A major function of the central nervous system (CNS) during locomotion is the ability to maintain dynamic stability during threats to balance. The CNS uses reactive, predictive, and anticipatory mechanisms in order to accomplish this. Previously, stability has been estimated using single measures. Since the entire body works as a system, dynamic stability should be examined by integrating kinematic, kinetic, and electromyographical measures of the whole body. This thesis examines three threats to stability (recovery from a frontal plane surface translation, stepping onto and walking on a compliant surface, and obstacle clearance on a compliant surface). These threats to stability would enable a full body stability analysis for reactive, predictive, and anticipatory CNS control mechanisms. From the results in this study, observing various biomechanical variables provides a more precise evaluation of dynamic stability and how it is achieved. Observations showed that different methods of increasing stability (eg. Lowering full body COM, increasing step width) were controlled by differing CNS mechanisms during a task. This provides evidence that a single measure cannot determine dynamic stability during a locomotion task and the body must be observed entirely to determine methods used in the maintenance of dynamic stability.

Congestion control in wireline networks has been studied extensively since the seminal work by Mazumdar et al in 1998. It is well known that this global optimization problem can be implemented in a distributed manner. Stability and fairness are two main design objectives of congestion control mechanisms. Most literatures make the assumption that the number of flows is fixed in the network and each flow has infinite backlog for transfer in developing congestion control schemes. However, this assumption may not hold in reality. Thus, there is a need to study congestion control algorithm in the presence of dynamic flows. It is only until recently that short-lived flows have been taken into account. In this thesis, we study utility maximization problems for networks with dynamic flows. In particular, we consider the case where each class of flows arrives according to a Poisson process and has a length given by a certain distribution. The goal is to maximize the long-term expected system utility, which is a function of the number of flows and the rate (identical within a given class) allocated to each flow. Our investigation shows that, as long as the average work brought by the arrival processes is strictly within the network stability region, the fairness and stability issues are independent. While stability can be guaranteed by, for example, a FIFO policy, utility maximization becomes an unconstrained optimization. We also provide a queueing interpretation of this seemingly surprising result and show that not all utility functions make sense under dynamic flows. Finally, we use simulation results to show that our algorithm indeed maximizes the expected system utility.

The piano action is the mechanism that transforms the finger force applied to a key into the motion of a hammer that strikes a piano string. This thesis focuses on improving the fidelity of the dynamic model of a grand piano action which has been already developed by Hirschkorn et al. at the University of Waterloo. This model is the state-of-the-art dynamic model of the piano in the literature and is based on the real components of the piano action mechanism (key, whippen, jack, repetition lever, and hammer). Two main areas for improving the fidelity of the dynamic model are the hammer shank and the connection point between the key and the ground. The hammer shank is a long narrow wooden rod and, by observation with a high-speed video camera, the flexibility of this part has been confirmed. In previous work, the piano hammer had been modelled as a rigid body. In this work, a Rayleigh beam model is used to model the flexible behaviour of the hammer shank. By comparing the experimental and analytical results, it turns out that the flexibility of the hammer shank does not significantly affect the rotation of the other parts of the piano mechanism, compared with the case that the hammer shank has been modelled as a rigid part. However, the flexibility of the hammer shank changes the impact velocity of the hammer head, and also causes a greater scuffing motion for the hammer head during the contact with the string. The connection of the piano key to the ground had been simply modelled with a revolute joint, but the physical form of the connection at that point suggests that a revoluteprismatic joint with a contact force underneath better represents this connection. By comparing the experimental and analytical results, it is concluded that incorporating this new model significantly increases the fidelity of the model for the blows. In order to test the accuracy of the dynamic model, an experimental setup, including a servo motor, a load cell, a strain gauge, and three optical encoders, is built. The servo motor is used to actuate the piano key. Since the purpose of the motor is to consistently mimic the finger force of the pianist, the output torque of the motor is controlled. To overcome the problem associated with the motor torque control method used in previous work, a new torque control method is implemented on a real-time PC and a better control of the motor torque output is established. Adding a more realistic model of the piano string to the current piano action model and finding a better contact model for the contacts that happen between the surfaces that are made of felt (or leather), are two main areas that can be worked on in the future research. These two areas will help to further increase the fidelity of the present piano action model.

