A Coupled Approach of Spatial Dielectric Mapping and Unsupervised Machine Learning for Damage Detection and Severity Segmentation in Polymer Matrix Composites
Rishabh D. Guha,Katherine Berkowitz,Landon R. Grace
DOI: https://doi.org/10.1016/j.ndteint.2024.103137
IF: 4.683
2024-05-05
NDT & E International
Abstract:Historically speaking, significant functional improvements in the engineering sectors has often been accompanied by breakthroughs at the frontier of materials discovery. In the aerospace and automobile industry, the development of fiber reinforced composites in the mid-twentieth century was one such event which led to rapid advances in terms of performance. When compared to conventional structural materials like steel, composites offer a higher strength to weight ratio [1], [2], resistance to chemical contamination [3] and a higher thermal stability [4], [5]. Composites also gained a lot of favor in the defense and military sectors when researchers discovered its ability to transmit radar signatures without considerable signal absorption [6]. The advantages this class of materials offer has led to a global composite market which was estimated at 86.4 billion USD in 2020. The industry is expected to expand at a compound annual growth rate (CAGR) of 6.6% from 2021 to 2028 [7]. Despite their obvious superiorities over existing material classes, it cannot be overlooked that the majority of applications where composites are employed are safety critical [8] where the failure of any constituent can trigger a catastrophic event without warning. Accurate damage detection and quantification in composites has been an active area of research for the past couple of decades [9], [10], [11] and the problem is complicated because, unlike metals, composites often do not show any visible signs of damage initiation or accumulation [12], [13]. The multitude of applied loads and stresses from different sources during the operating lifetime makes it even more challenging to develop tools for residual life prediction [14], [15]. Damage can initiate in a composite laminate from a seemingly benign event like a hail strike or a tool drop by an operator during a routine maintenance overhaul [16]. The barely visible matrix cracks generated will accumulate in-service as the structure is subjected to fluctuating cycles of temperature [17], [18], freeze-thaw cycling [19], [20], and fatigue loading [21], [22], [23]. Moreover, the polar nature of the polymer matrix of a composite coupled with exposure to low levels of precipitation or prolonged high humidity makes them amenable to low levels of moisture absorption [24], [25]. Neat epoxy-based resins can absorb anywhere between 1-7% moisture by wt. [26] which can translate to a moisture content of ∼3% by wt. [27] in a polymer composite. Earlier studies have provided a detailed documentation of the deleterious effect of absorbed moisture on the mechanical performance [28], [29], [30], [31], [32], thermal stability [33] and electrical insulation properties [34], [35], [36] of polymer matrix composites (PMC). The combination of performance unpredictability, mechanical loads and environmental stressors makes it imperative to develop safe, reliable, and affordable methods of non-destructive evaluation (NDE) geared specifically for composites. Currently, there are a variety of available NDE techniques which aim to detect damage at different phases as it evolves in a composite [37], [38], [39]. The strengths and weaknesses of each technique have been discussed in previous review papers [40], [41], [42], [43] but a persistent limitation across techniques is their inability to reliably predict sub-micron scale damage during the early lifetime of the composite. Some techniques like radiography [44] which perform well in the damage initiation phase are burdened by safety and economic restrictions. Consequently, despite the development of highly sophisticated NDE techniques at the academic level, a widespread adoption on the commercial scale is lacking and manual techniques that rely on operator skill like 'tap tests' [45] are still prevalent. We have briefly discussed the known degradation induced by moisture in composites, but some degree of absorption is inevitable even in low-humidity operating environments [46]. Molecular simulation studies [47], [48], [49], [50], [51], [52], [53] have demonstrated that the state of the ingressed water molecules is dependent on the intermolecular secondary bonding interactions with the polymer network and the physical morphology of the matrix. The presence of nanopores and voids can lead to local clusters of water molecules which do not engage in strong hydrogen bonds with the polar groups in the network. Spectroscopy [54], [55], [56], [57] and dielectric relaxation [58], [59] have been successful in identifying two distinct states of moisture in an epoxy – one which is clustered as water droplets ("free") and the other which is "bound" to the polymer matrix. The presence of internal damage sites creates additional free volume in the form of microcracks where moisture can exist as "free" water and exhibit characteristics similar to bulk water [60], [61], [62], [63], [64], [65]. In the case of dry non-condu -Abstract Truncated-
materials science, characterization & testing