No description available.

Conventional packed bed reactors of low tube to particle diameter ratio suffer from poor heat transfer near the tube wall, and also from increased axial dispersion compared with wide beds. In this dissertation the potential of a new reactor design aimed at overcoming these deficiencies is investigated. The radially-stratified bed provides more particles in the wall region to support improved heat transfer and also to flatten the voidage profile in an attempt to reduce the axial dispersion. An experimental study on the effect of stratification on voidage profiles using an image analysis technique showed that both the voidage and velocity profiles could be flattened. A study into the effect of different packing arrangements on axial dispersion has shown that the best arrangement for flattening the voidage profile can lead to over-compensation, resulting in channelling through the core and an increase in dispersion. However, a packing arrangement consisting of a binary mixture of large and small particles near the wall with the core of the bed packed solely with the larger, was shown to exhibit dispersion characteristics no worse than monosized packing. This same packing arrangement was also found to support improved heat transfer. At Re<SUB>p ></SUB> 1100 the heat transfer coefficient appropriate to the one-dimensional plug flow model was shown to increase by ca. 15%. A novel analysis of the one-dimensional reactor model has shown that use of a stratified bed of the same voidage, heat transfer coefficient, tube diameter, and feed flowrate as a monosized bed, results in a pressure drop which is 57% of that across a conventional bed of the same bed length. A further advantage of the stratified bed is that, on average, the catalyst particles are nearer the wall than in the conventional monosized bed. The magnitude of this and the other advantages was assessed by simulation of a reaction performed at high Reynolds numbers, the partial oxidation of ethylene. For this purpose, a new, plausible but simple two-dimensional model of the packed bed reactor was devised. A stratified bed was predicted to then have a pressure drop only some 34% of that of a conventional bed for the same overall conversion. This further reduction of some 23% in pressure drop stems largely from the reduction in packed length necessary to achieve a stipulated conversion. Alternatively, for the same selectivity, a stratified bed could reduce the pressure drop to about half that of a conventional bed of the same voidage. The potential of this novel design has thus been demonstrated: the next stage is optimisation.

No description available.

In this thesis a theoretical model that describes particles impacting with a circular plate, the resulting flexural vibrations of the plate, and its detection by an ultrasonic transducer, is described. Using a variety of approximations, an analytic expression for the plate vibrations is derived. This is then expressed as a Fourier series and an analytic expression for the frequency of the first main peak of the spectrum is obtained.

The initial objective of the work was to update the reactor pilot-scale rig command and control system. This enabled a variety of self-tuning control strategies [self-tuning PID (STPID), generalised minimum variance (GMV) and generalised predictive control (GPC)] to be applied in an attempt to achieve the desired control objectives. The different self-tuning control procedures were designed, then tested by simulation, and finally implemented on the experimental rig. It has been found that GPC gives superior control (in comparison to STPID and GMV) when the system dead-time varies and or has been poorly estimated. The GPC strategy is demonstrated further to cope well with system non-linearities. The preheater outlet temperature on the pilot plant is shown to be successfully controlled using all three self-tuning strategies. Generalised predictive control is found to give excellent results when employed to control the inlet temperature and the position and temperature of the hot-spot of the fixed-bed reactor where poorly known and variable dead-times are present. The hot-spot temperature on the rig is shown to be well controlled also by employing a new version of the generalised predictive controller, viz. extremum GPC. Finally both STPID and GPC strategies are found to be successful when based on a design which employs a neural network for system identification and prediction instead of the normal least squares procedure.

No description available.

The development of quantitative predictive models is of great significance for the successful application of membrane separation processes in the process industries. Successful models exist for the ultrafiltration of macromolecules and particles with sizes 2 to 5nm and 500nm to 10μm, respectively. None of the approaches could predict colloidal ultrafiltration rates in the intermediate size range of 5 to 500nm. This is due to interparticle interactions, especially electrostatic interactions which play an important role and which have been widely neglected or even overlooked by many membrane scientists. As a result of such interactions, changes in the ionic environment and particle zeta potential give <i>order of magnitude</i> changes in the rate of ultrafiltration. However, there has been no rigorous mathematical model for the ultrafiltration of colloids in this range that takes into account such interparticle interactions. The aim of the present dissertation was to develop and test rigorous mathematical models for membrane ultrafiltration of charged colloidal particles, which include quantitative calculations of particle-particle interactions within filter cakes, which are responsible for controlling permeation rates. This has been realized by two approaches: 1) an extended pairwise summation of interaction energies due to electrostatic or double layer forces and London-van der Waals forces through the DLVO theory and 2) by a multiparticle interaction approach based on the use of a Wigner-Seitz cell model to calculate the grand canonical electrostatic potential energy of a concentrated particle system. Dynamic ultrafiltration models for filter cakes of charged particles have been developed ('<i>From Physics To Filtration</i>') based on disjoining pressure calculations between hexagonal close packed particle layers which take the former interaction calculations into proper account. The models require no adjustable parameters. A comparison with experimental ultrafiltration results revealed that the dynamic model based on pairwise summation of interaction energies was in good agreement with the experimental filtration flux over a limited range of conditions, whereas the cell model predictions were in excellent quantitative agreement over a wide range of ionic and zeta potential conditions.

The control of a tubular fixed-bed catalytic reactor (TFBCR) system has been studied both experimentally and by simulation. A description is given of the modifications made to a tubular reactor pilot-plant (designed originally for steady-state operation) in order to facilitate full computer control under both steady-state and dynamic experimental conditions. Details are presented of self-tuning control algorithms that were employed to adjust the principal process variables. The reactor involved oxidises sulphur dioxide over a vanadium pentoxide catalyst at temperatures in excess of 673K. For dynamic work, good control of the process variables (viz. air and sulphur dioxide mass flowrates and the gas temperature at the reactor inlet) is required. As the oxidation reaction is exothermic, an additional control problem arises whereby the maximum temperature within the reactor must be limited such that catalyst deactivation or 'reactor runaway' do not take place. Changes made to the pilot-plant included improvements in the control microcomputer, the interfacing with the plant and the installation of mass flow controllers to adjust reliably both air and sulphur dioxide flowrates. A fast acting heater was added to allow rapid changes to be made in the temperature of the reactor inlet gases. Desired experimental conditions could be achieved using the process computer, which also had the capability to carry out full data logging, reactor start-up, shut-down and emergency shut-down. The control study involved two single-input single-output loops i.e. that controlling reactor gas inlet temperature and that controlling the maximum temperature within the reactor and the multi-variable control of the 'hot-spot' temperature and position using both inlet gas temperature and flowrate. Control difficulties were encountered since all loops exhibited considerable time delays coupled with large time-constants; additionally there appeared to be no linear relationships between controlling and controlled variables over the range of reactor operating conditions. The performances of a number of different feedback and feedforward/feedback self-tuning algorithms were assessed with respect to the process variables and comparisons were made with the standard proportional-integral-derivative (PID) controller and its variants. It has been demonstrated that the applications of self-tuning generalised predictive control (GPC) algorithms can provide robust control, with better performance than both self-tuning generalised minimum variance and (both self-tuning and fixed parameter) PID techniques, over a wide range of process operating conditions.

No description available.

No description available.