Adrian A. Hopgood

Fuzzy logic in a blackboard system for controlling plasma deposition processes

Abstract A blackboard system, ARBS, has been used to control a plasma processing unit, which is used for depositing coatings on the surface of electronic or mechanical components. Previous applications of ARBS have been based on crisp logic, but fuzzy logic was added in this study for plasma deposition control. Fuzzy rules have been introduced into ARBS without changing either the rule syntax or the existing inference engines, thereby demonstrating the flexibility of the software. Consequently crisp and fuzzy rules can coexist within a single knowledge source (i.e. module). An efficient technique for defuzzification has been employed in which the membership functions are replaced by Dirac delta functions. The technique is equivalent to standard methods of defuzzification, without loss of precision or accuracy, but with a reduced number of calculations. Multi-variable control of DC-bias (an electrical parameter) by automatic adjustment of pressure and RF (radio frequency) electrical power is demonstrated.

Formation of misfit dislocations during growth of InxGa1-xAs/GaAs strained-layer heterostructures

Transmission electron microscopy of GaAs/InxGa1-xAs/GaAs double heterostructures has enabled the onset and subsequent development of misfit dislocations to be followed for increasing strained-layer thicknesses, from sub- to supercritical. It has been observed that misfit segments are introduced into threading dislocations when the strained-layer thickness is close to, but below, the critical thickness predicted by the Matthews-Blakeslee (M-B) model. Analysis shows that threading dislocations may be able to glide to form interfacial misfit dislocation segments even though the critical thickness predicted by the M-B model has not been reached. It has also been observed that the total dislocation density rises slowly as the layer thickness increases above its critical value, until a sudden increase occurs. It is suggested that the sudden increase in dislocation density is associated with a different mechanism of misfit dislocation formation, which dominates the global relaxation of the structure.

Optimum Work Roll Profile Selection in the Hot Rolling of Wide Steel Strip Using Computational Intelligence

The finishing train of a hot strip mill has been modelled by using a constant volume element model. The accuracy of the model has been increased by using an Artificial Neural Network (ANN). A non-linear Rank Based Geaetic Algorithm has been developed for the optimization of the work roll profiles in the finishing stands of the simulated hot strip mill. It has been compared with eight other experimental optimization algorithms: Random Walk, Hill Climbing, Simulated Annealing (SA) and five different Genetic Algorithms (GA). Finally, the work roll profiles have been optimized by the non-linear Rank Based Genetic Algorithm. The quality of the strip from the simulated mill was significantly improved.

Rule-based control of a telecommunications network using the blackboard model

Abstract A rule-based system has been built for automated control of a telecommunications network. The system has been tested with a simulation of the British Telecom trunk network, which it monitors and controls. Knowledge of network management is encoded as rules within a blackboard system (ARBS). At runtime, a model of the problem and its solution evolves in a global memory area, i.e. the blackboard. The knowledge-base is divided into separate modules, or knowledge sources, which may be rule-based or procedural. Different rule-based knowledge sources can use different inference mechanisms, whose merits are compared here. Four different aspects of network control have been addressed, along with the facility to reduce or disable previously issued commands. The delay in receiving a response to a request for routing information has been simulated with ARBS, and the intervening period used to perform other related tasks.

Analysis of dark-line defect growth suppression in InxGa1−xAs/GaAs strained heterostructures

The driving force of 〈100〉 dark-line defect (DLD) climbing growth based on vacancy unsaturation is discussed. In InxGa1−xAs/GaAs strained structures, it is found that compressive strain can reduce the osmotic (climb) force and can suppress the climb of DLDs in 〈100〉 direction. The percentage of indium in InxGa1−xAs/GaAs strained heterostructures for the suppression of 〈100〉 DLD propagation is calculated under different material growth temperatures and doping levels. For an n-type doping level higher than 5×1016u2002cm−3, an indium percentage less than approximately 9% in InxGa1−xAs/GaAs heterostructures is sufficient to stop the 〈100〉 DLDs growth and agrees well with the experimental observation. These results are useful for the design and fabrication of high reliability strained heterostructure devices.

