AUTHOR=Natrayan L. , Janardhan Gorti , Paramasivam Prabhu , Dhanasekaran Seshathiri TITLE=Enhancing mechanical performance of TiO2 filler with Kevlar/epoxy-based hybrid composites in a cryogenic environment: a statistical optimization study using RSM and ANN methods JOURNAL=Frontiers in Materials VOLUME=10 YEAR=2023 URL=https://www.frontiersin.org/journals/materials/articles/10.3389/fmats.2023.1267514 DOI=10.3389/fmats.2023.1267514 ISSN=2296-8016 ABSTRACT=

This research aims to investigate the mechanical performance of the different weight proportions of nano-TiO2 combined with Kevlar fiber-based hybrid composites under cryogenic conditions. The following parameters were thus considered: (i) Kevlar fiber mat type (100 and 200 gsm); (ii) weight proportions of TiO2 nanofiller (2 and 6 wt%); and (iii) cryogenic processing time (10–30 min at −196°C). The composites were fabricated through compression molding techniques. After fabrication, the mechanical characteristics of the prepared nanocomposites—such as tensile, bending, and impact properties—were evaluated. The optimal mechanical strength of nanofiller-based composites was analyzed using response surface methodology (RSM) and artificial neural networks (ANNs). Compositions, such as four weight percentages of nano-TiO2 filler, 200 gsm of the Kevlar fiber mat, and 20 min of cryogenic treatment, were shown to produce the maximum mechanical strength (65.47 MPa of tensile, 97.34 MPa of flexural, and 52.82 J/m2 of impact). This is because residual strains are produced at low temperatures (cryogenic treatment) due to unstable matrices and fiber contraction. This interfacial stress helps maintain a relationship between the reinforcement and resin and improves adhesion, leading to improved results. Based on statistical evaluation, the ratio of correlation (R2), mean square deviation, and average error function of the experimental and validation data sets of the experimental models were analyzed. The ANN displays 0.9864 values for impact, 0.9842 for flexural, and 0.9764 for tensile. ANN and RSM models were used to forecast the mechanical efficiency of the suggested nanocomposites with up to 95% reliability.