Assessment of textile sludge pyrolysis behaviour through advance predictive models for bioenergy production

Abstract

Managing the textile sludge has become a significant global challenge because it is often disposed of in landfills without recovering its energy potential. Pyrolysis has emerged as a promising technology to extract energy and valuable chemicals from textile sludge. This study investigates the pyrolysis of textile sludge through thermogravimetric analysis (TGA) conducted from ambient temperature to 1000 °C at three different heating rates of 2.5, 5 and 7.5 °C/min in an inert environment. The kinetic study complemented was isoconversional method which includes both differential and integral (Friedman, Ozawa-Flynn-Wall & Kissinger-Akahira-Sunose) method. Moreover, five pseudo-components (pc1, pc2, pc3, pc4, pc5) were obtained by multi distributed activation energy model (M-DAEM). Additionally, combined kinetics retrieves the Ea of 98.3(±3.6) kJ/mol with R2 value of 0.9924. Furthermore, pyrolysis of textile sludge was performed in fixed bed reactor (FBR) at 500 °C that results in a pyro-oil, non-condensable gases and the biochar yields 17.5(±1.7) %, 13.7 (±3) % and 68.8 (±1.3) % respectively. To optimize the pyrolysis of textile sludge, it is essential to comprehend the complex kinetics involved in its degradation. Machine learning models like Artificial Neural Networks (ANN), Classification and Regression Trees (C&RT), Boosted regression trees (BRT), and K Nearest Neighbors (KNN) employed to predict the activation energy (Ea) with ANN emerging as superior predicting capabilities (R2=0.999) over other models. This study demonstrates the remarkable ability of ANN, C&RT, BRT and KNN to accurately analyze complex relationships, predicts the textile sludge pyrolysis kinetics and confirms the potential of textile sludge pyrolysis as a sustainable and efficient energy source.