Efficient training of RBF networks for classification.

Radial Basis Function networks with linear outputs are often used in regression problems because they can be substantially faster to train than Multi-layer Perceptrons. For classification problems, the use of linear outputs is less appropriate as the outputs are not guaranteed to represent probabilities. We show how RBFs with logistic and softmax outputs can be trained efficiently using the Fisher scoring algorithm. This approach can be used with any model which consists of a generalised linear output function applied to a model which is linear in its parameters. We compare this approach with standard non-linear optimisation algorithms on a number of datasets.

Flexible Gaussian process wind field models

This technical report builds on previous reports to derive the likelihood and its derivatives for a Gaussian Process with a modified Bessel function based covariance function. The full derivation is shown. The likelihood (with gradient information) can be used in maximum likelihood procedures (i.e. gradient based optimisation) and in Hybrid Monte Carlo sampling (i.e. within a Bayesian framework).

A scatterometer neural network sensor model with input noise

The ERS-1 satellite carries a scatterometer which measures the amount of radiation scattered back toward the satellite by the ocean's surface. These measurements can be used to infer wind vectors. The implementation of a neural network based forward model which maps wind vectors to radar backscatter is addressed. Input noise cannot be neglected. To account for this noise, a Bayesian framework is adopted. However, Markov Chain Monte Carlo sampling is too computationally expensive. Instead, gradient information is used with a non-linear optimisation algorithm to find the maximum em a posteriori probability values of the unknown variables. The resulting models are shown to compare well with the current operational model when visualised in the target space.

Efficient training of RBF networks for classification

Radial Basis Function networks with linear outputs are often used in regression problems because they can be substantially faster to train than Multi-layer Perceptrons. For classification problems, the use of linear outputs is less appropriate as the outputs are not guaranteed to represent probabilities. In this paper we show how RBFs with logistic and softmax outputs can be trained efficiently using algorithms derived from Generalised Linear Models. This approach is compared with standard non-linear optimisation algorithms on a number of datasets.

A scatterometer neural network sensor model with input noise

The ERS-1 satellite carries a scatterometer which measures the amount of radiation scattered back toward the satellite by the ocean's surface. These measurements can be used to infer wind vectors. The implementation of a neural network based forward model which maps wind vectors to radar backscatter is addressed. Input noise cannot be neglected. To account for this noise, a Bayesian framework is adopted. However, Markov Chain Monte Carlo sampling is too computationally expensive. Instead, gradient information is used with a non-linear optimisation algorithm to find the maximum em a posteriori probability values of the unknown variables. The resulting models are shown to compare well with the current operational model when visualised in the target space.

Dynamic simulation of chemical plant

A CSSL- type modular FORTRAN package, called ACES, has been developed to assist in the simulation of the dynamic behaviour of chemical plant. ACES can be harnessed, for instance, to simulate the transients in startups or after a throughput change. ACES has benefited from two existing simulators. The structure was adapted from ICL SLAM and most plant models originate in DYFLO. The latter employs sequential modularisation which is not always applicable to chemical engineering problems. A novel device of twice- round execution enables ACES to achieve general simultaneous modularisation. During the FIRST ROUND, STATE-VARIABLES are retrieved from the integrator and local calculations performed. During the SECOND ROUND, fresh derivatives are estimated and stored for simultaneous integration. ACES further includes a version of DIFSUB, a variable-step integrator capable of handling stiff differential systems. ACES is highly formalised . It does not use pseudo steady- state approximations and excludes inconsistent and arbitrary features of DYFLO. Built- in debug traps make ACES robust. ACES shows generality, flexibility, versatility and portability, and is very convenient to use. It undertakes substantial housekeeping behind the scenes and thus minimises the detailed involvement of the user. ACES provides a working set of defaults for simulation to proceed as far as possible. Built- in interfaces allow for reactions and user supplied algorithms to be incorporated . New plant models can be easily appended. Boundary- value problems and optimisation may be tackled using the RERUN feature. ACES is file oriented; a STATE can be saved in a readable form and reactivated later. Thus piecewise simulation is possible. ACES has been illustrated and verified to a large extent using some literature-based examples. Actual plant tests are desirable however to complete the verification of the library. Interaction and graphics are recommended for future work.

