Authors

Abstract

The intelligent techniques are used successfully in a broad band of applications
one of these applications is the cathodic protection system. Examples of these
techniques used in cathodic protection are fuzzy logic and genetic algorithms. The
present work aims to use the neural network to predict the minimum current
density required in impressed current cathodic protection to protect low carbon
steel pipe which have been related previously.[1].
This work deals with choosing the best network architecture for cathodic
protection system. This step used multilayer feed forward network four
environment variables (concentration C%, temperature T, distance D and pH) as
input to identify the minimum current density as output in a feed forward network
structure with one hidden layer using the practical results data for the learning
process. The best number of neurons in the hidden layer is chosen by trial and error
and it is found to be 25 neurons. the decision function used is the tan training
algorithm with one variable learning rate. Then, neural network training is done
using 25 data samples from the experimental data for the current density in the
above four variables conditions. The stopping criterion for training was to obtain a
sum square error of 0.001 or read 10000 Epochs. An (SSE) than 0.001 were
obtained after 5226 Epochs.
Generalization test used 5 data samples taken from the experimental results
other than those data samples used in the learning process to check the
performance of the neural network on these data. The SSE for these samples was
0.0053 and it shows a good generalization results for our application. The
comparison between the actual experimental output and the neural network out put
after the learning process are almost identical which indicates that good learning
process was achieved.
The practical results indicate that neural network system can be used
successfully to obtain minimum cathodic protection current density to protect low
carbon steel pipes.