PSI - Issue 79

Osman Bayrak et al. / Procedia Structural Integrity 79 (2026) 413–420

414

1. Introduction Ceramics play an essential role in advanced engineering applications owing to their unique physical and chemical properties, which are often unattainable with metals or polymers (Randhawa, 2024; Wang et al., 2025). Despite these advantages, their intrinsic brittleness remains a critical limitation that restricts broader structural use. One widely adopted strategy to overcome this drawback is the incorporation of nanofillers, which can improve toughness and reliability (Dlouhy et al., 2019; Mustafa et al., 2021). Among various ceramic materials, silicon nitride (Si ₃ N ₄ ) has been extensively investigated as a matrix material in ceramic composites (Ayas et al., 2008; Klemm, 2010; Krstic & Krstic, 2012). This material is particularly attractive due to its exceptional high-temperature capability, superior resistance to thermal shock, corrosion and wear, chemical stability, and excellent mechanical strength (Heimann, 2023). Graphene has emerged as a promising filler for ceramics (Mustafa et al., 2024) as well as polymers (Sánchez et al., 2023) and metals (Borand et al., 2024). Its superior mechanical (Papageorgiou et al., 2017) and electrical properties (Castro Neto et al., 2009), combined with an extremely large surface area-to-volume ratio (Zhu et al., 2010), make it an excellent candidate for ceramic-matrix composites. In recent years, studies showed that the addition of graphene nanoplatelets (GNPs or multilayer graphene) into ceramics can enhance properties such as electrical conductivity and fracture toughness (Huang & Wan, 2020). While toughness improvements were achieved [16], some other mechanical properties of Si ₃ N ₄ ceramics—such as Young’s modulus, bending strength, and hardness—were reported to degrade upon GNP incorporation [17–19]. Results remain inconsistent across literature, emphasizing the need to better understand the underlying toughening mechanisms. Recent studies on graphene-reinforced Si ₃ N ₄ nanocomposites have demonstrated that incorporating GNP can activate multiple toughening mechanisms, such as crack deflection, crack branching, crack bridging, crack pull-out (Bódis et al., 2019; Cygan et al., 2016; Dusza et al., 2012; Kvetková et al., 2012; Ramírez et al., 2018; Zhang et al., 2018), and interface separation (also called “interface debonding”) (Ramírez et al., 2021), which significantly improve fracture resistance. Therefore, having the knowledge of those mechanisms in the developed nanocomposites is of critical importance. Extensive experimental investigations, including microstructural analyses and mechanical testing, have provided valuable insights into these mechanisms and their dependence on processing conditions (Ramirez et al., 2014; Ramírez et al., 2018; Sedlák et al., 2016). Alongside experimental approaches, a limited number of numerical studies (Chen et al., 2021, 2022) modelled mechanical behavior of the nanocomposites. While they provided an important insight into the micromechanics of the nanocomposites, there is still a need for numerical models that could simulate major toughening mechanisms simultaneously. Factors such as the orientation distribution of GNPs and the presence of interfacial voids are often overlooked in numerical models, despite their critical influence on mechanical behavior of composites (Bayrak et al., 2025; Bódis et al., 2019; Tapasztó et al., 2016; Zhang et al., 2018). Therefore, there is a need for microstructure-informed finite element (FE) models that can realistically reproduce major toughening mechanisms and strengthen the link between experimental observations and predictive simulations. This study aims to cover this gap by microstructure-informed FE models. In this study, two-dimensional finite element models were developed based on the microstructural data given in the literature. The data included orientation distribution, average length and average thickness of GNPs. Dispersion and dimensions of interfacial voids were applied with a quasi-random manner in the models through a loading unloading process; there is no known statistical data in this regard in literature so far. Following the induction of the porosities, tensile and bending test simulations were performed on the models. Post-processing of the analyses showed that the