Most studies that have examined chloride-induced corrosion of steel in concrete have focused on sound concrete. However, reinforced concrete is seldom uncracked and very few studies have investigated the influence of cracked concrete on rebar corrosion. Furthermore, the studies that have examined the relationship between cracks and corrosion have focused on unloaded or statically loaded cracks. However, in practice, reinforced concrete structures (e.g. bridges) are often dynamically loaded. Hence, the cracks in such structures open and close which could influence the corrosion of the reinforcing steel. Consequently, the objectives of this project were (i) to examine the effect of different types of loading on the corrosion of reinforcing steel, (ii) the influence of concrete mixture design on the corrosion behaviour and (iii) to provide data that can be used in service-life modelling of cracked reinforced concretes. In this project, cracked reinforced concrete beams made with ordinary Portland cement concrete (OPCC) and high performance concrete (HPC) were subjected to no load, static loading and dynamic loading. They were immersed in salt solution to just above the crack level at their mid-point for two weeks out of every four (wet cycle) and, for the remaining two weeks, were left in ambient laboratory conditions to dry (dry cycle). The wet cycle led to three conditions of exposure for each beam: (i) the non-submerged region, (ii) the sound, submerged region and (ii) the cracked mid-section, which was also immersed in the solution. Linear polarization resistance and galvanostatic pulse techniques were used to monitor the corrosion in the three regions. Potentiodynamic polarization, electrochemical current noise and concrete electrical resistance measurements were also performed. These measurements illustrated that (i) rebar corroded faster at cracks than in sound concrete, (ii) HPC was more protective towards the rebar than OPCC even at cracks and (iii) there was a minor effect of the type of loading on rebar corrosion within the period of the project. These measurements also highlighted the problems associated with corrosion measurements, for example, identifying the actual corroding area and the influence of the length of rebar. The numbers of cracks and crack-widths in each beam were measured after the beam’s initial exposure to salt solution and, again, after the final corrosion measurements. HPC beams had more cracks than the OPCC. Also, final measurements illustrated increased crack-widths in dynamically loaded beams, regardless of the concrete type. The cracks in both statically and dynamically loaded OPCC and HPC beams bifurcated at the rebar level and propagated parallel to the rebar. This project also examined the extent of corrosion on the rebars and the distribution of corrosion products in the concrete and on the concrete walls of the cracks. Corrosion occurred only at cracks in the concrete and was spread over a larger area on the rebars in HPC than those in OPCC. The damage due to corrosion was superficial in HPC and crater-like in OPCC. Regardless of the concrete type, there was a larger distribution of corrosion products on the crack walls of the dynamically loaded beams. Corrosion products diffused into the cement paste and the paste-aggregate interface in OPCC but remained in the crack in HPC. The most voluminous corrosion product identified was ferric hydroxide. Elemental analysis of mill-scale on rebar which was not embedded in concrete or exposed to chlorides was compared to that of the bars that had been embedded in uncontaminated concrete and in cracked concrete exposed to chlorides. In uncontaminated concrete, mill-scale absorbed calcium and silicon. At a crack, a layer, composed of a mixture of cement paste and corrosion products, developed between the mill-scale and the substrate steel. Based on the results, it was concluded that (i) corrosion occurred on the rebar only at cracks in the concrete, (ii) corrosion was initiated at the cracks immediately upon exposure to salt solution, (ii) the type of loading had a minor influence on the corrosion rates of reinforcing steel and (iv) the use of polarized area led to a significant underestimation of the current density at the crack. It is recommended that the effect of cover-depth on (i) the time to initiation of corrosion and (ii) the corrosion current density in cracked concrete be investigated.

In a data centre, server clusters are typically used to provide the required processing capacity to provide acceptable response time performance to interactive applications. The workload of each application may be time-varying. Static allocation to meet peak demand is not an efficient usage of resources. Dynamic resource allocation, on the other hand, can result in efficient resource utilization while meeting the performance goals of individual applications. In this thesis, we develop a new interactive system model where the number of logon users changes over time. Our objective is to obtain results that can be used to guide dynamic resource allocation decisions. We obtain approximate analytic results for the response time distribution at steady state for our model. Using numerical examples, we show that these results are acceptable in terms of estimating the steady state probabilities of the number of logon users. We also show by comparison with simulation that our results are acceptable in estimating the response time distribution under a variety of dynamic resource allocation scenarios. More importantly, we show that our results are accurate in terms of predicting the minimum number of processor nodes required to meet the performance goal of an interaction application. Such information is valuable to resource provisioning and we discuss how our results can be used to guide dynamic resource allocation decisions.

