Over the past decade, environmental concerns and safety regulation have led to increasing demand for vehicles with higher passenger safety and fuel economy. This has spurred intensive research on advanced high strength steels (AHSS). The quench and partitioning (Q&P) heat treatment is a novel approach that has led to development of one group of third generation AHSS alloys. In recent years most of the studies on the Q&P process were dedicated to the effect of the heat treatment parameters on microstructural evolution and mechanical properties. However, micromechanical deformation behavior of constituent phases and damage evolution in Q&P steels are not fully understood. In this study, damage micromechanisms in a commercial QP980 were investigated with the aid of in-situ tensile tests under a scanning electron microscope (SEM) followed by local strain mapping using microscopic digital image correlation (µ-DIC) analysis so as to quantify the microstructural deformation of constituent phases. Nano-hardness measurements were conducted to correlate the amount of plastic deformation of each phases to its strength. Ex-situ tensile tests coupled with electron back scattered diffraction (EBSD) and X-ray diffraction (XRD) were conducted to study the influence of transformation induced plasticity (TRIP) of the retained austenite phase on microstructural damage and deformation. It was found that average local true strain in ferrite was approximately two times and three times greater than that of martensite and blocky retained austenite respectively, which was with the good agreement with nano-hardness measurements showing that retained austenite blocks was three times and two times harder than martensite and ferrite respectively. Damage in both ferrite and martensite starts at the same total strain; however, damage growth is faster in martensite leading to the formation of large cavities. The average local true strain ratio of ferrite to martensite decreases after total true strains higher than 0.1 and the reduction is more pronounced in regions with higher martensite volume fraction. EBSD results showed that at total true strain of 0.07 some of the retained austenite blocks located at the ferrite and martensite interfaces were almost fully transformed to martensite. According to XRD results at the point of necking 57% of retained austenite transformed to martensite. There is evidence of brittle cracking of large blocky retained austenite in regions with strain localization starting at relatively low strains but appear to have little impact on the final failure process. The good deformation ability of QP980 is attributed primarily to co-deformation of ferrite and martensite and secondarily to the TRIP effect. / Thesis / Master of Applied Science (MASc)

This paper presents a novel and efficient analytic framework for the performance analysis and capacity-assignment optimisation of a wireless GSM cell employing the Re-Use Partitioning (RUP) policy. RUP splits hierarchically the available bandwidth into multiple layers of frequencies and allows tighter frequency re-use in order to achieve a higher network capacity. In this context, a queueing network model (QNM) of a wireless cell is proposed consisting of a hierarchical layer configuration which is decomposed into individual GE/GE/c/c loss systems each of which is analysed in isolation via a more general maximum entropy (ME) state probability solution, subject to appropriate GE-type flow formulae and mean value constraints. Moreover, a new performance optimisation index is proposed as the weighted average non-blocking probability of traffic over all frequency layers. For illustration purposes, the proposed index is utilised to formulate and solve two capacity-assignment optimisation problems. Numerical examples are included to validate the relative accuracy of the analytic GE-type performance metrics against simulation and assess the optimal re-use partitioning policy of the available bandwidth.

The mobility of arsenic is controlled by the mineral source of arsenic and a host of biogeochemical factors such as pH, oxidation-reduction reactions, precipitation-dissolution reactions, adsorption-desorption processes, and the activity of microorganisms. In this study, sequential extraction and microscopic methods were used to evaluate arsenic partitioning in different phases in sediments and host rock at the Brinton arsenic mine (BAM) site. Results demonstrate spatial variability of arsenic in sediments, although the partitioning of arsenic in different phases was similar in both mine tailing and stream channel sediments. The sequential extraction results demonstrate that between 60 and 80 % of the total arsenic in sediments is associated with iron oxides, and an additional phosphate extraction showed that the majority (80%) of arsenic associated with the oxides is adsorbed. Imaging and analysis by scanning electron microscopy (SEM) and electron microprobe analysis (EMPA) show the presence of three arsenic bearing minerals, arsenopyrite, scorodite and arsenic-rich iron oxides, in both sediment and the host rock. In sediment, the minerals are present as individual grains, but in the host rock, they are present together, often with arsenopyrite at the core, surrounded by scorodite and/or elemental sulfur, which is rimmed by iron oxides. This spatial arrangement illustrates two weathering patterns of arsenopyrite, one that involves oxidation to form scorodite, which further dissolves to form arsenic-rich iron oxides; in this weathering series, sulfur presumably forms dissolved species which migrate away from the mineral. Another pattern, observed in several samples of host rock, involves formation of elemental sulfur in addition to scorodite and iron oxides. Results of this study have implications for arsenic mobility at the Brinton site and other mine sites where arsenic minerals are present. Although arsenopyrite is the main ore mineral, the main reservoir of arsenic in sediments is iron oxides. However, in the end it is the biogeochemical mechanism that releases arsenic from the mineral that will control arsenic mobility. In the case of iron oxides, desorption or reductive dissolution will promote arsenic release, whereas oxidizing conditions are required for arsenopyrite to release arsenic. / Master of Science

