Prediction and validation of fire parameters for a self-extinguishing and smoke suppressant electrospun PVP-based multilayer material through machine learning models

Electrospinning is a technology largely employed to obtain polymer fibers with different functionalities. The electrospinning of polyvinylpyrrolidone (PVP) in the presence of silica nanoparticles, and the subsequent thermal treatment of these electrospun PVP-silica fibers, allows for the manufacturing of a self-extinguishing material stable in polar solvents. However, this material lacks consistency and does not sustain any load: this strongly limits its application in many industrial fields (e.g., the aerospace sector). Herein, we used cross-linked electrospun PVP-silica blankets and TiO2 nanoparticles to coat hemp blankets, producing a multilayer material (MM) by surface charge interaction. The MM exhibited lower stiffness than the original hemp fabric but still good mechanical behavior, V0 class at the UL 94 vertical burning test, and good stretchability even after direct flame exposure. Further, burn-through and cone calorimetry tests revealed that MM is an excellent smoke suppressant and fireproof fabric, with very low total smoke release values (as low as 4.9 vs. − 33.3 m2/m2 measured for hemp) and its structure remained intact for at least − 1 min. Finally, as all the aforementioned experimental activity, though necessary and unsubstantial, is usually quite time-consuming, two Machine Learning models were developed and exploited to predict the fire performances related to the multilayer material. Despite the incomplete starting datasets, the implemented models accounted for a successful prediction of the target parameters (namely, Time to Ignition and peak of Heat Release Rate), thanks to the assistance of ChatGPT and the exploitation of made-on-purpose decision trees.