No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

This dissertation presents a novel method for generating the computational geometry of three spool aero engine for the Virtual Engine (VE) design environment, which is capable of carrying out full integration of preliminary and detailed design and analysis activities. The multi-fidelity and multi-disciplinary analysis of complex gas-turbine assemblies is the prime requirement of industrial design. The present challenges for gas-turbine industry, namely excessive design times, lack of decisive information till the late design stages, mismatching of databases and high costs associated with testing etc. necessitate creation of an integrated design framework such as VE. This thesis presents the work on selected parts of VE design environment. The present work initially focuses on the development of an integrated geometry modelling system for VE. The architecture of geometry module is derived from the requirements of VE design environment and Computer Aided Design (CAD) geometry systems. A dedicated geometry modeller is developed to represent gas turbine components. The building blocks of this modeller, known as features, are constructed as object-oriented data structures. A taxonomy of turbomachinery design features is defined to generate 2D axisymmetric geometry model of three-spool aero-engine. Such system supports intra-analysis information augmentation, data updating and automated data transformation for any kind of analysis. In the next phase, the Computer Aided Engineering (CAE) capabilities of VE are demonstrated by carrying out flow network analysis of Secondary Air System (SAS). Particular attention is devoted to the automated extraction of SAS network from the geometry. The generation of SAS network involves interrogation of component models, automatic identification of flow links and pre-processing of network elements. A library of physics-based linearized pressure loss models is prepared, validated and incorporated in VE. A new loss model for straight-through labyrinth seal is developed from numerical experiments and it is validated against those in the literature. A linearized flow solver for the whole engine SAS network model is also developed. Finally, low fidelity steady-state flow analysis is demonstrated on a limited domain of SAS network model. This work meets its objectives by showing a way of constructing important modules of VE design framework. It bridges the gap in current features technology by creating the design features that represent turbomachinery components. The work demonstrates quick generation of low-fidelity geometry model for whole aero-engine assembly, thus endorsing its utility for the analysis in early stages of design. A methodology for performing automated CAE analysis of secondary air system (SAS) of aero engine has also been developed. The development of new loss model for labyrinth seal and the demonstration of low-fidelity steady-state SAS analysis confirm the successful implementation of selected part of VE project.

In this thesis, a discretised Population Balance Equation (PBE) model is coupled with a detailed in-house Computational Fluid Dynamics code to study soot formation in axisymmetric diffusion flames with comprehensive gas-phase chemistries for C2H4 and CH4 fuels. The main aim of this study is to predict the complete Particle Size Distribution (PSD) of soot particles in turbulent non-premixed flames via a transported Probability Density Function approach. The PSD is obtained from the solution of the PBE without any prior assumption on its shape, using volume or diameter to describe the size of soot particles. However, due to a great number of uncertainties that appear from the turbulence interactions with chemistry, radiation and particle formation, the main objective is divided into smaller tasks where these complexities are avoided. Initially, the performance of the PBE is assessed under several numerical methods on an initial distribution-convection test, 0D reactors and 1D flamelet framework. The PBE has been originally discretised via a collocation type finite element method, and in the present work Finite Volume (FV) methods are used. The PBE with Total Variation Diminishing (TVD) scheme demonstrates better performance. The FV-TVD PBE with suitable soot kinetics is employed in 2D laminar flames where overall good agreement is achieved for the velocity, temperature, mole fraction of C2H2 and OH species and the mean properties of PSD (i.e. total number density and soot volume fraction). However, the temperature and soot volume fraction profiles on the turbulent flames do not exhibit similar accuracy as the laminar flames and there is still room for improvement. The evolution of the PSD is computed for both flames in the entire flame region exhibiting weak bimodal distribution in some points. The performance of complex coupled phenomena in PBE modelling via soot kinetics, detailed chemistry, radiation and turbulence interactions is explored.

During long-term exposure in creep or creep-fatigue environment, the creep strength of 9-12%Cr steel drops significantly due to the damage developed from metallurgical and microstructural changes occurring in complex subgrain structures. As creep time increases, failure strain of the material is reduced and the reduction is further enhanced under multiaxial condition, where a localised damage zone formed thus leading to crack initiation and growth under increased constraint. The present work focuses on high-temperature creep and fatigue behaviour, and life prediction of Grade 92 steel. Long-term data from literature is utilised to establish creep constitutive properties and to develop the predictive model. The expected scatter in the data is accounted for and it is shown that the important upper and lower bound trend under short and long-term tests can be identified irrespective of the variability of the data. Modelling approach is first applied to the stress-based Continuum Damage Mechanics (CDM) model coupled with a concept of representative stress to predict notched bar failures. Exponential-type predictive model which consists of two material constants, compared to standard CDM model with six is proposed as an alternative to the prediction. Furthermore, a semi-empirical constraint-based approach to predict creep rupture and crack growth under simple and complex stress state condition is presented. The model relates the Monkman-Grant failure strain to the constraint which arises from the geometry and time-dependent creep damage process in microscopic level. The model is consistent with the NSW (Nikbin-Smith-Webster) crack growth model which predicts the crack initiation and growth under the range of plane stress and strain. Finally, finite element (FE) analysis is performed on plain and notched bar to examine the influence of stress state on creep rupture and damage behaviour, and to validate the proposed predictive approach. The model is sufficiently simple yet reliable in predicting long-term failures.

The major drive to advance the understanding of flames comes from industry as many technical devices propelled by flames are in constant need of design revisions and efficiency improvements. Lean-premixed combustion technology was one of such areas, which saw major advancements. Unfortunately, combustion devices operating in premixed mode often exhibit undesired flame oscillations, which may lead to damage or even failure of the combustor. The present work is dedicated to better understanding and predicting flame oscillation. A modelling framework based on high-order accurate Discontinuous Galerkin methods is presented to investigate the interaction of steadily propagating flames with pressure waves. The analysis of the results show that the time scale associated with the pressure wave incident upon the flame must be of the same order as the time scales in the primary reaction zone to lead to partial wave reflection and wave transmission at the flame. Hence providing a mechanism for initiating flame oscillation. The predictive capabilities of the Large Eddy Simulation technique in conjunction with the Probability Density Function model of the unresolved turbulence-chemistry interaction are demonstrated in the simulation of flame oscillation in industrially relevant combustors. Two test cases with increasing complexity are considered: a forced oscillating flame in a bluff-body stabilised combustor, and a self-excited flame oscillation in a swirl combustor with complex geometry. Good agreement between the measurements and the simulations is obtained. The flame dynamics are well captured by both simulations. The results obtained in the study of the bluff-body flame are used to identify suitable chemical markers which correlate well with the total heat release rate. The product of molecular oxygen and the ketenyl radical is found to correlate better with the total heat release rate than the commonly used formaldehyde-based markers in the investigated premixed fuel-lean ethylene-air flame. In the case of the swirl burner the predictions lead to the suggestion of a self-excited flame oscillation in the lateral direction as the result of an interaction of the flame with a vortex ring. The difference between the predicted and experimentally determined frequency of oscillation is 11 %.