Prediction of Corrosion Rate for Carbon Steel in Soil Environment by Artificial Neural Network and Genetic Algorithm
الموضوعات :
Amir Akhtari-Goshayeshi
1
(
Advanced Materials Research Center, Department of Materials Engineering, Najafabad Branch, Islamic Azad University, Najafabad, Iran.
)
Moslem Ghobadi
2
(
Department of Technical Inspection, Abadan Facility of petroleum engineering, Petroleum University of Technology, Abadan, Iran.
)
Ehsan Saebnoori
3
(
Advanced Materials Research Center, Department of Materials Engineering, Najafabad Branch, Islamic Azad University, Najafabad, Iran.
)
Alireza Zarezadeh
4
(
Department of Safety engineering, Abadan Facility of petroleum engineering, Petroleum University of Technology, Abadan, Iran.
)
Mohammad Rostami
5
(
Department of chemical Engineering, Abadan Facility of petroleum engineering, Petroleum University of Technology, Abadan, Iran.
)
Mohammad Nematollahi
6
(
Department of materials Engineering, Isfahan University of Technology, Isfahan, Iran.
)
الکلمات المفتاحية: Genetic Algorithm (GA), Corrosion, Artificial Neural Network (ANN), weight loss, Soil Environment,
ملخص المقالة :
In this study, the corrosion rates for St37 carbon steel in some soil types with different conditions were measured. The effects of the parameters of moisture amount, soil’s particle size and salt’s concentration were determined by the mass loss method. An Artificial Neural Network (ANN) model with three inputs and one output was established to simulate the experimental data. It was observed that Levenberg–Marquardt algorithm with hyperbolic tangent sigmoid transfer function provided the best results in training with the lowest MSE and MAE compared to the other methods in the model. The R values for training, validation, and test were presented and the value of 0.98684 was achieved for the complete data set which demonstrate a high level of ANN performance. The Genetic Algorithm (GA) was also used to find optimum inputs for the target of minimum corrosion rate value. The results showed a good agreement between the model prediction and experimental values.
