Uncertainties and process-induced deformations and their influence on lifetime in thermosetting composite parts

Thermosetting composite materials are extensively utilized in the aerospace industry because of their exceptional stiffness and outstanding strength-to-weight ratio. Over the past few decades, research has demonstrated that the selection and optimization of composite materials (resin and fiber), mold characteristics, and other factors require extensive iterative analysis. Despite significant advancements in virtual manufacturing frameworks, considerable uncertainty still surrounds the processing and testing of aerospace components. The thesis investigates how inherent variability in material properties and processing parameters contributes to the development of process-induced deformations (PID) and residual stress fields in composite structures post-manufacturing. The work further examines the implications of these phenomena on structural integrity during mechanical testing. The research specifically investigates these phenomena at the coupon and element levels, focusing on load-bearing components manufactured from Hexcel 8552/AS4 prepreg materials. The initial phase of the study is dedicated to enhancing cure simulation tools. This involves comprehensive characterization of key material states and properties, manufacturing phenomena, and processing variables that significantly influence the development of PID and residual stresses. Characterizing the material state, however, remains challenging due to its strong dependence on variables governed by parametric models that describe the curing evolution of thermoset resins. These models are a critical component of simulation tools, offering insights into the development of mechanical properties in composite parts during manufacturing. To address this uncertainty from a passive manufacturing control perspective, a Finite Element-based process simulation tool is developed. This tool integrates a constitutive viscoelastic model with a non-parametric neural network trained on experimental characterization data, effectively replacing conventional parametric cure models. The predictive capability of the enhanced simulation tool is validated across several case studies by comparing PID, specifically spring-in angles, and demonstrates improved accuracy. While passive manufacturing control approaches offer valuable insights, active manufacturing control is essential for further enhancing process reliability and performance. Building upon the theme of cure state uncertainty, a novel Monte Carlo sampling scheme based on Particle Filters is developed to enable accelerated predictions of PID under a wide range of thermal loading conditions. It is observed that uncertainty associated with the cure state significantly diminishes after the gelation point, as evidenced by the convergence of PID predictions. The Particle Filter framework is further leveraged to conduct parametric exploration aimed at optimizing extrinsic parameters beyond the gelation point. Subsequently, an innovative methodology is proposed that couples the non-parametric neural network model with the Particle Filter approach, further improving the predictive robustness and control of the manufacturing process. Lastly, extending the scope of numerical modeling within the virtual manufacturing framework, this research addresses the challenge of predicting composite failure during testing, particularly when manufacturing defects such as voids and residual stresses coexist at the lamina level. To this end, a comprehensive 3D mixed-mode phase field framework is developed to simulate the progressive failure mechanisms of delamination and matrix cracking, while incorporating the influence of internal residual stresses induced during processing. The numerical approach to modeling progressive damage provides an early assessment of the impact of manufacturing defects and enables the optimization of cure cycle profiles, thereby reducing risk and enhancing the overall quality of composite manufacturing. Overall, the thesis underscores the critical role of numerical methods in advancing performance and reliability, providing a foundation for more efficient optimized manufacturing practices.

Not available. The document is only provided in English.