ANALYSIS OF WAX DEPOSITION IN TRANSPORTATION PIPE USING MACHINE LEARNING APPROACH

• Main Author: Septiano, Jalest
• Format: Theses
• Language: Indonesia
• Subjects: Pertambangan dan operasi berkaitan
• Online Access: https://digilib.itb.ac.id/gdl/view/65322
• Institution: Institut Teknologi Bandung

Wax deposit is one of the most serious issues encountered during the oil production process from subsurface to surface. Oil wax is a solid molecule with a carbon number greater than 18, which is common in paraffinic oils. Wax deposited in the pipe occurs as a result of a drop in temperature near the oil flow in the pipe, where wax crystals will begin to emerge when they come into contact with the wax appearing temperature (WAT). In this case, the wax will continue to be deposited in accordance with the wax's solubility in the oil until it reaches the lowest temperature in the oil flow in the pipe. The development of computer science, in this case machine learning, helps in the prediction and classification of petroleum problems. One of them is forecasting the rate of wax deposition in the pipe. Machine learning algorithms' high predictive ability can be utilized to compare them to other prediction methods. To forecast the wax deposition rate, four machine learning techniques are used: Random Forest, Adaboost, and Gradient Boosting for ensemble types, and linear regression algorithms. Based on the experimental results, it was discovered that the Adaboost algorithm had the best prediction performance, with MAE, MSE, and RMSE for the test data of 0.014, 0.003, and 0.052, respectively, and R2 on the test data of 0.997. The Adaboost approach is then implemented in various scenarios utilizing calculations to demonstrate the robustness of the machine learning made. On the 15th, 20th, and 25th days of forecasting the average thickness of the wax and estimating the maximum thickness of the wax below 10%, the relative error of each case was acquired. This demonstrates a good statistical value, implying that the Adaboost was robust.