Table 1.
Summary of Heterogeneous Catalysts Used in Biodiesel Production Along with Influential Process Parameter.
Fig 1.
The schematic representation of steps involved in biodiesel production.
Fig 2.
Overview of ML work flow a) ML models b) steps involve in work flow.
Table 2.
Hyper Parameters for all four ML learning models.
Table 3.
Experimental Design Matrix Based on L27 Orthogonal Array for Process Parameter Optimization.
Fig 3.
Visualization of Experimental data collection a) Reaction Temperature, Methanol to molar ratio verus biodiesel yield b) Reaction Temperature, Catalyst concentration versus biodiesel yield.
Fig 4.
Basic concept of five k-fold cross-validation.
Fig 5.
Comparison of experimental output verus prediction output by all four learning models a) LR, b) PR, c) DTs, d) RF.
Fig 6.
Validation metric for all four learning models a) R2, b) RMSE, c) MSE, d) MAE.
Fig 7.
Concept of bias-variance tradeoff for all learning model.
Table 4.
K-fold cross validation on average R2 results.
Fig 8.
model interpretation a) SHAP b) Feature Importance c) Partial dependence.
Fig 9.
Model interpretation on Heatmap.
Fig 10.
Optimum process parameters by best model a) Reaction Temperature (RT) b) Methanol to oil molar ratio (MOR), c) catalyst concentration (CC).
Fig 11.
3D illustration of optimum process parameters on maximum yield a) Reaction temperature (RT) versus Methanol to oil molar ratio (MOR), b) Reaction temperature (RT) versus catalyst concentration (CC), c) Methanol to oil molar ratio (MOR) versus catalyst concentration (CC).
Table 5.
Optimum process parameters for maximum biodiesel yield.