Mechanical Engineering - Theses

Permanent URI for this collection

Search Results

Now showing 1 - 10 of 42
  • Item
    Thumbnail Image
    The Effect of Crutch Usage on Human Walking with Lower Limb Exoskeleton
    Chen, Xin ( 2021)
    Lower limb exoskeleton robots allow patients with spinal cord injury (SCI) to perform overground gait. However, most lower-limb exoskeleton robots require users to use crutches to balance themselves during walking. It has been observed that long-term use of crutches will lead to potential harm to the shoulder joints due to the repetitive high load on the shoulder. Investigations of the shoulder reaction force experienced during exoskeleton use are needed to understand this effect better. Studies have shown that different crutch gaits can lead to the variation of shoulder reaction force. This study compares the effects of different gait patterns on the shoulder reaction force in an experiment involving six able-bodied individuals walking with the exoskeleton robot. Specifically, the shoulder reaction force during exoskeleton walking is studied with two commonly observed gait patterns: (1) the four-point parallel crutch gait (Gait-P) and (2) the four-point reciprocal crutch gait (Gait-R). Contact forces between the ground and crutches were recorded and indicated the shoulder reaction force. Three metrics (maximum rate-of-loading (MaxROL), maximum force (MaxForce), force- time-integral (FTI)) to evaluate the measured force have been adopted, and Wilcoxon signed-rank test has been used to check the difference significance. The results suggest a significant difference between FTI of different gait patterns, and Gait-R shows a higher accumulated force load on the shoulder joint. This difference in the accumulated load can be even larger with time, but current research results cannot estimate its effect on leading to shoulder injury. The result also indicates that there may be person to person variation in the metrics. The effect of changing the gait type on three metrics can be stronger, weaker, and sometimes are opposite to the effect shown at the population level. This difference may be caused by the subjects’ balance control ability or walking habits. In conclusion, the results suggest it is beneficial for SCI users to test the effect of each gait before using the exoskeleton robot in the long run. Otherwise, the parallel gait (Gait-P) was found to cause lower load on the shoulder at the population level.
  • Item
    Thumbnail Image
    Inducing human movement pattern change
    Xu, Yangmengfei ( 2021)
    Human’s movement pattern shaping is widely used in neurorehabilitation and sports training. Recent studies have shown that robotic device has its potential to become an efficient tool for clinicians to induce this change. To understand human's movement, different computational models were proposed and studied to explain how human resolves their redundancy. Although some arguments are still existing, the general idea of optimization has been well accepted. Based on these computational models, the motor learning studies showed that through practice in the new environment, the reward-based optimization could drive human to search for a better movement pattern 1) to maximize the performance and 2) to minimize the motor cost. Leveraging this optimization idea in human motor learning, this work aims to induce the movement pattern changes in an experimental setup solely relying on the motor cost without any explicit kinematic error. In this strategy, the intervention space and adaptation space are decoupled: while the force field only applies to the hand linear velocity, the adaptation is expected to happen in the redundant arm joint space (\textit{i.e.} the swivel angle). This work, therefore, explores the following topics: * Investigating the feasibility of inducing human motor adaptation in the redundant space by providing a task space intervention without explicit error feedback or instruction; * Evaluating the contribution of a progressively changing goal in this implicit motor adaptation, assuming that this adaptation may be further promoted through subtle prompts to explore the cost space; * Demonstrating a motor cost analysis based on the upper limb kinematics and dynamics model to validate the relationship between observations and motor cost.
