Building Ships That Work: A Computational Blueprint

Most naval architecture texts stop at the equations. This one starts there and keeps going — through meshing nightmares, turbulence model arguments, the pain of free-surface capture, and the discipline required to make a simulation credible enough for a classification society to stamp. Kathryn Dunn has written the reference she wished existed when she was building simulation pipelines: a single volume that treats the entire computational lifecycle as an engineering problem, not a menu of disconnected solvers.

What the book is about

Naval Architecture Computational Methods is a 25-chapter technical guide aimed at students, researchers, and practicing engineers who need to set up, run, and trust numerical simulations for vessel design. It moves chronologically through the design process: geometry representation (Chapter 2), mesh generation (Chapter 3), potential-flow panel methods (Chapter 4), boundary-element seakeeping (Chapter 5), RANS turbulence modeling (Chapter 6), free-surface capturing (Chapter 7), resistance and propulsion (Chapter 8), propulsors and cavitation (Chapter 9), maneuvering (Chapter 10), irregular-seas seakeeping and seaway loads (Chapter 11), hydrostatics and stability (Chapter 12), fluid-structure interaction (Chapter 13), finite-element structural analysis (Chapter 14), fatigue and vibration (Chapter 15), HPC solver selection (Chapter 16), verification/validation/uncertainty quantification (Chapter 17), experimental correlation (Chapter 18), optimization algorithms (Chapter 19), surrogate modeling (Chapter 20), multidisciplinary design optimization (Chapter 21), Python automation (Chapter 22), data management and reproducibility (Chapter 23), regulatory compliance (Chapter 24), and three end-to-end case studies (Chapter 25). The through-line is credibility: how to build simulations that are not only accurate but also efficient, traceable, and defensible.

Credibility as a design constraint

The book treats Verification, Validation, and Uncertainty Quantification (VVUQ) not as a post-processing checkbox but as a design constraint that shapes every upstream choice. Chapter 17 makes this explicit: "Verification asks, 'Are we solving the equations correctly?' Validation asks, 'Are we solving the correct equations?' And Uncertainty Quantification asks, 'How much confidence do we have in the answer, considering all the potential errors?'" The text then walks through mesh convergence studies using the Grid Convergence Index, time-step convergence for unsteady runs, and the dangers of "validation by tuning" — adjusting turbulence coefficients to match tank data without physical justification. Dunn insists that a simulation without quantified uncertainty is just a guess in colorful contours.

The mesh as a physics filter

Chapter 3 reframes meshing from a preprocessing chore into a physics filter. The author explains that "a well-constructed mesh is the silent hero of many a successful simulation" and then details exactly what that means for marine CFD: prismatic inflation layers targeting y+ ≈ 1 for low-Re models, wake refinement zones that stretch gradually aft, and a dedicated free-surface box with enough cells per wavelength to avoid numerical damping. The chapter doesn't just list metrics — aspect ratio, skewness, orthogonality — it connects each to specific failure modes: "Highly skewed cells can introduce interpolation errors and reduce solution accuracy" and "poor orthogonality, especially near boundaries, can degrade accuracy and slow convergence." The practical payoff is a checklist mindset: if the mesh doesn't resolve the boundary layer and the wave field simultaneously, the solver cannot save you.

Two-way coupling where it matters

Fluid-structure interaction gets a full chapter (13) and reappears in the slamming/whipping case study (Chapter 25). Dunn distinguishes one-way coupling — pressures from CFD fed to a static FEA — from the two-way partitioned loops needed for whipping and slamming: fluid pressures deform the structure, structural displacements feed back to the fluid mesh, and the loop iterates until interface forces and displacements are consistent within each time step. The book is honest about cost: "Full, direct, two-way coupled CFD-FEA simulations for an entire ship are currently beyond routine engineering practice." The recommended compromise is targeted FSI on critical zones (bow impact, propeller blade) calibrated by global hydroelastic BEM for whipping. This pragmatic tiering — high fidelity where physics demands it, reduced order elsewhere — runs through the entire book.

Automation as the enabler of exploration

Chapters 22 and 23 make the case that scripting and data governance are not IT overhead but the infrastructure that makes optimization and MDO possible. Python becomes the "central nervous system" connecting parametric CAD, automated meshing, solver APIs, HPC job submission, and post-processing into reproducible batch workflows. The text shows a concrete loop: read design variables from CSV, update geometry via CAD API, trigger mesher, launch solver, extract resistance and sinkage, store results, generate comparison plots — all unattended. Chapter 23 then extends this to configuration control: versioned inputs, containerized software environments, metadata-tagged outputs, and traceability from a sea-trial anomaly back to the exact CFD run that informed the original scantling. Without this discipline, the Pareto fronts in Chapter 21 are just pretty pictures.

Case studies that close the loop

The final chapter demonstrates the workflow on three realistic programs: a high-speed container feeder optimized for resistance and seakeeping via Kriging surrogates trained on RANS/VOF and BEM; a propeller redesign validated through unsteady sliding-mesh CFD, cavitation tunnel tests, and vibro-acoustic FEA; and an offshore patrol vessel where full two-way FSI on a bow section calibrates a faster time-domain BEM slamming model for long-term fatigue assessment. Each study shows the handoffs — geometry to mesh, CFD pressures to FEA loads, surrogate predictions to high-fidelity verification — and the VVUQ evidence required at each gate. The result is a template any team can adapt: define the mission, parameterize the form, automate the pipeline, quantify the uncertainty, and document the chain.

Who should read this

This book is written for the naval architect who already knows the governing equations but needs to turn them into a reliable, auditable design process. Graduate students will appreciate the worked examples and checklists; practicing engineers will recognize the meshing traps, turbulence-model debates, and Class-society documentation demands. Readers looking for a first introduction to CFD or FEA theory should start elsewhere — the text assumes fluency in the physics and focuses on the how of industrial application. For anyone building or managing a simulation-based design loop, it's the most complete single reference currently on the shelf.

Read “Naval Architecture Computational Methods” on MixCache.com →