An artificial neural network model for the corrosion current density of steel in mortar mixed with seawater
Corrosion is a very complicated phenomenon in the field of science and engineering. Over the years, several numerical models have been developed to predict the damage caused by the corrosion process. The use of the artificial neural network in modeling corrosion has gained popularity in recent years...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | text |
| Published: |
Animo Repository
2019
|
| Subjects: | |
| Online Access: | https://animorepository.dlsu.edu.ph/faculty_research/3057 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | De La Salle University |
| Summary: | Corrosion is a very complicated phenomenon in the field of science and engineering. Over the years, several numerical models have been developed to predict the damage caused by the corrosion process. The use of the artificial neural network in modeling corrosion has gained popularity in recent years. Many of the factors affecting corrosion are difficult to control. Thus, the artificial neural network may be a better technique to consider due to its ability to tolerate relatively imprecise, noisy or incomplete data, less vulnerability to outliers, filtering capacity and adaptability. This study aims to generate a corrosion current density prediction model using the artificial neural network approach. Microcell corrosion current density is defined as the rate of corrosion expressed in electric current per unit area of cross-section. Several variables were considered as input variables namely: age, water to cement ratio, cement content, compressive strength, type of mixing water, corrosion potential, solution resistance, and polarization resistance. These variables were entered into the neural network architecture and simulated in MATLAB. The feedforward backpropagation technique was used to generate the best model for the corrosion current density. The best neural network architecture consists of 8 input variable, 8 neurons in the hidden layer and one output variable. The resulting neural network model satisfactorily predicted the corrosion current density with a coefficient of correlation values of 0.96536, 0.80817, and 0.7662 for training, validation and testing phases, respectively. © Int. J. of GEOMATE. |
|---|