  • Item
    Thumbnail Image
    Measurements and analysis of turbulent boundary layers subjected to streamwise pressure gradients
    Romero, Sylvia ( 2021)
    This thesis details the changes in structure between an adverse pressure gradient turbulent boundary layer (APG TBL), zero pressure gradient (ZPG) TBL, and channel flow by means of the mean momentum balance (MMB). In order to understand the effects of a pressure gradient on a turbulent boundary layer, aspects of the physical flow are studied via mean statistics, turbulence measurements, and spectra analysis. This study uses new experimental measurements that are conducted along an APG ramp as well as measurements downstream of the ramp insert to study the flow as it relaxes towards equilibrium. In the present experimental set-up the boundary layer is under modest APG conditions, where the Clauser pressure-gradient parameter $\beta$ is $\leq 1.8$. Well-resolved hot-wire measurements are obtained at the Flow Physics Facility (FPF) at the University of New Hampshire. Comparisons are made with ZPG TBL experimental data at similar Reynolds number and computational data at lower Reynolds number. Present measurements are also compared to existing APG TBL lower Reynolds number experimental and computational data sets. Finally, it is shown how these findings relate to an analytical transformation. The primary takeaways from the MMB analysis presented herein are $(i)$ distance-from-the-wall scaling can result from an assumption of self-similar mean dynamics, and does not require primacy of a single velocity scale, and $(ii)$ distance-from-the-wall scaling does not necessarily imply a logarithmic mean velocity profile; a power-law velocity scale hierarchy along with self-similar mean dynamics simultaneously produces distance-from-the-wall scaling \textit{and} a power law mean velocity profile. The choice to refer to the (potentially) self-similar subdomain as the `inertial sublayer' in the present study (rather than the `log' layer) is therefore deliberate.
  • Item
    Thumbnail Image
    Discrete Tone Identification and Heat Transfer of Under-expanded Impinging Jets
    Li, Minghang ( 2021)
    Inspired by the rocket launching process and the hovering mode of the Short Take-off and Vertical Landing aircraft (STOVL), the high-amplitude discrete tones and the heat transfer performance of under-expanded impinging jets have been studied using data from large-eddy simulation (LES). Regarding the discrete tones, the characteristics of the upstream-propagating waves and their travelling paths are first investigated. By examining the existing two resonance models: a classical aeroacoustic feedback model and a modified model combining the classical model and a vortex sheet model, the upstream-propagating waves are found to be associated with both the free-stream acoustic waves and the intrinsic jet modes. This finding is further confirmed by a quantitative comparison between the current data and the eigenfunction distribution provided by the vortex sheet model. With respect to the travelling path, it is found that the upstream-propagating waves can travel inside the shear layer and approach the nozzle lip where the flow is not supersonic. In addition to the downstream-travelling Kelvin-Helmholtz (K-H) instability waves, the current study observes another family of downstream-travelling waves at a higher wavenumber. This family of waves is found to be the product of the interactions between the K-H waves and the shock structures only for jets with an axisymmetric mode. In order to identify how the discrete tones are generated, a geometrical acoustic model is first introduced and applied to the impinging jets considered in this study. It is found that the model can well capture the acoustic leakage, which is produced by the interaction between the large-scale vortical structures and the shock tip. However, the leakage locations are found not to be the dominant noise sources based on the predictions of the classical feedback model. Hence, a ray tracing and cross-correlation method is proposed. It is found that the dominant noise sources are located at the impingement plate, which also agrees well with the instantaneous observations. Finally for the jet impingement heat transfer, the underlying physics of the peaks in the heat transfer coefficients is studied. Through the observations of the near-wall instantaneous fields (pressure fluctuation and vorticity norm), the analysis of the near-wall statistics and the reconstructed amplitude and phase fields at the discrete tone frequencies, it is found that the first peak in the heat transfer coefficient is related to the jet impingement or jet mixing, and is driven by the impingement of the non-periodic turbulent structures. A second peak in the heat transfer coefficient, which is only observed for the jet with higher compressibility, is found to be associated with the reattachment of the recirculation bubbles in the wall jet.
