Various problems regarding the behaviour of fixed bed catalytic reactors involving highly exothermic reactions have been studied in relation to optimal design and control. Steady and unsteady state mathematical models of various degrees of complexity have been used from those considering axial and radial diffusion to the simple one dimensional representation neglecting both mechanisms. Since these models would be used repeatedly, they must be relatively simply solved by a computer in a reasonable time and without loss of detail necessary to take full advantage of control or optimization processes. Orthogonal collocation has proved a very efficient method of solution for solution of the radial diffusion and axial diffusion models. It has been shown that in the former case, an optimal distribution of the collocation points in the radial direction requires the minimum number of points. Double collocation, under certain conditions, is an efficient integration procedure both for steady and unsteady state models. In the case of the axial diffusion model, some orthogonal polynomials converge faster than others depending on the profiles to be approximated. It has been recognised that further reduction in computing time is usually coupled with a reduction in model dimensionality. A model reduction technique has been used to lump the radial profiles in the unsteady state radial diffusion model. This lumped model has the ability to regenerate the radial profiles from simple algebraic expressions with reasonable accuracy compared with the distributed parameter system. Studies on the transient behaviour of the reactor have indicated that the major dynamic factor is the solid heat capacitance and that the inlet temperature and concentration may be manipulated to effectively control the reactor in a multi-variable mode. Consideration has also been given to the response of the reactor to sinusoidal and damped sinusoidal perturbations at the inlet. It has been found that for certain frequencies severe hot spots may be formed over a part of the radial profiles before a safe quasi-stationary state is reached. A detailed examination of this behaviour has shown that the differences in the speeds of propagation of the concentration and temperature waves along the reactor were significant factors in determining the resulting behaviour. A steady state axial diffusion model in which the radial variation in temperature is approximated by a parabolic radial temperature profile has been considered. The limitations of this approximation have been identified and treated by the model reduction technique. Thus the model developed gives adequate representation of axial and radial dispersion processes. Axial dispersion becomes important if the axial temperature and concentration gradients increase beyond a certain value. This value may be calculated from the axial profiles of the one dimensional model which neglects axial diffusion. Consideration has been given to a dynamic model based on the above and the collocation and the reduction technique used to solve the model. The solution time is reduced to reasonable levels making, it suitable for detailed studies. Including the axial dispersion in the dynamic model did not alter the qualitative behaviour of the reactor. The exceptional cases are those related to parametric sensitivity or temperature runaway studies. Instability arising from parametric sensitivity or multiple states in either the radial or axial diffusion models has been considered. The criteria developed indicate that if instability is to occur in the reactor, it is likely to originate from the solid phase regardless of the mechanisms considered in the fluid. Thus, conditions of catalyst particle stability are essential in establishing local stability of the reactor.

The analysis of operating procedures in the early stages of design can lead to safer and higher performance plants. Qualitative reasoning techniques hold considerable promise in supporting generations of operating procedures, since they are able to describe possible trajectories of a system based on non-quantitative information and provide explanation about process behaviour in a way which gives insight into the underlying physical processes. Despite this potential, existing techniques still present limitations related to the tendency for generating non-real behaviour patterns and the inability to describe distributed parameter systems. This study presents a qualitative reasoning methodology, weighted digraph (WDG) approach, for describing the dynamics of complex chemical processes, and in particular of distributed parameter systems, with a considerable reduction in the generation of spurious solutions. It is based on a generalisation of the signed digraph approach and retains its main advantages, such as the ability to easily represent intuitive and causal knowledge and a graph structure which makes apparent the flow of information between variables. In addition, it incorporates several new features, making use of functional weighting, differential nodes and temporal edges, which enable the procedure to qualitatively describe complex patterns of behaviour. The effectiveness of the approach is demonstrated by considering the qualitative modelling and simulation of the dynamic behaviour of several chemical processes: heat-exchanger, CSTR with and without temperature control and distillation column. The proposed weighted digraph approach is used to support generation of start-up procedures with reference to two case studies: a network of heat-exchangers and an integrated system composed of a CSTR and a feed/effluent heat-exchanger. It is shown that the digraph based strategy has the ability to generate feasible operating procedures in the presence of operational constraints and identify the need for modifications of the process topology in order to allow the start-up of the system. Results also indicate that work is still needed in order to further improve the methodology and create an interactive computer based interface to help with reasoning about complex patterns of behaviour.