This thesis addresses the partitioning problem for configurable computing machines. Specifically, this thesis presents algorithms to partition chain-structured task graphs across configurable computing machines. The algorithms give optimal solutions for throughput and total execution time for these problems under constraints on area, pin count, and power consumption. The algorithms provide flexibility for applying these constraints while remaining polynomial in complexity. Proofs of correctness as well as an analysis of runtime complexity are given. Experiments are performed to illustrate the runtime of these algorithms. / Master of Science

Spatial analysis is of primary importance in forestry. Many factors that affect tree development have spatial components and can be sampled across geographic space. Some examples of spatially structured factors that affect tree growth include soil composition, water availability, and growing space. Our goals for this dissertation were to test the efficacy of spatial analysis tools in a forestry setting and make recommendations for their use. Reliable spatial analysis tools will lead to more effective statistical testing and can lead to useful mapping of spatial patterns. The data for this project is from simulated even aged loblolly pine stands (Pinus taeda L.). These simulated stands are grown at regular spacing and we impose a range of parameters on the stands to simulate many possible scenarios. In chapter 3 of this dissertation we perform a sensitivity analysis to determine if our methods are suitable for further research and applications. In chapter 4 we perform our analysis on more realistic data generated by a spatially-explicit stand simulator, PTAEDA 4.1. In chapter 3 we performed a statistical simulation of plantation stands without effects of competition and mortality. We used redundancy analysis (RDA) to quantify spatial variability, partial redundancy analysis (pRDA) to test for spatial dependence, and spatially constrained cluster analysis to map soil productivity. Our results indicated that RDA and pRDA are reliable methods and future evaluation is appropriate. The results from the spatially constrained cluster analysis were less clear. The success or failure of the clustering algorithm could not be disentangled from the success or failure of the selection criterion used to predict the number of clusters. Further investigations should address this concern. In chapter 4 we used PTAEDA 4.1, a loblolly stand simulator, to simulate a range of site conditions and produce data that we could use for analysis. The results showed that RDA and pRDA were not reliable methods and ready for the field. Spatially constrained cluster analysis performed poorly when more realistic data was used and because of this further use was uncertain. It was clear from the results that levels of variation and spatial pattern complexity of microsites influenced the success rate of the methods. Both RDA and pRDA were less successful with higher levels of variation in the data and with increased spatial pattern complexity. In chapter 5 we related the coefficient of variation from our simulations in (chapters 3 and 4) to two sets of real plot data, including a clonal set and open pollinated set. We then implemented a spatial analysis of the real plot data. Our spatial analysis results of the two comparable data sets were unaffected by genetic variability indicating that the primary source of variability across plots appears to be soil and other factors, not genetic related. / Ph. D.

As the IT industry marks a paradigm shift from the traditional system design model to System-On-Chip (SOC) design, the design of custom hardware, embedded processors and associated software have become very tightly coupled. Any change in the implementation of one of the components affects the design of other components and, in turn, the performance of the system. This has led to an integrated design approach known as hardware/software co-design and co-verification. The conventional techniques for co-design favor partitioning the system into hardware and software components at an early stage of the design and then iteratively refining it until a good solution is found. This method is expensive and time consuming. A more modern approach is to model the whole system and rigorously test and refine it before the partitioning is done. The key to this method is the ability to model and simulate the entire system. The advent of new System Level Modeling Languages (SLML), like SystemC, has made this possible. This research proposes a strategy to automate the process of partitioning a system model after it has been simulated and verified. The partitioning idea is based on systems modeled using Process Model Graphs (PmG). It is possible to extract a PmG directly from a SLML like SystemC. The PmG is then annotated with additional attributes like IO delay and rate of activation. A complexity heuristic is generated from this information, which is then used by a greedy algorithm to partition the graph into different architectures. Further, a command line tool has been developed that can process textually represented PmGs and partition them based on this approach. / Master of Science

No description available.