  • Item
    Thumbnail Image
    Large-eddy simulation of a natural gas DISI engine
    Yosri, Mohammadreza ( 2021)
    Direct Injection (DI) of natural gas in Spark Ignition (SI) engines can improve efficiency and reduce CO2 emissions compared with gasoline-fuelled equivalents. In the CNG DISI engine, as compressible gas flows through the injector's small passages and into the combustion chamber, complex phenomena such as choking, shock waves, and boundary layer separation occur. These flow features affect the fuel/air mixing and can considerably affect flame propagation and the combustion quality. To fully exploit the benefits of CNG DISI engines, these phenomena should be investigated. In this thesis, the gas dynamics, flow field, fuel/air mixing and flame propagation resulting from natural gas injection are studied using Large-Eddy Simulation (LES). First, LESs of methane DI as a CNG surrogate into a Constant Volume Chamber (CVC), considering the full internal geometry of a prototype injector, are performed to examine the effect of the injector internal geometry on the external flow characteristics of methane DI at conditions relevant to the production engines. The LES results are validated against high-speed, schlieren imaging experiments and empirical correlations. A new post-processing method is introduced, which permits the three-dimensional LES field to be projected into a two-dimensional density gradient field that can be compared to a schlieren image. Also, it is shown that a short version of the injector can represent the full injector if the pressure loss in the full injector is taken into account for imposing the inlet boundary conditions of the short injector. Second, LESs of methane DI into a modern CNG DISI production engine are presented to study the fuel/air mixing characteristics at various injection timings of the late, moderate and advanced. It is found that the Coanda effect exists in the cylinder and substantially impacts the gaseous jet development. The fuel/air mixing process is found to have two stages: a) during injection, mixing occurred strongly in the far-mixing zone where the jet was subsonic; and b) after injection, tumble flow motion was the key driving phenomenon for mixing. Different injection timings also result in varying levels of mixture homogeneity within the cylinder. The mixing duration decreases from the most advanced to the late injection timing. It is then expected that advanced injection, with a longer mixing time, leads to a better mixture homogeneity. However, it is shown that this is not the case because of the jet impingement on the piston crown in this case and therefore increased level of inhomogeneity.
  • Item
    Thumbnail Image
    Structure, energetic, and electronic properties of Layered perovskites and influence of a key structure feature: A-site ion movements
    Lou, Yaoding ( 2021)
    Perovskites, known as ABC3 compounds with the same type of crystal structure as calcium titanium oxide, have been studied for decades. Benefiting from the various compositions and structures, ferroelectricity, ferromagnetism, superconductivity, photoelectricity and other properties have been discovered and can be tuned in perovskites, which enable perovskites to be applied into various fields such as solar cells, catalysis, lasers, LEDs and so on. Meanwhile, with the developing requirements of modern applications, small scale devices with a density of properties require the multi-functional materials on the micrometre or nanometre scale. These requirements lead the way to two-dimensional (2D) materials. The fascinating novel properties of graphene exfoliated from graphite have encouraged the intense exploration of other 2D materials. 2D perovskites could be another promising type of 2D material. However, the first freestanding perovskites with several unit cell thicknesses were just reported recently. It is a new area that will attract lots of interest in the coming years. At present, it is highly desirable to get an overview of this new type of 2D material, including material chemistry, crystal structures, and the relevant physical properties (such as ferroelectricity). My thesis is devoted to achieving this aim. We start from the Ruddlesden-Popper (RP) perovskites, which is one kind of layered perovskite. It has been suggested that monolayer perovskites can be exfoliated from RP perovskites. It is a good material system that serves as the basis for a further understanding of 2D perovskites. Considering the systematic knowledge of RP perovskites system is limited by the shortage of attention, the data mining method was applied on the open-source database to sort out the datasets of the compositions, band gaps and structures of RP perovskites. The properties difference between ABC3 counterparts and RP perovskites are studied to understand how the thin layered crystal structures will influence material properties. A unique structure feature: A-site ion movement in RP perovskite, was identified, which potentially influence another structure feature: octahedral rotations. The A-site ion movements could decrease the tendency of octahedral rotations and even inhibit them. We perform the density functional theory (DFT) calculations and calculate the Coulomb energies to learn the origin of octahedral rotations, which suggests that A-C Coulomb energy reduction might be a major driving force for octahedral rotation in ABC3. Further study of Sr4Ti3O10, Ca4Ti3O10, Sr4Hf3O10 and Sr4Zr3O10 RP perovskite systems provide further insight into octahedral rotations. In this research, we separate the process of losing or decreasing octahedral rotation in these systems. Firstly, we find that A-site cation inward movements (ACIM) decrease the Coulomb energy, which weaken the driving force of octahedral rotation. Then the losing or decreasing of octahedral rotation could shift the electronic energy levels to lower values in electronic band diagrams, which provides the resistance force for octahedral rotations. As a result, RP perovskites system could lose or decrease the octahedral rotation on the surface. We also find the A-site cation inward movements could induce spontaneous polarization in RP perovskites. With the replacement of A-site cations by other A’-site cations on the surface layer, we can break the inversion symmetry and produce the spontaneous polarization in the whole system. By building the A’Sr3Ti3O10 and A’Ba3Ti3O10 (A’ as other alkaline earth metals or Pb) systems, we find that the smaller A’-site cations exhibit more significant inward movements and thus induce a larger spontaneous polarization. Further exploration of A’SrnTinO3n+1 and A’BanTinO3n+1 (n = 1, 2, 3) systems shows that changing the layer thickness n could effectively control the spontaneous polarization. The BeBaTiO4 with Ba cation inward movements has the highest polarization of 0.89 C/m2, which is even higher than the classic PbTiO3.
