A Neural Network-Based Model for Predicting Saybolt Color of Petroleum Products.

Area of Science:

• Petroleum Engineering
• Analytical Chemistry
• Artificial Intelligence

Background:

• Saybolt color is crucial for assessing petroleum product quality and guiding refinement.
• Current laboratory-based color measurement methods are time-consuming and expensive.

Purpose of the Study:

• To develop an automated artificial neural network (ANN) model for predicting Saybolt color.
• To compare the ANN model's performance against traditional multiple linear regression (MLR).

Main Methods:

• An ANN model was designed with five input variables: density, kinematic viscosity, sulfur content, cetane index, and total acid number.
• Two backpropagation algorithms were evaluated with varying transfer functions and neuron counts.
• Model performance was assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and coefficient of determination (R²).

Main Results:

• The ANN model, utilizing the Levenberg-Marquardt algorithm, tangent sigmoid transfer function, and three neurons, achieved superior performance (R² = 0.995, MAE = 1.000, RMSE = 1.658).
• The ANN model significantly outperformed MLR, which yielded an R² of 0.830.
• The developed ANN model demonstrated high accuracy in predicting Saybolt color.

Conclusions:

• The ANN model presents a highly effective and accurate method for real-time Saybolt color prediction.
• This automated approach offers a significant improvement over conventional laboratory-based techniques.
• The findings highlight the potential of AI in optimizing petroleum quality control processes.