Edge-type misfit dislocations produced by thermal processing of pre-relaxed InxGa1−xAs/GaAs heterostructures

Dislocation structures are presented for GaAs/InxGa1−xAs/GaAs heterostructures before and after thermal processing. Cathodoluminescence has allowed nondestructive examination of bulk specimens, while transmission electron microscopy has been used to establish the details of the dislocation configurations. In each as-grown specimen the thickness of the InxGa1−xAs layer was above its critical value, so 60° misfit dislocations were already present. It is shown that new pure edge, i.e., 90°, dislocations are formed at the interfaces by thermal processing at 1040 K. Their Burgers vectors are a/2〈101〉 perpendicular to their 〈010〉 directions. Although individual 90° misfit dislocations are more effective relievers of strain than 60° ones, the self-energy for an array of such dislocations is higher and hence 60° misfit dislocations form first. A model of the formation of 90° edge misfit dislocations is proposed in which the climb of vacancy-producing jogs on pre-existing 60° dislocations leaves a trailing disloca...

Kinetics of non-radiative-defect-related degradation in GaAs/AlGaAs heterojunction bipolar transistors

We show that a vacancy-controlled model can explain the experimentally observed negative dependence on the stress temperature of GaAs/AlGaAs heterojunction bipolar transistors degradation by non-radiative defects. The driving force (osmotic force) of the degradation versus the bias stress temperature is calculated and a decreasing temperature dependence is found. This model, which has previously been applied to strained-layer lasers and light emitting diodes, will help us understand the reported experimental observations on the temperature-acceleration tests for the GaAs/AlGaAs-based heterostructure devices such as heterojunction bipolar transistors.

An inference mechanism for selection, and its application to polymers

Abstract This paper describes two knowledge-based computer systems which have been built to assist product designers in the choice of appropriate polymer(s) from which to manufacture a component or product. The first system is novel in its application of inference under uncertainty to problems of selection. Probabilities (expressed as odds) are used to represent performance values of polymers for each property under consideration. The method of Bayesian updating is used to combine the probabilities to reach an overall conclusion. The shortcomings of this approach are critically discussed and are used to justify the development of a more appropriate inference mechanism (AIM) for dealing with problems of selection. AIM does not purport to be a general and rigorous means of manipulating probabilities. Instead it has been designed specifically as a useful and realistic means of combining performance values for separate criteria, when making a selection. AIM has been adopted in the second polymer selection system.

Intelligent control of low pressure plasma processing

Several parameters characterize systems for materials processing that use radio frequency electrical discharges in gases at low pressure. These include directly measurable quantities such as a DC bias voltage, an ion current, an energy flux, masses of charged species, and spectrally resolved optical emission. None of these is directly controllable but all are dependent on several variables that can be controlled such as radio-frequency (RF) power, chamber pressure, and gas flow rates. There is a rich parameter space that must be painstakingly searched for optimum conditions for any particular process. In place of the relatively slow manual procedure, an artificial intelligence (AI) approach has been used to map out contours for all of the above characteristic parameters in the control space. Automatic characterization of plasma systems in this way could significantly reduce the time to re-configure them and to transfer processes between different systems.

Hybridization Techniques in Optical Emission Spectral Analysis

The utilisation of formal artificial intelligence (AI) tools has been implemented to produce a hybrid system for optical emission spectral analysis that combines a multilayer perceptron neural network with rule-based system techniques. Even though optical emission spectroscopy is extensively used as an in-situ diagnostic for ionised gas plasmas in manufacturing processes, ways of interpreting the spectra without prior knowledge or expertise from the users stand-point has encouraged the use of Al techniques to automate the interpretation process. The hybrid approach presented here combines a modified network architecture with a simple rule-base in order to produce explicit models of the identifiable chemical species.