This thesis presents theoretical and practical issues of local optimal control, which is one of the advanced control methods. It can be counted as an optimal modelbased multivariable control technique. The main contributions of this work can be summarized as follows. • A comparative robustness study of local optimal controller with other conventional controllers is performed for gas turbine engine as a multivariable system. • As the original local optimal control is incapable to deal with non-minimum phase systems, a modified local optimal control is proposed to deal with non-minimum phase systems as well as minimum phase systems. • The local optimal controller performance is investigated for reduced order models. Because of its effectiveness, genetic algorithm is used with certain predefined controller structures as an alternative method to estimate the controller parameters without obtaining the model parameters. • A new tuning technique of digital PID controller is introduced for both multivariable and single-input single-output systems based on the relations deduced with the local optimal controller. As such, the PID controller is turned into model-based controller. • As tlie PIO and the local optimal controller are model based multivariable controllers, their parameters can be tuned online based on online identification techniques. The recursive lease squares algorithm is used as an online . closed loop identification technique to achieve such online tuning of those controller parameters. • Local optimal controller is generalized to deal with non-linear systems as a non-linear controller. • Most of the above techniques are tested on a laboratory-based test rig.

Flow in a turbocharger turbine is highly unsteady in nature as it responds to the exhaust manifold of an internal combustion (IC) engine. Despite this it is conventional to use quasi-steady turbine models in one-dimensional turbocharged engine simulations, even though they cannot reproduce the known hysteresis of turbine mass flow and performance characteristics recorded under pulsating flow conditions. Using filling-and-emptying models improves the situation by permitting mass accumulation in the turbine volute. Depending on the unsteadiness level, this approach may still be insufficient to capture true turbine operation since neither method can resolve unsteady effects due to pressure wave action in the flow. It is unclear when transition occurs between filling-and- emptying and wave action modes. To this end, a proprietary computational gas dynamics code in C++ is presented to simulate the unsteady, compressible flow inherent to IC engine exhaust manifolds. The Euler equations for one-dimensional inviscid flow are discretized to provide second-order, conservative, shock-capturing finite difference schemes able to resolve wave propagation in ducts with area variation, wall friction and heat transfer. A wave action turbine volute model is constructed using bespoke boundary conditions. Validation against experimental data shows satisfactory agreement for pulse frequencies up to 40 Hz, and improved instantaneous swallowing capacity prediction at all tested frequencies compared to quasi-steady calculations. Fourier series characterization of on-engine pulse waveforms reveals multiple harmonic components, causing significant regions of divergence between filling-and-emptying and wave action predicted hystereses. Comparison of concurrent wave action and filling-and-emptying simulations applying simpler sinusoidal waveforms allows development of the unsteadiness measures FSt and FSt(p). An approximate guideline to ensure a filling-and-emptying mode stipulates FSt [Symbol appears here. To view, please open pdf attachment]0.15 and FSt(p) [Symbol appears here. To view, please open pdf attachment]0.02. Evaluation of FSt and FSt(p) for an example on-engine case indicates certain wave action already by 1600 rev/min, borne out by subsequent inspection of the swallowing capacity traces.

A pulsating combustor producing longitudinal acoustic oscillations was constructed and a mathematical model of the system developed. The combustor was a closed - open tube, combustion taking place at the closed end into which were fed air and propane. The two lowest modes of longitudinal, acoustic vibration were obtained. The fundamental occurring at low fuel flowrate up to a maximum flowrate corresponding to an energy • input of 12Kw, at which either the fundamental or first harmonic occurred and above which only the first harmonic was obtained up to the maximum flowrate of the system corresponding to an energy input of 20 Kw. The analysis of the system used the conservation equations of mass, momentum and energy from these suitably formed equations could be derived which were solved by the method of characteristics. The combustion model was governed by a simple overall-reaction rate equation. Plug flow was assumed with perfect radial mixing and no axial mixing, conduction or diffusion. The convective heat transfer coefficients were evaluated by means of the quasi-steady-state theory. The mathematical model predicted the gas temperature gradient and the distribution of pressure and velocity standing waves. Owing to the use of a much simplified combustion model, it was not possible to predict the acoustic energy required to determine the amplitude and frequency of oscillation. The amplitude was found to be highly dependent on the fuel injection system, Air/Fuel ratio and mode of oscillation. The practical results confirmed the higher rates of heat transfer associated with pulsating flow.

No description available.

No description available.

No description available.

Co-firing coal with biomass has gained much interest in recent times by power generators keen on exploiting the environmental and economic benefits. Various trials have been undertaken on small substitution levels of typically below 10% of the total thermal input. Higher substitution levels would expose potential problems in terms of slagging and fouling on heat transfer surfaces. The research study investigated the use of a novel small scale combustor to simulate the conditions of real industrial furnaces. The design and manufacture of the novel combustor is explained with detailed discussion on the developments to suit the combustor for co-firing trials. Successful simulations of a 500kW semi-industrial and a 235 MWe full scale furnaces were achieved. Co-firing trials were performed with three types of waste biomass dried sewage sludge, sawdust and refuse derived fuel. Numerous valuable deposition data was generated during the research study. The data included deposition observations, fouling deposition rates, fuel and fly ash analyses, slag deposition analyses and online flue gas analyses. These would form part of an advanced slagging and fouling predictor. References to traditional empirical indices for slagging and fouling are also included.

