Basic Informations

C.V

not found

Master Title

Intelligent Model for Preventing Oil Spill Using Deep Learning

Master Abstract

It's quite simple to learn about accidents after they've occurred, and dealing with their solutions is done right away. However, when looking for a remedy to these accidents, don't discount the factors that contributed to their occurrence. Accidents can be prevented in the future by being aware of their cause. It's crucial to use a strong technique to define these causes as the research point in spill accidents and predict accidents, as some human lives depend on the safety of such a field, and to predict the occurrence of an accident before it happens based on the factors that contributed to that accident in the past. Similar circumstances could result in an accident, but this requires enough data and a potent technology like deep learning techniques. The issue of oil and gas flaws that result in spills or explosions that cause numerous human casualties, as well as oil field exploitation and costs, are covered in this research. Petrol is an important field in our lives because it controls all aspects of human life and their way of life, so our research focused on petrol and its problems in order to introduce a better way of life. The data used in this research was taken from the 3w database that was prepared by Petrobras, the Brazilian oil holding. The 9 classes classified in that work include Abrupt Increase of Basic Sediment and Water (BSW), Spurious Closure of the Downhole Safety Valve (DHSV), Severe Slugging, Flow Instability, Rapid Productivity Loss, Quick Restriction in the Production Choke (PCK), Scaling in the Production Choke (PCK), and Hydrate in the Production Line. the normal state that indicates the factors that will not lead to a problem. Deep learning classification techniques were used in this study. 99% accuracy was obtained in that model, and it refers to a successful prediction and classification of each class. Different results were observed when different deep neural networks such as: hidden layers, optimizers, neurons, epochs, and activation functions were used. Our research gived that result by using Adam's optimizer and Tanh's activation function.

PHD Title

Not found

PHD Abstract

not found