With the increasing demand for optimal control and operation of plant, in particular chemical reactors, the need for detailed models which can be solved by computer in a reasonable time is apparent. Clearly, since the model is used repeatedly in each iteration of the computation, this will nor.mal~ mean that it must be relatively simple. Unfortunately, this results in loss of detail necessar,y to take full advantage of the optimisation. A technique for model reduction, suitable for the two dimensional heterogeneous catalytic reactor has been developed, which results in substantial reduotion in dimensionality of the system, but which retains the essential detail. A general reaction scheme with first order kinetics has been considered. Furthermore, it is possible to relate the well defined physical parameters, e.g. transport coefficients and rate constants, etc. to the parameters in the reduced model. Significant savings in computation time are made which makes it feasible to use the reduced model in optimisation and control schemes.

This thesis discusses the use of thermodynamic Second Law analysis in the context of chemical process network design. It is divided into two parts. Part I is based on the study of entire processes while Part II concentrates on the problem of heat exchanger network design. This division into two parts facilitates a clear presentation of the results obtained. Second Law analyses are frequently referred to in the academic literature as giving a more valid account of inefficiencies in engineering systems than simple heat balances, a view that would seem to be well founded in thermodynamic theory. On the other hand, process design engineers in industry do not seem to make much use of this type of analysis. They usually comment that the results obtained either state the obvious (e.g., "... do not degrade heat...", etc.) or lead to recommendations that are not practical (e.g., "... use fuel cells instead of thermal reactors...", etc.) Thus, there seems to be a conflict between theoretical claims and practical experience. The present thesis attempts to clarify this situation by giving a balanced view of both the potential value of Second Law analysis as well as its shortcomings. In Part I, it is shown that Second Law analyses are both difficult to produce and difficult to interpret in the context of chemical process design. Consequently, an approach is developed to overcome these difficulties. However, the approach somewhat transforms the meaning of the words "thermodynamic analysis". Namely, it is no longer a strict application of Second Law textbook theory that is implied, but a rather more broad minded approach involving the use of carefully considered thermodynamic concepts. In other words, a somewhat "slackened" form of thermodynamic analysis is recommended. This slackened form is less well defined than the classical one but easier to produce and more meaningful to interpret. (A more detailed explanation of the concepts involved is given in the "Extended Abstract of Part I" on page iii) . When applied to the case studies, the approach leads quickly to attractive design suggestions. In Part II, the approach is applied to the rather specialised and recently well researched problem of heat exchanger network design. This leads to the development of several explicit design methods which consistently identify better solutions for identical problems than methods described previously in the literature. Also, an understanding of the subject matter is achieved that is quite unprecedented (see "Extended Abstract of Part II" on page v). These results seem to support the conclusions arrived at in Part I. An interesting observation made in Part I and Part II is that some of the designs put forward would not only be cheap to run but also cheap to build. Thus, the thermodynamic approach developed here appears to be capable not only of identifying energy savings but capital savings, too. This gives rise to a very ambitious claim (compare pp. 270-271) : namely, that the approach may lead to designs that are "driving force conscious" in general. In other words, it may help the engineer to develop a better feel for the natural driving forces in his problem and may therefore stimulate him to create, in a quite general sense, more appropriate, simpler, and more elegant processes.

An integrated framework for process monitoring is developed in this study which consists of three components: (1) feature extraction from dynamic transient signals using multiscale wavelet transform; (2) operational state identification using unsupervised and recursive learning methods; and (3) automatic extraction of knowledge rules from process operational data and embedding of the extracted knowledge in the structure and weights of fuzzy-neural networks. The methodologies and the prototype system which have been developed are illustrated and evaluated using data collected from a dynamic simulator of a refinery catalytic cracking process. Methods for pre-processing dynamic transient signals for feature extraction, dimension reduction and noise removal are investigated and a new method is developed which makes use of wavelet transform to determine the singularities and irregularities of a dynamic transient signal by identifying the extrema from wavelet multiresolution analysis. The method is able to reduce the dimensionality of the data and removes noise components in a single step as well as capturing the most significant components of the dynamic response. A modified version of the unsupervised neural network ART2, designated ARTNET, has been developed which uses wavelet feature extraction to provide a substitution of the data pre-processing part of ART2. ARTNET is shown to be more effective in avoiding the adverse effects of noise, less sensitive to user defined parameters and faster in computation, as well as still retaining the advantages of unsupervised and recursive learning. Based on this, a fuzzy neural network is developed which is able to automatically extract knowledge rules from process data. The knowledge rules which are generated are transparent and explicit to operators. The method is therefore able to bridge the gap between numerical data and qualitative knowledge and takes advantage of the features of neural networks for capturing concepts and so provides an effective and robust method for learning knowledge from process data. Various methods for integrating different facets of a problem, and making use of this information in parallel to mutually compensate for drawbacks of any single approach are also exploited. Data obtained from a dynamic simulator of a refinery fluid catalytic cracking process (FCC) has been used to illustrate the methodologies and to evaluate a prototype system for using these new approaches. FCC provides a very useful case study because of the highly non-linear dynamics arising from the strong interactions between the reactor and fluidised bed regenerator derived from the mass and momentum balance. The use of simulation data makes it possible to look at the results in detail so that the methods can be fully tested. The case studies illustrate the potential of the methods developed.