Judgments of spatial relationships are often made when the object or observer are moving. Behaviourally, there is evidence that these ‘dynamic’ judgments of spatial extent differ from static judgments. Here I used three separate techniques for exploring dynamic judgments: first, a line bisection paradigm was employed to study ocular and pointing judgments of spatial extent while manipulating line length, position, speed, acceleration, and direction of scanning (Experiments 1-4); second, functional MRI (fMRI) was used to examine whether distinct brain regions were involved in dynamic versus static judgments of spatial extent (Exp 5); and finally, a mathematical and computational model of dynamic judgments was developed to provide a framework for interpreting the experimental results. In the behavioural experiments, substantial differences were seen between static and dynamic bisection, suggesting the two invoke different neural processes for computing spatial extent. Surprisingly, ocular and pointing judgments produced distinct bisection patterns that were uncorrelated, with pointing somewhat more impervious to manipulations such as scan direction and position than ocular bisections. However, a new experimental task for probing dynamic judgments (the ‘no line’ Experiment 4) found that scan direction can influence both hand behaviour. Functional MRI demonstrated that dynamic relative to static judgments produced activations in the cuneus and precuneus bilaterally, left cerebellum, and medial frontal gyrus, with reduced activation relative to static judgments observed in the supramarginal gyrus bilaterally. Dynamic bisections relative to a control condition produced activations in the right precuneus and left cerebellum, as well as in left superior parietal lobule, left middle temporal gyrus, and right precentral gyrus. It may be the case that velocity processing and temporal estimates are integrated primarily in the cuneus and precuneus bilaterally to produce estimates of spatial extent under dynamic scanning conditions. These results highlight the fact that dynamic judgments of spatial extent engage brain regions distinct from those employed to make static judgments, supporting the behavioural results that these are separate and distinct. Finally, a mathematical model was proposed for dynamic judgments of spatial extent, based on the idea that, rather than using an ‘all-or-none’ approach, spatial working memory actually takes about 100 ms to reach full representational strength for any given point in space. The model successfully explains many of the effects seen in the behavioural experiments including the effects of scan direction, velocity, line length, and position. In conjunction with the neuroimaging data, it also suggests why neglect patients may fail to show rightward bisection biases when making dynamic judgments of spatial extent. Overall, this work provides novel insights into how the brain executes dynamic judgments of spatial extent.

This thesis aims to reveal the magnitude of the income elasticity of health expenditure and the impact of non-income determinants of health expenditures in the Canadian Provinces. Health can be seen as a luxury good if the income elasticity exceeds unity and as a necessity good if the income elasticity is below unity. The motivation behind the analysis of the determinants of health spending is to identify the forces that drive the persistent increase in health expenditures in Canada and to explain the disparities in provincial health expenditures, thereby to prescribe sustainable macroeconomic policies regarding health spending. Panel data on real per capita GDP, relative price of health care, the share of publicly funded health expenditure, the share of senior population and life expectancy at birth have been used to investigate the determinants of Canadian real per capita provincial total, private and government health expenditures for the period 1975-2002. Dynamic models of health expenditure are analyzed via Generalized Instrumental Variables and Generalized Method of Moments techniques. Evidence confirms that health is far from being a luxury for Canada and government health expenditures are constrained by the relative prices. Results also cast doubt upon the power of quantitative analysis in explaining the increasing health expenditures.

