Failure as a design feature: A computational framework for fracture-driven optimization

Abstract

Traditional structural design focuses on preventing fracture, limiting applicability in systems that require controlled, predictable failure. This dissertation developed a dual-physics optimization framework that couples finite element analysis (FEA) with peridynamic fracture simulation to design structures where engineered failure is a primary objective.The framework linked DOLFINx-based transient FEA for impact stress evaluation with Peridigm peridynamic fracture modeling through a fracture-aware thickness cap that pre vents stress-driven thickening from compromising fracture performance. This cap broke the feedback cycle in which thickness increases suppress fracture damage, enabling the stress and fracture objectives to be resolved within a single automated loop. The parameterized thin shell structure, with independent thickness and fracture zone definitions per sectoral region, enabled automated mesh generation, multi-orientation drop simulation, and crackline connectivity evaluation within a closed optimization loop. Eight optimization methodswere compared across band-count and initial-thickness sensitivity studies.The breakaway weight shell for Hawaiian pelagic longline fisheries served as the validation application. Flyback events, in which a failed fishing line accelerates the weighted swivel at high speed, pose serious injury risks to crew; the breakaway shell must survive normal handling while fracturing reliably upon flyback impact. Prior full-scale flyback testing and tensile characterization of monofilament line established the loading environment, predicting a maximum flyback velocity of 234 m/s for the standard 45 g swivel under 1000 N line tension and demonstrating an 86% impulse reduction and 99% impact energy reduction with the break-away prototype. These experimental results provided the boundary conditions (impact velocity, projectile mass, shell material, and fracture geometry) that parameterized the computational framework. Peridynamic simulations verified that fracture propagated along designed failure planes.Energy conservation in the transient FEA solver was verified to within 1.6% over multiple bounce cycles. The optimization achieved stress feasibility at the five-band parameterization (worst-case stress ratio 0.98, total shell thickness 9.02 mm) while maintaining fracture performance, and all seven methods converged to qualitatively similar pole-thickened thickness profiles. FSD-G achieved stress feasibility at the ten-band parameterization (σ/σt = 0.96) with the lowest material volume (0.90×baseline), while PGD was the most consistent method across all configurations. The peridynamic fracture simulations confirmed that geometric selectivity alone, using a single uniform critical stretch across all regions, directed crack prop- agation along designed failure planes without requiring differentiated material properties. The fully automated framework demonstrated that structural optimization and fracture optimization can coexist within a single dual-physics loop, and that geometry alone is sufficient to control crack paths in frangible structural systems.