Vassilis Gaganis

Brief Paper: Real-time control of manufacturing cells using dynamic neural networks

In this paper, a control aspect of the non-acyclic FMS scheduling problem is considered. Based on a dynamic neural network model derived herein, an adaptive, continuous time neural network controller is constructed. The actual dispatching times are determined from the continuous control input discretization. The controller is capable of driving system production to the required demand and guaranteeing system stability and boundedness of all signals in the closed-loop system. Modeling errors and discretization effects are taken into account thus rendering the controller robust. A case study demonstrates the efficiency of the proposed technique.

Accurate determination of aromatic groups in heavy petroleum fractions using HPLC-UV-DAD

The purpose of this work is to identify specific aromatic component groups in heavy petroleum fractions based on their elution times and UV spectrum patterns using high performance liquid chromatography with ultraviolet diode array detection (HPLC-UV-DAD). Multivariate statistical methods such as evolving factor analysis and k-means clustering were used to interpret the signal obtained from the analysis of gas oil samples, in which strong overlapping of the eluting aromatic component groups occurs. The presented method allows the precise determination of the elution profiles of a series of aromatic component groups and therefore can be applied for their accurate quantitative determination.

A recurrent neural network model to describe manufacturing cell dynamics

A neural network approach to the manufacturing cell modelling problem is discussed. A recurrent high-order neural network structure (RHONN) is employed to identify cell dynamics, which is supposed to be unknown. The model is constructed in such a way that enables the design of a controller which will force the model and thus the original cell to display the required behaviour. The control input signal is transformed to a continuous one so as to conform with the RHONN assumptions, thus converting the original discrete-event system to a continuous one. A case study demonstrates the approximation capabilities of the proposed architecture.

A novel approach for the characterization of aromatics in petroleum fractions using HPLC-UV-DAD and evolving factor analysis

The scope of this work is to investigate the applicability of the evolving factor analysis (EFA) in the characterization of the non-saturated fraction of petroleum mixtures by high performance liquid chromatography coupled with UV diode array detection (HPLC-UV-DAD). The EFA was used to interpret the signal obtained from the analysis of fluids, the aromatic peaks of which overlap to a great extent. It is shown that the method can precisely define the elution profile of each one of the constituents and determine the start and the end elution times of the main aromatic hydrocarbon groups based on their differences in the spectrum pattern. In addition, the method gives the UV spectra of the eluting constituents which can be further employed for the compositional characterization of the aromatic part of oil fractions.

Neuro schedulers for flexible manufacturing systems

In this paper a dynamic neural network (DNN)-based controller is constructed to provide the basis upon which a scheduler is developed to guarantee that system production will reach the required demand while satisfying buffer capacity constraints. Lyapunov stability theory is used to prove boundedness of all signals in the closed loop.

Neural networks in manufacturing cell design

This paper presents a neural network approach in determining the appropriate manufacturing cell configuration that meets the required performance measures. Simulation experiments were conducted with many possible combinations of design changes to calculate cell performance measures, and thus generate training pairs for a neural network. Three different static neural network structures have been trained using the above data. Comparison of neural network efficiency and computational effort required is made through a case study, for every neural network architecture.

Dynamic neural networks for real time control of FMS

A control aspect of the non-acyclic FMS scheduling problem is considered. Based on the dynamic neural network model derived herein, an adaptive, continuous time neural network controller is constructed. The actual dispatching times are determined from the continuous control input discretization. The controller is capable of driving the system production to the required demand and guarantees system stability and boundedness of all signals in the closed loop. Modeling errors and discretization effects are taken into account thus rendering the controller robust. A case study demonstrates the efficiency of the proposed technique.

Rapid phase stability calculations in fluid flow simulation using simple discriminating functions

Abstract This paper presents a new phase stability method that is applicable when repeated phase behavior calculations are needed as it is the case with multiphase fluid flow compositional simulation in upstream petroleum engineering. Two discriminating functions act as classifiers in such a way that a positive value of one of the two functions determines the stability state of the mixture. The two functions are generated off line, prior to the simulation, and their expressions are very simple so that they can be evaluated rapidly in a non-iterative way for every discretization block and at each timestep during the simulation. The CPU time required for phase stability calculations is dramatically reduced while still obtaining correct classification results corresponding to the global minimum of the system Gibbs energy function. The method can be applied to any chemical engineering problem where the class of several objects needs to be determined repeatedly and quickly.

High order neural networks to control manufacturing systems-a comparison study

In this paper the neuro adaptive scheduling methodology is evaluated by comparing its performance with conventional schedulers, through simulation studies. The case study chosen constitutes an existing manufacturing cell, which can be viewed as a highly complex nonacyclic FMS, with extremely heterogenous part processing times. The results reveal superiority of our algorithm in terms of backlogging and inventory cost, system stability and work-in-process.

Neural Nets and Multichannel Image Processing Applications

The application of neural network technology to multichannel image processing is presented in this paper. Topics such as image restoration, segmentation, transformation and compression are discussed, covering a wide range of image processing and analysis areas. The problems are converted to optimisation or interpolation problems through the appropriate mathematical interface. This alternative interpretation enables the application of well-known neural network structures, perfectly suited for such purposes due to the accuracy and high speed of computation. The proposed framework handles multichannel data in a compact form and, thus, it is directly applicable to remote sensing.