Design and Implementation of Intelligent Classifier and Size Estimator for Leakage in Oil Pipelines

Dr. Hanan A. R. Akkar, Dr. Wael A. H. Hadi, Ibraheem H. M. Al-Dosari

Almost industrial activities play an important role in human life by delivering the adequate resources. However oil and gas pipelines represent the back bone for these industrial resources due to oil and gas transportations around the world. But unfortunately sometimes great losses are recordedannually due to leakages in thesepipelines. So this idea attracts most of the researcher to shrink the heavy losses by incorporating the artificial intelligence theory to predict the behavior for these pipelines and estimate the probability for leakage occurrence and provide enough information about leak position and size rather than a prior protective actions against the corrosion and environmental risks if exist. This work adopts a new proposed neural network model with back propagation algorithm as one of the popular methods for leak detection in a pipeline which is usually used to classify the leak size and position along the pipeline. Achieved results of the work in this paper explained the difference between various transfer functions used for hidden and output neurons in ANN, also the confusion matrix for each learning algorithms shows the outperformance for the batching method against the incremental method for weights updating in the BPNN.Also different neuron based transfer functions with various characteristics are compared from classification accuracy and training performance points of view.