Green composites based on thermoplastic starches and various natural plant fibers: Impacting parameters of the mechanical properties using machine-learning

Multivariate analyses on formulation and mechanical behavior of nonwoven and nonoriented natural fibers reinforced thermoplastic starch (TPS) composites were performed. Glycerol and water were considered as TPS plasticizers. Fibers composition (i.e., cellulose, hemicellulose, lignin), fibers morphology (fibers length), starch composition (i.e., amylose/amylopectin ratio) as well as the processing conditions (i.e., temperature, rotor speed, relative humidity during aging) were evaluated for their ability to affect the elastic modulus, tensile strength, and elongation at break of the final materials. Multivariate linear regressions were computed to unveil the importance of each variable on the mechanical behavior. Fibers composition impacted the most the models: cellulose maximization improved the elastic modulus and tensile strength while lignin reduced the elastic modulus and hemicellulose decreased the tensile strength. TPS plasticizers, temperature, and rotor speed of the process were negatively impacting the elastic modulus but in a lesser extent than the fiber composition. Within the range of the created database, the selected variables and attributed coefficients were permitted to explain the variability. The produced models revealed that complex and yet uninvestigated interactions are to be considered within TPS-based biocomposites. Therefore, this work discusses and suggests a “must-have” list of variables for comparable analyses of new TPS-based biocomposites using natural fibers as reinforcement.