where the key step exploits a strictly convex objective function, resulting in computational advantages. A contribution of my approach is that the model is large dimensional and semiparametric and so does not require the use of parametric distributional assumptions. The second chapter studies the estimation of semiparametric monotone index models when the sample size n is extremely large and conventional approaches fail to work due to devastating computational burdens. Motivated by the mini-batch gradient descent algorithm (MBGD) that is widely used as a stochastic optimization tool in the machine learning field, this chapter proposes a novel subsample- and iteration-based estimation procedure. In particular, starting from any initial guess of the true parameter, the estimator is progressively updated using a sequence of subsamples randomly drawn from the data set whose sample size is much smaller than n. The update is based on the gradient of some well-chosen loss function, where the nonparametric component in the model is replaced with its Nadaraya-Watson kernel estimator that is also constructed based on the random subsamples. The proposed algorithm essentially generalizes MBGD algorithm to the semiparametric setup. Since the new method uses only a subsample to perform Nadaraya-Watson kernel estimation and conduct the update, compared with the full-sample-based iterative method, the new method reduces the computational time by roughly n times if the subsample size and the kernel function are chosen properly, so can be easily applied when the sample size n is large. Moreover, this chapter shows that if averages are further conducted across the estimators produced during iterations, the difference between the average estimator and full-sample-based estimator will be 1/\sqrt{n}-trivial. Consequently, the averaged estimator is 1/\sqrt{n}-consistent and asymptotically normally distributed. In other words, the new estimator substantially improves the computational speed, while at the same time maintains the estimation accuracy. Finally, extensive Monte Carlo experiments and real data analysis illustrate the excellent performance of novel algorithm in terms of computational efficiency when the sample size is extremely large. Finally, the third chapter studies robust inference procedure for treatment effects in panel data with flexible relationship across units via the random forest method. The key contribution of this chapter is twofold. First, it proposes a direct construction of prediction intervals for the treatment effect by exploiting the information of the joint distribution of the cross-sectional units to construct counterfactuals using random forest. In particular, it proposes a Quantile Control Method (QCM) using the Quantile Random Forest (QRF) to accommodate flexible cross-sectional structure as well as high dimensionality. Second, it establishes the asymptotic consistency of QRF under the panel/time series setup with high dimensionality, which is of theoretical interest on its own right. In addition, Monte Carlo simulations are conducted and show that prediction intervals via the QCM have excellent coverage probability for the treatment effects comparing to existing methods in the literature, and are robust to heteroskedasticity, autocorrelation, and various types of model misspecifications. Finally, an empirical application to study the effect of the economic integration between Hong Kong and mainland China on Hong Kong’s economy is conducted to highlight the potential of the proposed method. / Thesis (PhD) — Boston College, 2024. / Submitted to: Boston College. Graduate School of Arts and Sciences. / Discipline: Economics.

Cette thèse s’intéresse à l’analyse de structures de machines électriques non conventionnelles destinées à la production d’électricité par des groupes électrogènes. Les topologies recherchées doivent utiliser moins de matières actives et/ou simplifier la procédure de fabrication par rapport aux machines actuelles.Une des structures est dédiée au système d’excitation de la machine. Elle possède un stator à griffes avec un bobinage toroïdal. Son dimensionnement est réalisé à l’aide d’un modèle de calcul par éléments finis en 3D, qui a été validé expérimentalement. Cette structure permet d’obtenir une réduction significative de la quantité de cuivre dans le bobinage.Les deux autres structures traitées concernent l’alternateur principal. Une machine synchro-réluctante à barrières deflux et à rotor bobiné a été proposée. L’objectif est d’augmenter la densité de couple grâce au couple de saillance. Elle est dimensionnée et comparée avec la structure conventionnelle. Ce cas met en évidence les limitations des structures proches de la structure actuelle.La dernière structure est une nouvelle topologie de machine électrique. Elle combine les caractéristiques des machines à flux axial et des machines à griffes, permettant de simplifier le système d’excitation. Une approche de modélisation originale a été développée pour l’analyse de cette topologie 3D en vue de son dimensionnement par optimisation. Le système d’excitation a un rôle important dans les performances de cette machine.Cette thèse constate aussi que l’évolution future des matériaux magnétiques pourrait tirer meilleur profit des structures non-conventionnelles avec des trajets de flux tridimensionnels. / The PhD project aims to analyse nonconventional structures of electric machines for electric power generation with diesel gensets. The researched topologies should use less active materials and/or simplify the manufacturing process compared to today’s machines.One of the structures is intended for the excitation system of the electric machine. It is composed of a claw-pole stator and a toroidal winding. It is designed using a 3D Finite Element model, previously validated by a prototype. This structure allows a significant reduction of the quantity of copper of the field winding.The other two structures under study are proposed for the main generator. A synchronous-reluctance machine with flux-barriers and a field winding in the rotor is analysed. The objective is to increase the torque density by increasing the reluctance torque. The designed machine is compared to the conventional structure. This case shows the limitations of structures similar to the current structure.The last structure is a new topology of electric machine. It combines the characteristics of claw-pole machines andaxial-flux machines. An original modelling approach is developed to analyse this 3D structure for its design based on an optimization algorithm. The study shows that the excitation system has a key role in its performances.This work also notes that the future evolution of magnetic materials should benefit the use of non-conventional structures with 3D flux paths.