  • Item
    Thumbnail Image
    Resolvent analysis pressure modelling for airfoil trailing-edge noise predictions
    Wagner, Georges Alexandre ( 2021)
    The design of quieter airfoils has been stymied by the lack of fast-turnaround, accurate noise prediction tools applicable to new designs. Herein, the suitability of incompressible resolvent analysis for predicting the acoustic source field is investigated. This is done by comparing resolvent predictions to high-fidelity counterparts. Combining resolvent method surface pressure predictions with Amiet's acoustic analogy is shown to be a very promising trailing-edge noise prediction framework. The accuracy of this noise prediction hinges on an accurate prediction of the surface pressure difference, in the region close to the airfoil trailing edge. In the first part of this thesis it shown, for a flat plate and a NACA0012 airfoil, that incompressible resolvent analysis accurately captures the surface pressure footprint of hydrodynamic instabilities that give rise to noise. Resolvent analysis can be relied upon to identify the frequencies at which dominant linear instability mechanisms occur that lead to tonal contributions to TE noise. At these frequencies, resolvent-based Amiet noise predictions replicate DNS-based Amiet noise predictions and are two orders of magnitude faster. Incompressible resolvent analysis is also applied to a controlled diffusion (CD) airfoil, at a high Reynolds number and large incidence. The physics-driven nature of resolvent analysis is demonstrated, as it captures subtle differences in the acoustic source field for changes in the flow conditions. However, the method struggles to capture structures that arise from nonlinear interaction and the breakdown to turbulence. The second part of this thesis consists of an investigation into which components of pressure are accounted for by resolvent analysis of the incompressible Navier-Stokes equations. Firstly, we analyse whether the resolvent method predicts solely one of the incident or scattered pressure fields or whether it accounts for both and their interaction, i.e. to total pressure field. The resolvent method is shown to capture the total pressure field, motivating its use for generating input to Amiet's acoustic analogy. Secondly, we analyse how accurately the incompressible resolvent formulation predicts the fast component of pressure and whether it can be used to retrieve the slow component of pressure. At frequencies where the dominant flow features arise from linear instability mechanisms, the fast pressure far exceeds the slow pressure in magnitude. Resolvent analysis provides accurate approximations of this fast pressure, and by extension of the total pressure. At frequencies where the dominant flow features arise due to the nonlinear interaction of modes from different frequencies, the slow component of pressure is non-negligible. To model the latter, the resolvent forcing needs to contain a non-solenoidal component. This is achieved by approximating the forcing using the triadic interaction of resolvent modes from frequencies where linear mechanisms dominate. Using the true forcing, evaluated from DNS data, we are able to capture the airfoil surface pressure with great accuracy, even at frequencies where the flow physics are dominated by nonlinear interaction, the resolvent operator is not low-rank, and the slow pressure constitutes a large contribution to the total pressure. This proves that even when it is not low-rank, the resolvent operator remains a transfer function capable of predicting the flow-physics in an accurate and low-cost manner.
