The thesis is concerned with the development of data-driven methods for fault diagnosis of plant-wide disturbances. Industrial plants producing large quantities of liquid or gaseous chemicals run continuously and under tight cost, safety, quality and environmental constraints. Any unwanted variability in the form of disturbances in the production process affects one of these constraints. Worse still, disturbances can spread and cause large parts of the process to be upset. Detection and diagnosis of disturbances is therefore an important subject for chemical companies. Chemical processes are well equipped with modern instrumentation technology so that measurements of process variables such as flow, temperature or pressure are abundantly available. The research has used time series analysis for process measurements in a novel way. In particular, it focuses on measures to decide about cause and effect of processes variables to address the question whether A causes B or B causes A. Knowing the causal relation ship finds the fault propagation path in case of a disturbance and eventually traces the disturbance back to the root cause. Three different measures are investigated and developed for the application to chemical process data: one straight forward algorithm based on the cross-correlation function and two statistics based on nearest neighbours methods and transfer entropy. Together with the automatic generation of causal maps these approaches lead to a breakthrough in fault diagnosis. Guidelines for the parameters of the methods are tailored to signatures caused by disturbances common in chemical processes. A significance level is introduced for automatic implementation of industrial applications. Case studies with process disturbances, in particular those from a three months placement with Eastman Chemical Company, are analysed with the developed tools. The results are compared and recommendations of choosing the best method for a data set are generalised from results of the case studies.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

Plasma processing on industrial scale is becoming increasingly complex, demanding new strategies for process control and monitoring. Of particular interest is the energy transport in the interface region between the non-equilibrium low-pressure plasma and the surface. The individual plasma components have varying properties and exhibit different dynamics, which enable numerous chemical and physical modifications of surfaces simultaneously. Measurements of the in-situ surface condition and important chemically active radical species are extremely challenging. The most promising approach to overcome these challenges to achieve advanced process control is the active coupling of numerical simulations and experiments. In this regard, numerical simulations are a well-established technique to study fundamental plasma parameters and plasma dynamics for a variety of discharge sources. The utilised numerical simulation is an experimentally benchmarked 1D fluid model, with semi-kinetic treatment of electrons and an improved energy dependent ion mobility treatment. This model is applied for a geometrically symmetric and asymmetric capacitively coupled oxygen RF discharge. Within the investigated pressure range of 10 Pa - 100 Pa the simulations predict that changing surface conditions have a significant effect on dynamics of the plasma-surface interface. In particular, the surface loss probability and lifetime of metastable singlet delta oxygen as well as the secondary electron emission coefficient are identified to substantially influence the electronegativity and the plasma sheath dynamics on a nanosecond timescale. Phase resolved optical emission spectroscopy measurements, utilising different surface materials, confirm these predictions by comparing measured and simulated excitation features for three different optical emission lines. Through the synergistic coupling of numerical simulations and experiments, the surface work functions as well as other key surface parameters are assessed. Furthermore, the use of an advanced actinometry technique, demonstrated by coupling simple electron kinetic simulations and optical measurements, enables measurements of the spatial distribution of radical atomic oxygen densities and local electron energies over the total discharge volume.

This thesis presents a formal method for the the design of optimal and provably correct procedural controllers for chemical processes modelled as Stochastic Discrete Event Systems (SDESs). The thesis extends previous work on Procedural Control Theory (PCT) [1], which used formal techniques for the design of automation Discrete Event Systems (DESs). Many dynamic processes for example, batch operations and the start-up and shut down of continuous plants, can be modelled as DESs. Controllers for these systems are typically of the sequential type. Most prior work on characterizing the behaviour of DESs has been restricted to deterministic systems. However, DESs consisting of concurrent interacting processes present a broad spectrum of uncertainty such as uncertainty in the occurrence of events. The formalism of weighted probabilistic Finite State Machine (wp-FSM) is introduced for modelling SDESs and pre-de ned failure models are embedded in wp-FSM to describe and control the abnormal behaviour of systems. The thesis presents e cient algorithms and procedures for synthesising optimal procedural controllers for such SDESs. The synthesised optimal controllers for such stochastic systems will take into consideration probabilities of events occurrence, operation costs and failure costs of events in making optimal choices in the design of control sequences. The controllers will force the system from an initial state to one or more goal states with an optimal expected cost and when feasible drive the system from any state reached after a failure to goal states. On the practical side, recognising the importance of the needs of the target end user, the design of a suitable software implementation is completed. The potential of both the approach and the supporting software are demonstrated by two industry case studies. Furthermore, the simulation environment gPROMS was used to test whether the operating speci cations thus designed were met in a combined discrete/continuous environment.

The research has included collaboration with number of different companies and consortiums involving spectroscopic measurements with the application of chemometric techniques. For the 'European Framework 5', Standards Measurements and Testing (SMT) chemometrics network consortium a certified reference dataset based on visible metals complex spectra was developed. An inter-laboratory study was carried out which demonstrated the between subject significant difference for chemometric data analysis. An industrial collaboration with BNFL, Springfield's, this work consisted of producing a PLS regression model which could be used to predict levels of uranyl and nitrate in uranyl nitrate liquors samples, which were analysed by Raman spectroscopy which was insensitive to temperature. A substantial amount of work has been in the development of GMS with multivariate calibration for process analysis. The GMS is designed for the analysis of flowing mixtures, slurries and moisture content. The method is currently hindered by the existing calibration method; here PCA, PLS and weighted ridge regression (WRR) have been applied to the broadband, complex spectra to successfully allow measurement of a range of samples including; aqueous, organic, fermentation and non-homogeneous samples.