One of the main thermal-hydraulic challenges of LWR modelling is the prediction of boiling phenomena. This thesis describes numerical and analytical studies aimed at modelling the heat transfer and hydrodynamics of a single steam bubble during nucleate boiling of water, aiming both to improve our current understanding of the phenomena - of the evaporation process itself, and of nucleate boiling heat transfer - and to improve our ability to predict such phenomena, at both single bubble and component scales. Analytical and CFD studies of bubble formation are described. These require accurate representations of evaporation from the liquid microlayer at the bubble base. This vaporization has been investigated from a molecular point of view, with modelling base on kinetic theory, and an apparent inconsistency in measurements of microlayer evaporation during bubble formation has been resolved, and an improved understanding of the molecular mechanism of phase-change thereby gained. The importance of including this improved representation of the evaporation process in single-bubble CFD simulations has been demonstrated. Aiming to improve the closure relations employed for component-scale CFD simulation of boiling flows, interface-tracking modelling of bubble growth and release has been used. Single-bubble interface-tracking models have been developed in an attempt to quantify the transient conduction ('quenching') component of nucleate boiling heat transfer, associated with bubble lift-off. These mechanistic models allowed detailed quantification of the complex physics associated with bubble growth, and with quenching of the dry area at the bubble base that takes place at bubble departure. A large discrepancy was observed between estimates of the quench heat transfer from these interface-tracking simulations, and that incorporated in the more approximate modelling embodied in the closure relations widely used in component scale CFD.

Polymers undergo physical, chemical and structural changes when exposed to heat and/or fire. Thermoplastics melt, decompose and burn; thermosets decompose, char and/or burn, depending on the temperature changes due to external incident heat flux. Detailed in this thesis is a theoretical and numerical heat transfer study, which is undertaken to simulate and experimentally validated temperature variations during melting, decomposition, charring and ignition phases of polymers. For melting, thermoplastic polymers (polypropylene, polyester, polyamide 6, polymethyl methacrylate, polycarbonate and polystyrene) have been used, whereas for decomposition, charring and ignition glass fibre – reinforced epoxy composites have been chosen. For each case a one-dimensional finite difference method, using Matlab as the operator has been developed to determine the transient temperature distributions within the different types of polymers materials. The convective and radiative heat transfer boundary conditions, at the exposed and unexposed sides of polymer samples, have also been taken into account accordingly. While some experimental results to validate the different numerical models built are from other researchers’ work at Bolton, in addition to these, other sets of experiments were specifically developed for this work. The melting behaviour of thermoplastics has been modelled in two scenarios: (i) vertically oriented sample where melt dripping occurs and (ii) horizontally oriented sample within a contained holder in order that the mass will not escape from the containment region. In the the first scenario the sample was placed in a tube furnace, where the radiant heat is uniform on all sides of the sample. This is based on the experimental methodology developed at Bolton University in an earlier project which studied the melt dripping behaviour of polymers. The thermogravimetric and rheological analysis of molten drops had indicated that, depending upon the temperature of the furnace (external heat flux) and the structure of the polymer, in some cases it was pure melting whereas in others it was accompanied by a partial decomposition of the polymer. A one-dimensional finite difference method based on a moving boundary approach has been developed to model the temperatures of the molten drops polymers. The simulated results showed good agreement with the molten drops’ temperatures measured by experiments. In addition, using kinetic parameters, degrees of decomposition in drops obtained at different furnace temperatures were also simulated, which were validated with previous experimental results. For the second scenario, in which the sample is placed horizontally in a container, experiments were conducted using a cone calorimeter with the heat applied only on the top surface, while the other sides of the polymer sample are insulated, A further one-dimensional finite difference method based on a Stefan approach involving phase changing material, has been developed to determine the melting point temperature and to estimate the temperature profile within the polymer slab, to simulate pure melting and melting plus partial decomposition which may or may not catch fire depending upon the degree of decomposition. The predicted results matched well with the experimental results. Furthermore, the heat transfer model was modified to simulate the temperature profiles through the thickness of a glass fibre - reinforced composite exposed to different heat fluxes in a cone calorimeter. This involved incorporating a kinetic model for the decomposition process taking into consideration the varying thermophysical properties as a function of temperature. This is achieved by using the critical heat flux that is the minimum incident heat flux leading to ignition, in the equation defining the ignition temperature, The simulated temperature profiles matched well with the experimental results obtained from previous works at the University of Bolton, giving a much better agreement than previously published models describing this condition. Ignition phenomenon is well described by the model showing a jumping step when the composite polymer ignites and burns. The last part of the work was to simulate the heat transfer in Intumescent coated glass fibre reinforced epoxy composites exposed to heat in a cone calorimeter. On exposure to heat the intumescent coating expands to form a char, the thickness and the thermal conductivity of which, depends on the type of coating. It was not the purpose of this work to model expansion of the coating; rather the emphasis was to understand the thermal barrier efficiency of the expanded char. However, changes to the surface, expansion of the local thickness and char region when exposed to heat were incorporated into the model to gain better agreement with experiment values.

The present investigation of the air motion in the prechamber of an indirect injection diesel engine is in two main parts. In the first part a mathematical model was developed to describe the air motion in the prechamber. In the second part experimental measurements were made of mean air velocities in two different cylindrical prechambers. The prechamber tests were carried out with the engine under motoring conditions. The computations of the mathematical model were compared with the present experimental results and also with previously published data by other authors. A discussion of these comparisons was made.