Mathematical models are developed for a steam reforming furnace in both the steady state and unsteady state modes of operation. The one-dimensional radiation field in a participating medium is resolved into various classes of rays which distribute radiant energy to the furnace tubes carrying the process gas. Results predicted by the model are found to compare satisfactorily with values measured on an actual plant. A preliminary analysis of the thermal shock effects that arise during transient conditions in the furnace is made and limits on the safe rote of change of load level are indicated.

Various aspects of the behaviour of fixed bed reactors supporting highly exothermic reactions, relevant to stable -optimal design and control, have been studied using detailed mathematical models. In order to establish the form of the simplest basic structure, two methods of describing radial heat transfer in two dimensional packed beds have been examined. It is shown that a lumped parameter single phase heat transfer model which implicitly incorporates the heterogeneous structure can account not only for the radial heat flux associated with both the fluid and solid phases but is also the more appropriate formulation since it allows the important reaction rate limitations due to intraparticle mass transfer to be properly estimated. Using this two dimensional, heterogeneous dynamic model of the reactor it has been possible to evaluate a simpler one dimensional formulation. It is shown that the latter gives an adequate-description of the dynamic behaviour of the system, provided that the overall heat transfer coefficient between the fluid and the coolant is suitably defined, and may, therefore, be used for general studies of reactor performance. Consideration has been-given to the response of the reactor to sinusoidal perturbations of the inlet conditions. It has been found that at certain frequencies of oscillation temperature runaway may develop before a safe quasi-stationary state is reached. A detailed examination of this behaviour has shown that in addition to the non-linear effects the difference in the speeds of propagation of the concentration and temperature waves along the reactor as a result of the heterogeneity of the system is also very significant. The effect of both cocurrent and countercurrent cooling of a single reactor tube has been examined. The behaviour for perturbations in coolant temperature is similar to that for inlet temperature and indicates potential difficulties in the design of control systems. A mathematical model of a multitubular reactor with crossflow cooling has been developed and used to identify some of the problems which may arise in these systems. In particular, considerable interaction between the individual reactor tubes occurs when significant conversion of the reactant takes place. This causes tubes in different parts of the bundle to exhibit different behaviour and with countercurrent cooling this may give rise to multiple steady states due to the feedback of heat within the system. A technique has been developed for predicting regions of parametric sensitivity and temperature runaway in heterogeneous fixed bed reactors. The relationship between this form of instability and that due to multiple states of the catalyst pellet has been demonstrated. Application of this method to both the design and control of a reactor is discussed and it is shown that it provides an insight as to the behaviour of the system since it makes possible the establishment of a relationship between local and global reactor stability and the operating variables.

Consideration has been given to various aspects of the behaviour of multitubular fixed bed catalytic reactors supporting highly exothermic reactions. In particular, the problem of adequately representing the detailed characteristics of such reactors has been investigated, and mathematical models describing the steady state and dynamic behaviour of both co- and counter-currently cooled systems have been introduced as a basis for design and control studies. Valuable insight into the operational characteristics of large multitubular reactors has been obtained by investigating the phenomena related to the interactive heat transfer within the system, and such results provide a very useful assessment of how flexible the final unit might be in its ability to accomodate new operating conditions. The role of the coolant, and the manner in which it is presented to the tubes, has been shown to be especially important. Not only is the routing of the coolant through the bundle significant, but also the flowrate. Relatively small variations in the rate can cause very large changes in the temperature profiles within the tubes. This is not primarily a result of the heat transfer coefficient being modified, but because of the change in the residence time of the cooling fluid and the consequent effect on the coolant temperature rise. The results of an investigation into the configuration and coolant flow direction have shown that there can be considerable economic advantages in using co-currently cooled reactors with more than two shell-side passes. Such an arrangement is significantly more stable than a counter-current configuration for a wide range of coolant flow conditions. Furthermore, the effective feed forward of the heat in the coolant gives more even temperature profiles inside the reactor tubes which can result in a greater overall conversion. A steady state method has been developed which is capable of representing the ranges of coolant and reactant conditions under which the reactor is stable. It is shown how this may be extended to enable the local stability of the tubeside to be related to the overall stability of the unit. Thus, the stability of the system is related to easily obtainable variables, namely the coolant and reactant inlet temperatures and coolant flowrate. Dynamic models of the reactor have been formulated and used to demonstrate that the initial transient response of the system can lead to temperature runaway even though both the initial and final stationary states are stable. This behaviöur, which would not be predicted by normal frequency response techniques, clearly has significant effects on the design of the system and its control strategy.

