AI-Optimized Tensile Strength Prediction of Biocomposites Using Agricultural Waste

The increasing demand for green and environmentally friendly composite materials have triggered a marked interest in the utilization of agricultural waste as reinforcement in polymer matrices. This paper shares a fresh and unique way of predicting the tensile strength of bio-composites of Rice Husk, Groundnut Shell, and Santa Maria Feverfew when they are dipped in epoxy resin and hardener. We propose a new machine learning model that combines XGBoost and Long Short-Term Memory (LSTM) networks as a new way to reduce the error for the prediction of the tensile strength of a given material. Data were extracted from the process of making composite samples in a laboratory, and then the samples have been tested for their tensile strengths. The XGBoost model can identify non-linear connections, and decrease the number of characteristics, while the LSTM model uses the information from the XGBoost predictions to make the output more accurate. The metrics that are used in the evaluation like MSE, MAE, and R² Score are the evidence that the proposed approach is working. The residual laboratory results, and information extracted from the charts about the most important features, are also the types of visual techniques that can be used as evidence of the models' robustness. This AI-fueled system is a very useful device for designing composites more efficiently and cutting the costs of the experiment, while at the same time it gets the limited development of the green materials and sustainable engineering fields speeded up.