L'ordonnancement de tâches sur un système temps réel dur correspond à trouver une façon de choisir, à chaque instant, quelle tâche doit être exécutée sur le processeur pour que chacune ait le temps de terminer son travail avant son échéance. Ce problème, dans le contexte monoprocesseur, est déjà bien étudié et permet des applications sur des systèmes en production (aérospatiale, bourse etc.). Aujourd'hui, les plateformes multiprocesseur se sont généralisées et ont amené de nombreuses questions telles que l'utilisation efficace de tous les processeurs. Dans cette thèse, nous explorons les approches existantes pour résoudre ce problème. Nous étudions tout d'abord l'approche par partitionnement qui consiste à utiliser les recherches existantes en ramenant ce problème à plusieurs systèmes monoprocesseur. Ici, nous proposons un algorithme générique dont les paramètres sont adaptables en fonction de l'objectif à atteindre. Nous étudions ensuite l'approche par semi-partitionnement qui permet la migration d'un nombre restreint de tâches. Nous proposons une solution avec des migrations restreintes qui pourrait être assez simplement implémentée sur des systèmes concrets. Nous proposons ensuite une solution avec des migrations non restreintes qui offre de meilleurs résultats mais est plus difficile à implémenter. Enfin, les programmeurs utilisent de plus en plus le concept de tâches parallèles qui peuvent utiliser plusieurs processeurs en même temps. Ces tâches sont encore peu étudiées et nous proposons donc un nouveau modèle pour les représenter. Nous étudions les ordonnanceurs possibles et nous définissons une façon de garantir l'ordonnançabilité de ces tâches pour deux d'entre eux / The scheduling of tasks on a hard real-time system consists in finding a way to choose, at each time instant, which task should be executed on the processor so that each succeed to complete its work before its deadline. In the uniprocessor case, this problem is already well studied and enables us to do practical applications on real systems (aerospace, stock exchange etc.). Today, multiprocessor platforms are widespread and led to many issues such as the effective use of all processors. In this thesis, we explore the existing approaches to solve this problem. We first study the partitioning approach that reduces this problem to several uniprocessor systems and leverage existing research. For this one, we propose a generic partitioning algorithm whose parameters can be adapted according to different goals. We then study the semi-partitioning approach that allows migrations for a limited number of tasks. We propose a solution with restricted migration that could be implemented rather simply on real systems. We then propose a solution with unrestricted migration which provides better results but is more difficult to implement. Finally, programmers use more and more the concept of parallel tasks that can use multiple processors simultaneously. These tasks are still little studied and we propose a new model to represent them. We study the possible schedulers and define a way to ensure the schedulability of such tasks for two of them

Large-scale graph-structured datasets are growing at an increasing rate. Social network graphs are an example of these datasets. Processing large-scale graphstructured datasets are central to many applications ranging from telecommunication to biology and has led to the development of many parallel graph algorithms. Performance of parallel graph algorithms largely depends on how the underlying graph is partitioned. In this work, we focus on studying streaming vertex-cut graph partitioning algorithms where partitioners receive a graph as a stream of vertices and edges and assign partitions to them on their arrival once and for all. Some of these algorithms maintain a state during partitioning. In some cases, the size of the state is so huge that it cannot be kept in a single machine memory. In many real world scenarios, several instances of a streaming graph partitioning algorithm are run simultaneously to improve the system throughput. However, running several instances of a partitioner drops the partitioning quality considerably due to the incomplete information of partitioners. Even frequently sharing states and its combination with buffering mechanisms does not completely solves the problem because of the heavy communication overhead produced by partitioners. In this thesis, we propose an algorithm which tackles the problem of low scalability and performance of existing streaming graph partitioning algorithms by providing an efficient way of sharing states and its combination with windowing mechanism. We compare state-of-the-art streaming graph partitioning algorithms with our proposed solution concerning performance and efficiency. Our solution combines a batch processing method with a shared-state mechanism to achieve both an outstanding performance and a high partitioning quality. Shared state mechanism is used for sharing states of partitioners. We provide a robust implementation of our method in a PowerGraph framework. Furthermore, we empirically evaluate the impact of partitioning quality on how graph algorithms perform in a real cloud environment. The results show that our proposed method outperforms other algorithms in terms of partitioning quality and resource consumption and improves partitioning time considerably. On average our method improves partitioning time by 23%, decreases communication load by 15% and increase memory consumption by only 5% compared to the state-of-the-art streaming graph partitioning.

Many problems appearing in scientific computing and other areas can be formulated as a graph partitioning problems. Examples include data distribution for parallel computers, decomposition of sparse matrices and VLSI-design. In this survey we present the graph partitioning problem, describe some applications and introduce many of the algorithms used to solve the problem.