A novel cyclic reactor system is proposed for heterogeneous, catalytic, gas-phase reactions. This system utilises the inherent characteristics of the thermal regenerator to impose favourable reaction temperature profiles along the catalyst bed without setting up radial temperature gradients. This control of the longitudinal profile enables higher conversions to be obtained than those from steady state reactors. The reactor system is investigated by computer simulation using the endothermic, reversible dehydrogenation of ethylbenzene to styrene in the presence of steam as an example. The higher conversions obtained from the proposed system produce utility cost savings in this process. Kinetics presented in the literature for this reaction are compared and assessed. None of these is entirely satisfactory and a more representative set is derived. Models for the reacting and regenerating bed are discussed and suitable models are presented. A comparative study of solution methods for these models is carried out in order to determine one which gives an accurate solution and also minimises computing requirements. The most suitable operating policy lor the system, with an endothermic reaction, is the use of constant heat inputs with constant flow during each period of operation, This allows the bed inlet temperature to vary with time, but it seems likely that the damping effect of the system will be large and the inlet temperatures may be assumed constant Counter-current, rather than co-current, operation of the system is preferred, A simple design procedure, which does not require the solution of the cyclic model, is described. This is found to give good predictions of the cyclic steady state performance of the system. Che effect of the various system parameters on the performance is investigated. The major parameters for a given bed size are the period time, reactor and regenerator steam flows and regenerator inlet temperature. It is shown that the system can give higher conversions than a steady state reactor but i.t may be desirable to operate at lower conversions to reduce the operating cost. Guidelines for optimising the system are discussed.

Abnormal operation of chemical processec sa used by equipment and sensor faults, such as plugging of pipes, control system failure or improper operation by personnel can result in poor product quality, equipment damage, or a catastrophe process failure leading to loss of equipment and worker injury, as well as significant economic losses. It is estimated that the cost attributable to preventable losses in the petrochemical industry only is around several billion pounds per year. Independent studies of case histories by the Health and Safety Commission in the UK and by a Honeywell led industrial consortium in the US and world wide show that human errors represent the major cause of failures. In contrast to this discovery, the majority of pervious studies on computer aided systems for fault detection and diagnosis has focussed on the process side only. It is now widely acknowledged that there is only limited information on how human factors can be assessed and even less that is specific to chemical industry, therefore research is much needed in this area. This study presents a methodology to involve human factors into the development of systems for automatic identification and diagnosis of abnormal operations and develops methods and techniques that can be used to simultaneously capture, characterise and assess the performance of operators as well as of the process. A joint process operator simulation platform was developed which was used as a test-bed for carrying out the studies. The process part is a simulator, which emulates in high fidelity the dynamic behaviour of the process, which is subject to influence of various disturbances and operators intervention. The operator module was developed as a real-time expert system, which emulates operator's behaviour in interpretation of received signals, planning and executions of the decisions. The interaction between the two modules is managed through an interaction module, which handles the real-time exchange of data using DDE (Dynamic Data Exchange). The interaction module also contains the toolkits for analysing the dynamic behaviour of the joint process-operator system. The operator simulation module was developed based on a theoretical model of human behaviour, which breaks operator's activities into perception of signals an interpretation of the received information, planning for actions and execution of the decisions. The system was implemented as a real-time expert system using visual Prolog. Numerical models were also integrated into the expert system, e. g. stress models of operators. This flexible system allows studies on individual operators actions, stress, intervene time, the frequency of intervene and near-miss or near-hit in operation. As part of the effort to use the platform to develop methods and tools for characterising and assessing the dynamic behaviour of the joint process-operator system, a digraph method for qualitative/quantitative modelling of the dynamic behaviour of the combined system was proposed. The method involves categorical characterisation of dynamic trends using principal component analysis and fuzzy c-means and sectioning of the clusters. An iterative method for determining the number of the clusters and sections based on the global performance was derived. Compared with pervious studies on qualitative process modelling, the proposed approach is more accurate and has higher resolution, and more importantly is able to deal with joint process-operator systems. The methods and systems developed were illustrated and fully tested using simulated and industrial case studies.