Simultaneous chemical reaction and distillation of formaldehyde

The thesis is concerned with the development and testing of a mathematical model of a distillation process in which the components react chemically. The formaldehyde-methanol-water system was selected and only the reversible reactions between formaldehyde and water giving methylene glycol and between formaldehyde and methanol producing hemiformal were assumed to occur under the distillation conditions. Accordingly the system has been treated as a five component system. The vapour-liquid equilibrium calculations were performed by solving iteratively the thermodynamic relationships expressing the phase equilibria with the stoichiometric equations expressing the chemical equilibria. Using optimisation techniques, the Wilson single parameters and Henry's constants were calculated for binary systems containing formaldehyde which was assumed to be a supercritical component whilst Wilson binary parameters were calculated for the remaining binary systems. Thus the phase equilibria for the formaldehyde system could be calculated using these parameters and good accuracy was obtained when calculated values were compared with experimental values. The distillation process was modelled using the mass and energy balance equations together with the phase equilibria calculations. The plate efficiencies were obtained from a modified A.I.Ch.E. Bubble Tray method. The resulting equations were solved by an iterative plate to plate calculation based on the Newton Raphson method. Experiments were carried out in a 76mm I.D., eight sieve plate distillation column and the results were compared with the mathematical model calculations. Overall, good agreement was obtained but some discrepancies were observed in the concentration profiles and these may have been caused by the effect of limited physical property data and a limited understanding of the reactions mechanism. The model equations were solved in the form of modular computer programs. Although they were written to describe the steady state distillation with simultaneous chemical reaction of the formaldehyde system, the approach used may be of wider application.

Developing a scalable model of recombinant protein yield from Pichia pastoris:the influence of culture conditions, biomass and induction regime

Background The optimisation and scale-up of process conditions leading to high yields of recombinant proteins is an enduring bottleneck in the post-genomic sciences. Typical experiments rely on varying selected parameters through repeated rounds of trial-and-error optimisation. To rationalise this, several groups have recently adopted the 'design of experiments' (DoE) approach frequently used in industry. Studies have focused on parameters such as medium composition, nutrient feed rates and induction of expression in shake flasks or bioreactors, as well as oxygen transfer rates in micro-well plates. In this study we wanted to generate a predictive model that described small-scale screens and to test its scalability to bioreactors. Results Here we demonstrate how the use of a DoE approach in a multi-well mini-bioreactor permitted the rapid establishment of high yielding production phase conditions that could be transferred to a 7 L bioreactor. Using green fluorescent protein secreted from Pichia pastoris, we derived a predictive model of protein yield as a function of the three most commonly-varied process parameters: temperature, pH and the percentage of dissolved oxygen in the culture medium. Importantly, when yield was normalised to culture volume and density, the model was scalable from mL to L working volumes. By increasing pre-induction biomass accumulation, model-predicted yields were further improved. Yield improvement was most significant, however, on varying the fed-batch induction regime to minimise methanol accumulation so that the productivity of the culture increased throughout the whole induction period. These findings suggest the importance of matching the rate of protein production with the host metabolism. Conclusion We demonstrate how a rational, stepwise approach to recombinant protein production screens can reduce process development time.

IT support for managing aircraft spares in a closed-loop supply chain

This paper reports on the development of elements of an e-supply chain management system for managing maintenance, repair and overhaul (MRO) relationships in the aerospace industry. A standard systems development methodology has been followed to produce a process model (i.e. the AMSCR model); an information model (i.e. business rules) and a computerised information management capability (i.e. automated optimisation). The proof of concept for this web-based MRO supply chain system has been established through the collaboration with a sample of the different types of supply chain members. The proven benefit is a reduction in the stock-holding costs for the whole supply chain whilst also minimising non-flying time of the aircraft that the supply chain supports. This type of system is now vital in an industry that has continuously decreasing profit margins, which in turn means pressure to reduce servicing times and increase the interval between maintenance actions.

Designing a support system for aerospace maintenance supply chains

Purpose - To develop a systems strategy for supply chain management in aerospace maintenance, repair and overhaul (MRO). Design/methodology/approach - A standard systems development methodology has been followed to produce a process model (i.e. the AMSCR model); an information model (i.e. business rules) and a computerised information management capability (i.e. automated optimisation). Findings - The proof of concept for this web-based MRO supply chain system has been established through collaboration with a sample of the different types of supply chain members. The proven benefits comprise new potential to minimise the stock holding costs of the whole supply chain whilst also minimising non-flying time of the aircraft that the supply chain supports. Research limitations/implications - The scale of change needed to successfully model and automate the supply chain is vast. This research is a limited-scale experiment intended to show the power of process analysis and automation, coupled with strategic use of management science techniques, to derive tangible business benefit. Practical implications - This type of system is now vital in an industry that has continuously decreasing profit margins; which in turn means pressure to reduce servicing times and increase the mean time between them. Originality/value - Original work has been conducted at several levels: process, information and automation. The proof-of-concept system has been applied to an aircraft MRO supply chain. This is an area of research that has been neglected, and as a result is not well served by current systems solutions. © Emerald Group Publishing Limited.

