A corrosion prediction model for oil and gas pipeline using CMARPGA

Pipelines are used as a medium to transport the oil, however, low maintenance causing not only the loss of the material itself but as well to the surrounding people and environment. In order to tackle the incidents, experts are assigned and experiments are conducted to analyze the source of the leakage. The leakage is often triggered by either natural disaster such as earthquake or human negligence such as low maintenance of oil pipeline. Natural disaster is unpredictable and it is difficult to prevent; therefore, researches are carried out in detecting corrosion of transmission pipelines. In this research, a new oil pipeline corrosion prediction model is proposed. An associative classification technique named classification based on multiple association rules is applied in the proposed prediction model. This proposed prediction model named CMARGA is then enhanced by using genetic algorithm in order build an optimum decision tree. The decision tree is said optimum in term of the genetic algorithm is used to examine the correlation between a group of association rules instead of using one single rule in predicting a case. The prediction model, CMARGA is tested against 15 datasets from UCI machine learning which yielded average accuracy of 80.2041%. After the validation, CMARGA is then tested against a simulated oil pipeline corrosion dataset consist of partial pressure carbon dioxide, velocity, and temperature. A good result of 96.6667% accuracy as single run validation is achieved; while, 96.0% accuracy obtained when runs through tenth cross validation.