Optimizing process parameters to predict hydrothermal caffeine extraction from tea residues

Tea is one of the most important parts of agricultural production in China. A large amount of waste tea residues can be produced every year with the continuous growth of the tea industry. Waste tea residue shares significant economic and medicinal value, due to its bioactive substances, especially caffeine. However, most waste tea residues cannot be fully utilized in a resourceful way, resulting in the waste of resources and pressure on the environment. In this study, an efficient caffeine extraction was proposed using hydrothermal extraction. A systematic investigation was implemented to explore the effects of different factors on extraction efficiency. Response surface ology (RSM) and artificial neural network (ANN) were also combined to optimize and predict the extraction conditions. Firstly, the efficient extraction of caffeine from tea residue was achieved by hydrothermal extraction. The special properties of water were exploited under high temperature and pressure. As such, the lignocellulose structure in tea residues was effectively destroyed to obtain the dissolution of active substances, such as caffeine. The high efficiency and environmental advantages of hydrothermal approaches were emphasized to treat the wet biomass under mild conditions, compared with the conventional one. The better performance was achieved in the high efficiency, low energy consumption, and reduced use of hazardous chemical solvents. Secondly, the key parameters of extraction (hydrothermal temperature, time, liquid-solid ratio, and pH) were systematically analyzed using RSM. The influencing level of each factor on the caffeine extraction was determined to derive the optimal extraction conditions. The experimental results showed that the hydrothermal temperature shared the most significant effect on the caffeine extraction. The extraction efficiency was significantly improved to appropriately increase the hydrothermal temperature. An ANN model was also established to further validate and optimize the extraction conditions. A large amount of experimental data was trained and then validated to learn and simulate complex nonlinear processes. The ANN model was also used to accurately predict caffeine extraction under different conditions. The prediction of the ANN model showed that the optimal extraction conditions were determined as the hydrothermal temperature of 220 °C, hydrothermal time of 3.5 h, liquid-solid ratio of 50 mL/g, and pH 9.0. A caffeine extraction of 66.19 mg/g was expected to be obtained. There was high agreement with the actual experiment (65.81 ± 0.47 mg/g), indicating a high prediction accuracy. The ANN model was verified to optimize and predict the caffeine extraction from waste tea residue, in terms of the high feasibility and accuracy. In conclusion, the organic solvents were avoided to be environmentally friendly and realize green chemistry, indicating the promising potential for the extraction of natural products. Meanwhile, the efficient extraction of caffeine from waste tea residue can provide a useful reference for the high-value recycling of waste biomass resources using RSM and ANN models. The environmental pressure can be alleviated to reduce the waste of resources for the considerable economic benefits in the tea industries.