  • Item
    Thumbnail Image
    Characterisation, Detection and Forecasting of Intra-hourly Wind Power Ramps
    Pichault, Mathieu Marc Bernard ( 2021)
    One of the challenges preventing greater integration of wind energy into the power grids is the occurrence of large changes in wind power generation over a short amount of time, also called "ramp events". With ever-expanding wind energy penetration, wind power ramps pose a severe risk to electric system security and their mismanagement can have dramatic consequences such as power blackouts. Accurate and timely short-term forecasts are crucial in predicting these ramps and facilitating power grid balancing, storage management, and load dispatch planning. The first part of the thesis is concerned with characterising wind power ramps and their underlying engineering and meteorological processes, with a view to reduce uncertainties in wind power forecasts. It is shown, amongst other things, that 46 % of the ramps at the study site are associated with frontal activity and that wind power fluctuations at the study site tend to plateau before and after the ramps. In the second part of the thesis, two methodologies to predict wind power based on upstream velocity field measurements are introduced and shown to outperform industry benchmarks. The forecasting skill of the proposed methods is particularly evident over the course of wind power ramps, reporting up to 19 % improvement over the persistence benchmark. The opportunities and challenges inherent in using wind LiDARs for ramp forecasting are also discussed. Finally, further analyses of the wind flow fields demonstrate the impact of wind gusts (coherent blocks of strong wind speeds) on ramping behaviour. The results show that wind gusts are associated with increased power variability, both at the wind farm scale and the wind turbine scale. It is also shown that gusts with length scales less than one kilometre tend to travel at the rate of the background flow, whereas larger gusts propagate on average faster. Overall, this research is part of a broader initiative to mitigate greenhouse gas emissions by making wind energy a more controllable resource.
  • Item
    Thumbnail Image
    Data-driven Reynolds-averaged turbulence closures for buoyancy-affected flow
    Xu, Xiaowei ( 2021)
    Turbulent flow subjected to buoyancy force is ubiquitous in daily life, e.g. in building ventilation, nuclear reactor containment, and geophysical flows. To improve the prediction accuracy of existing turbulence models, this thesis presents the results of the application of an in-house symbolic regression tool, i.e. gene expression programming (GEP), on buoyancy-extended Reynolds-averaged closure models for buoyant flows in a differentially heated vertical planar channel. In the first part of this study, attention is paid to understanding the turbulent Prandtl number's behaviour and improve the predictability of the linear eddy diffusivity models. By comparing the location of mean velocity maxima, there is an infinity anomaly for the eddy viscosity and the turbulent Prandtl number, as both terms are divided by the mean velocity gradient according to the standard definition, in vertical buoyant flow. To predict the quantities of interest, e.g. the Nusselt number, GEP is used with various cost functions, e.g. the mean velocity gradient, with the aid of the latest direct numerical simulation (DNS) dataset for vertical natural and mixed convection. It is found that the new machine-learnt algebraic models, as the reciprocal of $Pr_t$, successfully handle the infinity issue for both vertical natural and mixed convection. Moreover, the proposed models with embedded coordinate frame invariance can be conveniently implemented in the Reynolds-averaged scalar equation and are proven to be robust and accurate in the current parameter space, where the Rayleigh number spans from $10^5$ to $10^9 $ for vertical natural convection and the bulk Richardson number $Ri_b $ is in the range of $ 0$ and $ 0.1$ for vertical mixed convection. However, there are notable errors between the prediction and DNS data when incorporating the algebraic model of turbulent Prandtl number into full Reynolds--averaged Navier--Stokes (RANS) equations. As a result, the turbulence closure is upgraded with buoyancy-extended terms. The second part of this study re-examines the buoyancy-accounting algebraic scalar-flux model proposed by Kenjeres et al., Int. J. Heat Fluid Flow, Vol. 26, pp. 569-586 (2005). Based on a term-by-term analysis on the model with the aid of high-fidelity datasets, it is demonstrated that there are significant discrepancies in the predicted turbulent heat fluxes once the model is combined with the existing algebraic Reynolds stress models. Consequently, it is suggested that the quadratic terms in buoyancy-extended explicit algebraic Reynolds stress models should be included, and such non-linear Reynolds-stress and heat-flux closure models are then developed via GEP. The evaluation of these GEP-based models shows significant improvements in the prediction of mean quantities and second moments in an a-priori stage and in an a-posteriori stage, with the latter being realised by embedding the new models into the elliptic relaxation v^2-f equations, across different Rayleigh number cases. In comparison to passive scalar flow, the complexity of turbulence modelling for natural convection problems is increased as the velocity and scalar fields are strongly coupled by the buoyancy force. The above data-driven turbulence modelling approaches have treated the unclosed terms of the velocity and thermal fields separately, which has lead to inaccurate predictions when handling natural convection problems. Hence, the appropriate Reynolds-averaged closure models for natural convection ought to capture this interaction within the second-moment terms. In the last part of this study, we therefore develop fully coupled buoyancy-extended models by using a novel multi-objective and multi-expression machine-learning framework that is based on CFD-driven training (Zhao et al., J. Comput. Phys., 411, 109413, (2020)).The model candidates obtained from a Gene-Expression Programming approach, and thus available in symbolic form, are evaluated by running RANS solvers for different Rayleigh number cases during the model training process. This novel framework is applied to vertical natural convection, with the emphasis on the importance of coupling the explicit closure model formulations, the choice of cost functions, and the appropriate input flow features (i.e. a generalised flux Richardson number) for developing accurate models. It is shown that the resulting machine-learnt models improve the predictions of quantities of interests, e.g. mean velocity and temperature profiles, for vertical natural convection with Rayleigh numbers in the range of $10^5$ to $10^9$.
