Physics-informed neural network application in poly (vinyl alcohol) and orange peel powder based biocomposite film optimization for enhanced optoelectronic performance

This study introduces a Physics-Informed Neural Network (PINN) model to optimize biofilm fabrication using a solution casting method with poly (vinyl alcohol) (PVA) and biowaste orange peel powder (OPP) as constituents. Key process parameters include the PVA matrix temperature (80°C–98°C), magnetic bed spin rate (1000–1900 RPM), PVA-OPP mixture temperature (27°C–36°C), and proportions of PVA (50%–95%) and OPP (5%–50%). For initial input metrics, the films were fabricated with varying weight percentages of orange peel powder and PVA. The films were evaluated for surface morphology using Scanning Electron Microscopy (SEM) and for surface roughness through Atomic Force Microscopy (AFM) and the probe roughness test, which serve as primary input metrics alongside thermal and tensile properties. The PINN model identifies optimal settings, with 12.5% OPP achieving superior UV absorption (6.2% transmittance in the 200–400 nm range) through enhancing surface smoothness. The model also predicts an improved thermal stability of 355.08°C and a tensile strength of 8.67 MPa. Experimental validation at optimal conditions, achieved through PINN (84°C, 1150 RPM, 28.5°C, 87.5% PVA, and 12.5% OPP), demonstrates a 99.8% correlation with PINN predictions. The optimized biofilm composition offers better UV absorption, thermal withstanding capacity, tensile properties, and balanced environmental stability through moderate moisture absorption as well as biodegradation, making it highly suitable for optoelectronic applications.