AI Based Model for Prediction of Heavy Metals Using Physio-Chemical Characterization of Agricultural Waste Ashes

Abstract

The escalating volume of waste materials generated as byproducts is a growing concern in the context of recycling. These waste materials encompass a variety of heavy metals (HMs) that pose significant environmental hazards to plants, animals, and ecosystems. To address that HMs, there was a need to develop an artificial intelligence-based model capable of predicting the presence and quantity of HMs based on the chemical composition of the discards as AWAs. This study delved into a comprehensive analysis of the diverse origins of AWAs, exploring their multifaceted characteristics across different sources. In this research, a total of thirty-two types of SCBA and RHA were accumulated from various sources. The properties and attributes of residual ashes were assessed utilizing various methods of analysis, including X-ray fluorescence (XRF), Fourier-Transform Infrared Spectroscopy (FTIR), Scanning Electron Microscope (SEM), Energy dispersive X-Ray (EDX), X-ray Diffraction Analysis (XRD), Thermogravimetric Analysis / Differential Scanning calorimetry (TGA/DSC), and Atomic Absorption Spectroscopy (AAS). The results were presented in the light of existing literature and standards. The results accordingly revealed that AWAs can be categorized in three fractions based on loss on ignition. At the end some, recommendations for the utilization of SCBA and RHA based on the characterization results were also made for utilization as supplementary material in construction industry. Moreover, the machine learning model was constructed using input variables such as the physio-chemical properties of SCBA and RHA, element properties, and total HMs concentrations to predict the HM fractions. The application of machine learning tool to procured SCBA and RHA revealed that the model utilizing deep neural networks demonstrated performance robustly, possessing strong generalization capabilities (R2 = 0.99 on the testing set), enabling the rapid and accurate prediction of HMs fractions. The element properties were found to be the primary determinant of the HMs fractions. This study adds value to the creation of sustainable approaches for managing waste and provides a framework for the characterization of waste ashes for potential utilize as a primary substance in construction materials.