Because of the highly complex structure of the load-sensing pump, its compensators and controlling elements, simulation of load-sensing pump system pose many challenges to researchers. One way to overcome some of the difficulties with creating complex computer model is the use of black box approach to create an approximation of the system behaviour by analyzing input/output relationships. That means the details of the physical phenomena are not so much of concern in the black box approach. Neural network can be used to implement the black box concept for system identification and it is proven that the neural network have the ability to model very complex behaviour and there is a well defined set of neural and neural network structures. Previous studies have shown the problems and limitations in dynamic system modeling using static neuron based neural networks. Some new neuron structures, Dynamic Neural Units (DNUs), have been developed which open a new area to the research associated with the system modelling.<p>The overall objective of this research was to investigate the feasibility of using a dynamic neural unit (DNU) based dynamic neural network (DNN) in modeling a hydraulic component (specifically a load-sensing pump), and the model could be used in a simulation with any other required component model to aid in hydraulic system design. To be truly representative of the component, the neural network model must be valid for both the steady state and the transient response. Due to three components (compensator, pump and control valve) in a load sensing pump system, there were three different pump model structures (the pump, compensator and valve model, the compensator and pump model, and the pump only model) from the practical point of view, and they were analysed thoroughly in this study. In this study, the DNU based DNN was used to model a pump only model which was a portion of a complete load sensing pump. After the trained DNN was tested with a wide variety of system inputs and due to the steady state error illustrated by the trained DNN, compensation equation approach and DNN and SNN combination approach were then adopted to overcome the steady state deviation. <p>It was verified, through this work, that the DNU based DNN can capture the dynamics of a nonlinear system, and the DNN and SNN combination can eliminate the steady state error which was generated by the trained DNN. <p>The first major contribution of this research was in investigating the feasibility of using the DNN to model a nonlinear system and eliminating the error accumulation problem encountered in the previous work. The second major contribution is exploring the combination of DNN and SNN to make the neural network model valid for both steady state and the transient response.

Previous research has shown that the productivity of sows housed in an Electronic Sow Feeding (ESF) system is affected by the housing management (static vs. dynamic), stage of gestation at mixing and parity. Familiarity has also been shown to affect the behaviour of group-housed sows. Thus, the objective of this experiment was to determine how the previously mentioned social factors affect the behaviour, physiology and productivity of sows housed in an ESF system. Sows were regrouped into either the static and dynamic pens. Within an introduction group, a subgroup of up to 24 focals sows was observed. The focal sows were chosen based on whether they were mixed pre vs. post-implantation (<12 vs. >46 days post-breeding), familiar vs. unfamiliar with group mates and parity (1st vs. 2nd and 3rd vs. 4th +). Aggression at mixing and at the feeder, injury scores, feeder entry order, space usage, salivary cortisol and farrowing productivity was recorded. The data was analyzed using Proc-Mixed and the General Model for SAS. Housing did not have a significant effect on the any of the parameters examined. Young sows had significantly more piglets born alive when housed in a dynamic system, while old sows had more piglets born alive when housed in a static system (p=0.03). Pre-implant sows initiated more aggressive encounters than post-implant sows (p=0.01). Post-implant sows ate later in the feeding cycle (p=0.03), rested on the slats more (p<0.001) and had higher salivary cortisol concentrations (p=0.0008). However, the cortisol concentrations increased throughout gestation for all sows (p<0.001). Familiarity did not have an effect on any of the variables examined except, familiar sows spent more time lying against the wall (p=0.03) and unfamiliar sows spent more time lying in the centre of the solid area of the pen (p=0.02). Old sows were involved in more aggressive encounters (p=0.04), spent more time fighting at mixing (p=0.02) and laid against the wall more (p<0.001). Young sows tended to received more scratches (p=0.07), ate later in the feeding cycle (p<0.001) and spent more time lying on the slats (p<0.001). Intermediate sows had significantly lower salivary cortisol concentrations (p=0.003). There was not a difference between the static and dynamic management systems. Sows should not be mixed until after embryonic implantation because they are more docile. The intermediate sows underwent the least amount of social stress due to their intermediate position within the dominance hierarchy.

A novel and promising method of software development is the interactive style of development, where code is written and incrementally tested simultaneously. Interpreted dynamic languages such as Ruby, Python, and Lua support this interactive development style. However, because they lack semantic analysis as part of a compilation phase, they do not provide type-checking. The programmer is only informed of type errors when they are encountered in the execution of the programfar too late and often at a less-informative location in the code. We introduce a typing system for Ruby, where types will be determined before execution by inferring principal typings. This system overcomes the obstacles that interactive and dynamic program development imposes on type checking; yielding an effective type-checking facility for dynamic programming languages. Our development is embodied as an extension to irb, the Ruby interactive mode, allowing us to evaluate principal typings for interactive development.