Hybrid model diesel fuel degradation
Description
This data set contains MATLAB code to model experimental data on DF degradation dynamics in soil microcosm from different batch experiments under different C/N, moisture content. The experimental profile is interpolated, and the instantaneous degradation rate is calculated. An ANN model for rates is developed from the data set. This ANN model is incorporated into the dynamic model ODE to predict the degradation profile and also to predict the optimal C/N ratio and moisture content for degradation. Additionally, the ANN model explanation is presented in LIME and shap plot
Files
Steps to reproduce
b1_interpolation_Rates takes the experimental data and excludes data points of anomaly (not essential) and, for each data set, calculates 50-500 interpolated values. Those interpolated values are used to calculate the rates. b2_ninterpolation_Rates_PLOTS generates the plot, a swarmplot of estimated rates for different interpolation frequencies, and a DF degradation plot (interpolated) and an experimental one. b3_rates_to_ANN_revision WATE develops an ANN model for each scenario (rate of interpolation) considering different two-layer architectures. c2_Model_prediction_with_MLP is the hybrid model prediction adopting the developed ANN model. c3__LIME_Shaply is the generator for LIME and Shapely analysis datasets, which are used to generate plots using c4_LIME_Shaply_PLOTS. c5_amultiobj_opt is used to predict the process condition optimizing DF degradation.
Institutions
- Karunya University