Theses and Dissertations (Electrical & Mining Engineering)
http://hdl.handle.net/10500/2922
2016-08-28T00:26:24ZMeasurement, modelling and optimization of three-phase submerged ARC furnaces (SAF)
http://hdl.handle.net/10500/6588
Measurement, modelling and optimization of three-phase submerged ARC furnaces (SAF)
Amadi, Amos
This thesis investigates the modelling and optimization of electro-thermal variable parameters
applicable in obtaining an optimal operating point in SAFs. Graphite electrodes that are
symmetrically positioned around the furnace are used to convert electrical energy to heat
energy via three-phase arcs. The raw materials are fed via conveyor belts from the top of the
furnace and are smelted by the arcs produced by the electrodes. The charge constitutes the
resistance variable, whilst the heat emitted from the molten charge constitutes the temperature
variable. The supply voltage to the furnace constitutes the last variable and it suffers from the
network disturbances such as harmonics, dips, surges and others.
Although there are many variables that are involved in submerged arc furnace operations, the
scope of this thesis is restricted to three electro-thermal variable parameters namely,
resistance, voltage and temperature. The measurement of these parameters need to be done
accurately and controlled effectively in order to achieve optimum output power during the
furnace operation. An amalgamated variable parameter measurement (AVPM) system is
proposed for the accurate measurement of these variables by use of mathematically modeled
modules. The verification of this proposed measurement system is not considered in this
thesis as it is recommended for future study.
Modelling is difficult using mathematical functions according to the mechanisms of the actual
furnace plant system because of its complexity and many disturbances. The neural networks
have been chosen because of its easy to use in modelling nonlinear functions such as the
furnace plant. In this thesis, the furnace plant is modeled with the neural networks (NN)
algorithm to obtain the SAF NN model. The model is then optimized using the particle swarm
optimizer (PSO) algorithm. The formulated PSO based SAF NN modelâ€™s results are also
validated using the real SAF plant samples.
2012-06-01T00:00:00Z