ANN-Kriging hybrid model for predicting carbon and inorganic phosphorus recovery in hydrothermal carbonization

Modeling of hydrothermal carbonization (HTC) of poultry litter to high-value materials was conducted in order to understand the process and predict the influence of process parameters on product properties. Reaction temperature and time were considered as inputs, whereas carbon and inorganic phosphorous recovery were considered as responses in the model. Artificial neural network (ANN) model was used in order to correlate the process parameters to the outputs. The model was trained and validated using the data collected from HTC experiments carried out at temperatures between 150 ≤ T ≤ 300 °C, and residence time between 30 ≤ t ≤ 480 min. In order to improve the predictability of ANN, more theoretical data points were generated using Kriging approach based on the available measured data. Kriging interpolation improved the ANN model dramatically in training and validation phases, where the carbon recovery model fitting was improved by 0.94% and 9.2% in training and validation respectively, and the inorganic phosphorous (IP) recovery model fitting was improved by a staggering 16.4% and 19.6% in training and validation respectively. This improvement is also reflecting on the derived profiles of carbon and IP recovery in terms of the process parameters. The validated model was then used to understand the effect of process parameters on the response. It was revealed that temperature has more significant effect on the carbon and phosphorous recovery, while the effect of reaction time is more important at low reaction temperatures. The derived profiles shows a monotonic increase in IP recovery and a monotonic decrease in Carbon recovery at higher temperatures and time, this is due to multiple mechanism occurring simultaneously in the HTC reactor at various temperatures and times.