La machine asynchrone présente un intérêt majeur par rapport aux autres types de machines(courant continu, synchrone, ...), sa robustesse, son faible coût de fabrication etd'entretien en sont les principales raisons. Cependant ces avantages ont longtemps été inhibés par la complexité de la commande de celle-ci.De nos jours de nombreux industrielles proposent des variateurs de vitesse pour la machine asynchrone offrant à la fois la souplesse de contrôle, et la qualité de la conversion électromagnétique,naturellement obtenues jusqu'alors avec la machine à courant continu et de la machine synchrone.Depuis quelques années les industrielles font face à une nouvelle problématique, qui est la suppression du capteur mécanique dans le processus de régulation de vitesse de la machine asynchrone. Les travaux de cette thèse, effectués dansle cadre d'un support CIFRE entre l'entreprise GS Maintenance et le laboratoire ECS-Lab EA 3649, ont été orientésvers la réalisation d'un système de contrôle commande d'un variateur industrieldédié aux machines asynchrones sans capteur mécanique. De ce point de vue, l'objectifpremier du travail de thèse, est la conception des techniques de détermination des grandeursmécaniques (vitesse) de la machine asynchrone en utilisant comme seules mesuresles grandeurs électriques. Ces techniques, utilisées pour remplacer l'informationdonnée par les capteurs mécaniques, sont parfois appelées capteurs logiciels.Une attention particulière est donnée au fonctionnement de la machine asynchrone sanscapteur mécanique à basse vitesse. Dans un second temps l'objectif étant d'illustrer lesintérêts technologiques d'un observateur basé sur la technique des modes glissants dansle but d'intégrer celui-ci dans le système contrôle commande d'un variateur industriel. / Induction machine includes main interests compared with others electricals machines like brushed DC Motor,or synchronus electric Motor.Its robustness, its low cost manufacture, and maintenance are major reason of its success.However, for long time this advantages inhibited because of induction machine control complexity.Nowadays,many industrial propose speed drives for induction machine giving both control flexibility, and electromagnetic qualited conversion, naturally obtained with DC motor, and synchronus electric Motor.For several years now, many manufacturers face to a new problematic, wich is sensorless induction machine control.This thesis work, carried out in concert with the firm GS Maintenance and ECS-Lab EA 3649 laboratory under CIFRE financement.This work focused on conception plant dedicated to sensorless industrial speed drive control for induction machine.From this point of view, at first glance this work proposes technical strategies to identify mechanical induction machine variables, by using only electrical measurements.This strategies used to stand in for informations from a mechanical sensor, are the so called software sensor.Specific attention has been paid to induction machine sensorless working at very low speed. Secondly, we propose to illustrate the interest of a second order Sliding Mode Observer with final aim to be integrated into an industrial speed drive

In the past few years, evaluating on adversarial examples has become a standard procedure to measure robustness of deep learning models. Literature on adversarial examples for neural nets has largely focused on image data, which are represented as points in continuous space. However, a vast proportion of machine learning models operate on discrete input, and thus demand a similar rigor in understanding their vulnerabilities and robustness. We study robustness of neural network architectures for textual and graph inputs, through the lens of adversarial input perturbations. We will cover methods for both attacks and defense; we will focus on 1) addressing challenges in optimization for creating adversarial perturbations for discrete data; 2) evaluating and contrasting white-box and black-box adversarial examples; and 3) proposing efficient methods to make the models robust against adversarial attacks.