Research Advances in Machine Learning Techniques in Gas Hydrate Applications.

The complex modeling accuracy of gas hydrate models has been recently improved owing to the existence of data for machine learning tools. In this review, we discuss most of the machine learning tools used in various hydrate-related areas such as phase behavior predictions, hydrate kinetics, CO2 capture, and gas hydrate natural distribution and saturation. The performance comparison between machine learning and conventional gas hydrate models is also discussed in detail. This review shows that machine learning methods have improved hydrate phase property predictions and could be adopted in current and new gas hydrate simulation software for better and more accurate results.

Coconut oil and fermented palm wine biodiesel production for oil spill cleanup: experimental, numerical, and hybrid metaheuristic modeling approaches.

This paper for the first time synthesizes novel biodiesel experimentally using low-cost feedstocks of coconut oil, caustic soda, and fermented palm wine contaminated by microorganisms. The alkaline catalyzed transesterification method was used for biodiesel production with minimal glycerol. The produced biodiesel was biodegradable and effective in cleaning a shoreline oil spill experiment verified by our developed oil spill radial numerical simulator. For the first time, an adaptive neuro-fuzzy inference system (ANFIS) was hybridized with invasive weed optimization (IWO), imperialist competitive algorithm (ICA), and shuffled complex evolution (SCE-UA) to predict biodiesel yield (BY) using obtained Monte Carlo simulation datasets from the biodiesel experimental seed data. The test results indicated ANFIS-IWO (MSE = 0.0628) as the best model and also when compared to the benchmarked ANFIS genetic algorithm (MSE = 0.0639). Additionally, ANFIS-IWO (RMSE = 0.54705) was tested on another coconut biodiesel data in the literature and it outperformed both response surface methodology (RMSE = 0.72739) and artificial neural network (RMSE = 0.68615) models used. The hybridized models proved to be robust for biodiesel yield modeling in addition to the produced biodiesel serving as an environmentally acceptable and cost-effective alternative for shoreline bioremediation.