Mathematical modelling for process optimization of polyurethane (PU) biodegradation using combined response surface methodology and machine learning approaches

Polyurethane (PU) is one of the most widely used thermoset polymers across various industries. Its recalcitrant nature has resulted in its long term environmental accumulation. In this study, the PU biodegradation potential of Fusarium parceramosum was studied using statistical optimization and predictive modelling approaches. A design of experiments (DOE) approach was employed to screen seven process variables, of which four influential variables were subsequently optimized. Initial screening identified pH, and agitation speed as the significant factors affecting degradation, with a maximum degradation of 24.92 % during screening to 96.22 % degradation during further central composite design (CCD) based optimization. The observed weight loss was associated with PU biofragmentation and structural deterioration, indicating polymer breakdown. The developed RSM model showed good agreement with experimental results (R 2 = 0.84). Machine learning (ML) models, including Single-layer feedforward neural network (SLFNN), Radial basis function neural network (RBFNN), Support vector regression (SVR) and Random forest also demonstrated good predictive performances (R 2 ≈ 0.82 – 0.83), for modelling PU biodegradation under the optimized conditions. The predicted biodegradation performance was further validated through physicochemical characterization using gravimetric weight loss, enzyme assay, Scanning electron microscopy (SEM), Fourier transform infrared spectroscopy (FTIR), X-ray Diffraction (XRD) and, Gas-chromatography-Mass spectrometry (GC-MS), which confirmed significant structural and chemical modifications in PU foam, indicating active enzymatic degradation. GC-MS analysis further tentatively identified urethane, hydrogen isocyanate, glycerin, and N-(3-methylphenyl)urea as degradation intermediates, providing the basis for proposing a putative PU biodegradation pathway inferred from the detected metabolites and available literature. Overall, the study highlights the potential of combining microbial degradation, statistical optimization, and machine learning approaches for efficient PU waste treatment, contributing to sustainable plastic waste management and aligning with sustainable development goals 12 (Responsible consumption and production) and 13 (Climate action).