  • Item
    No Preview Available
    Pulsatile effects on turbulence dynamics in pipe flow
    Cheng, Zijin ( 2021)
    An increasing concern in medical industry on human diseases in artery flows (e, g,. Cardiovascular Disease (CVD)) has risen in the relevant research fields. The turbulent activities and pulsating forcing conditions of the blood flow in aorta increase dramatically the complexity of flow physics. In addition to that, pulsatile turbulent flows are well-encountered in a wide range of engineering applications and physical systems including environmental flows over ocean and reciprocating flows in internal combustion engine. Therefore, increased understanding of pulsatile forcing turbulent flow is of immense technological importance and is beneficial towards a future fluid mechanics field. The objective of this study is to use direct numerical simulations (DNSs) method to conduct a series of numerical experiments on pulsatile turbulent pipe flows and to complement the fundamental understanding of pulsatile forcing effects on turbulence dynamics. To this end, the investigation covers a range of friction Reynolds numbers and forcing conditions and is conducted via the following three topics: i) the turbulence dynamics in single-mode pulsatile forcing pipe flow; ii) the Reynolds-number effect on single-mode forcing conditions; iii) the turbulence dynamics at dual-mode pulsatile forcing condition. The study of single-mode pulsatile pipe flow ranged the Reynolds number at both low (180) and moderate-high (360, 540) friction Reynolds numbers and the forcing frequencies over high (type IV) and very-high (type V) forcing regimes at a fixed pulsatile amplitude of 0.64. In this topic, a new physics-informed method of forcing classification was achieved by directly comparing the applied frequencies with the instantaneous Reynolds shear stress co-spectra, – a rarely reported quantity in most experimental and DNS studies – in the frequency domain and thus a new forcing type – the upper-limit of the very-high frequency regime – was determined as ultra-high frequency (type VI). At the ultra-high forcing type, turbulent activities were fully decoupled from the pulsating flow field. The roadmap of our forcing classification method exhibited a good robustness where the turbulence dynamics presented consistent responses to the high, very-high and ultra-high forcing conditions across these three computational Reynolds numbers. In dual-mode forcing pulsatile flows, the forcing classification acquired from RSS frequency co-spectra gave a good outline. The simulations were conducted by combining either two of the three aforementioned forcing types (types IV, V and VI), and this work complemented the possible ambiguity of the dual-mode forcing effects in past classifications. To be specific, at the forcing combination of types IV and V, observable interactions between the two forcing modes were found and the phase-averaged turbulence statistics at each forcing mode showed difference with the corresponding single-mode values. At the forcing combination of types IV and VI, the first-mode phase-averaged turbulence statistics were no longer affected by the second forcing and showed good agreement with the corresponding single-mode values. In the meantime, the second-mode phase-averaged statistics became phase-independent but showed slight difference with the corresponding single-mode values. At the forcing combination of types V and VI, the two forcing modes were fully decoupled with each other and both modes showed good agreement with the corresponding single-mode results. In addition, my work extended the previous studies by complementing a detailed analysis of a series of time- and phase-averaged single- (including quadrant profiles, probability density functions, weighted joint probability density functions and skewness profiles) and two-point (including cross-correlations, one-dimensional and two-dimensional wave-number spectra) Reynolds shear stress and other statistics in both physical and Fourier domains.