Measurement efficiency and productivity in SAS/OR

This paper explores the use of the optimisation procedures in SAS/OR software with application to the measurement of efficiency and productivity of decision-making units (DMUs) using data envelopment analysis (DEA) techniques. DEA was originally introduced by Charnes et al. [J. Oper. Res. 2 (1978) 429] is a linear programming method for assessing the efficiency and productivity of DMUs. Over the last two decades, DEA has gained considerable attention as a managerial tool for measuring performance of organisations and it has widely been used for assessing the efficiency of public and private sectors such as banks, airlines, hospitals, universities and manufactures. As a result, new applications with more variables and more complicated models are being introduced. Further to successive development of DEA a non-parametric productivity measure, Malmquist index, has been introduced by Fare et al. [J. Prod. Anal. 3 (1992) 85]. Employing Malmquist index, productivity growth can be decomposed into technical change and efficiency change. On the other hand, the SAS is a powerful software and it is capable of running various optimisation problems such as linear programming with all types of constraints. To facilitate the use of DEA and Malmquist index by SAS users, a SAS/MALM code was implemented in the SAS programming language. The SAS macro developed in this paper selects the chosen variables from a SAS data file and constructs sets of linear-programming models based on the selected DEA. An example is given to illustrate how one could use the code to measure the efficiency and productivity of organisations.

A heuristic approach for component scheduling on a high-speed PCB assembly machine

The collect-and-place machine is one of the most widely used placement machines for assembling electronic components on the printed circuit boards (PCBs). Nevertheless, the number of researches concerning the optimisation of the machine performance is very few. This motivates us to study the component scheduling problem for this type of machine with the objective of minimising the total assembly time. The component scheduling problem is an integration of the component sequencing problem, that is, the sequencing of component placements; and the feeder arrangement problem, that is, the assignment of component types to feeders. To solve the component scheduling problem efficiently, a hybrid genetic algorithm is developed in this paper. A numerical example is used to compare the performance of the algorithm with different component grouping approaches and different population sizes.

The mechanics of the conform continuous extrusion process

The work describes the programme of activities relating to a mechanical study of the Conform extrusion process. The main objective was to provide a basic understanding of the mechanics of the Conform process with particular emphasis placed on modelling using experimental and theoretical considerations. The experimental equipment used includes a state of the art computer-aided data-logging system and high temperature loadcells (up to 260oC) manufactured from tungsten carbide. Full details of the experimental equipment is presented in sections 3 and 4. A theoretical model is given in Section 5. The model presented is based on the upper bound theorem using a variation of the existing extrusion theories combined with temperature changes in the feed metal across the deformation zone. In addition, constitutive equations used in the model have been generated from existing experimental data. Theoretical and experimental data are presented in tabular form in Section 6. The discussion of results includes a comprehensive graphical presentation of the experimental and theoretical data. The main findings are: (i) the establishment of stress/strain relationships and an energy balance in order to study the factors affecting redundant work, and hence a model suitable for design purposes; (ii) optimisation of the process, by determination of the extrusion pressure for the range of reduction and changes in the extrusion chamber geometry at lower wheel speeds; and (iii) an understanding of the control of the peak temperature reach during extrusion.

Time complexity and convergence analysis of domain theoretic Picard method

We present an implementation of the domain-theoretic Picard method for solving initial value problems (IVPs) introduced by Edalat and Pattinson [1]. Compared to Edalat and Pattinson's implementation, our algorithm uses a more efficient arithmetic based on an arbitrary precision floating-point library. Despite the additional overestimations due to floating-point rounding, we obtain a similar bound on the convergence rate of the produced approximations. Moreover, our convergence analysis is detailed enough to allow a static optimisation in the growth of the precision used in successive Picard iterations. Such optimisation greatly improves the efficiency of the solving process. Although a similar optimisation could be performed dynamically without our analysis, a static one gives us a significant advantage: we are able to predict the time it will take the solver to obtain an approximation of a certain (arbitrarily high) quality.

All-Raman amplified transmission at 40 Gbit/s in standard single mode fibre

We investigate a 40 Gbit/s all-Raman amplified standard single mode fibre (SMF) transmission system with the mid-range amplifier spacing of 80-90 km. The impact of span configuration on double Rayleigh back scattering (DRBS) was studied. Four different span configurations were compared experimentally. A transmission distance of 1666 km in SMF has been achieved without forward error correcting (FEC) for the first time. The results demonstrate that the detrimental effects associated with high pump power Raman amplification in standard fibre can be minimised by dispersion map optimisation. © 2003 IEEE.