Hydrogen safety systems, evidence-qualified scientific models, and reproducible research
I develop scientific software at the intersection of chemical engineering, process safety, and computational research. My work spans operator-facing hydrogen-safety prototypes, document-grounded knowledge systems, atmospheric-dispersion models, source-term calculations, fire analysis, and atomistic materials research.
Across these projects, physical assumptions, evidence, validation status, and limitations are treated as part of the software—not as afterthoughts.
| System | What it does | Status and scope |
|---|---|---|
| hydrogen-station-sim | Physics-based hydrogen-refuelling-station digital twin with dynamic equipment models, 82 HAZOP sensor channels, faults, safety logic, consequence analysis, and 3D monitoring | Modelling, monitoring, and training prototype; not a certified controller |
| SAGA-PY | Python/FastAPI safety-knowledge assistant for Korean gas standards, law, and HAZOP analysis | Supports standalone document-grounded chat and dedicated digital-twin APIs; requires a lawful local corpus and provider credentials |
| SAGA-STS | Earlier Java/Spring Boot edition of the SAGA gas-safety assistant | Preserved separately for reproducibility and further development; different runtime and APIs from SAGA-PY |
The digital twin and SAGA-PY are independent services that can work together: the simulator supplies calculated process and sensor context, while SAGA-PY provides an advisory knowledge layer. Neither replaces current regulations, site procedures, or qualified engineering judgment.
| Project | Purpose | Evidence boundary |
|---|---|---|
| slabx | Python reimplementation of the SLAB dense-gas dispersion model | Compared with the original Fortran and evaluated against 38 field and wind-tunnel trials |
| slabx-lh2 | Liquid-hydrogen extensions and an applicability diagnostic for SLABx | Seven-test, 210-record FFI research configuration with explicit routing limits |
| degadisx | Python reimplementation of EPA DEGADIS 2.1 | Retains the legacy thermodynamic and numerical route; deliberately excludes LH₂ research extensions |
| degali | Modern DEGADIS-based research model for cryogenic-hydrogen releases | Regression-tested compatibility path and an alpha-stage LH₂ extension with documented validation limits |
| lh2poolx | Evidence-qualified, quasi-steady LH₂ pool source term for a declared time window | Makes the release-to-ground and footprint assumptions explicit; not a dispersion or impact model |
| DynamicLH2PoolX | Time-dependent LH₂ pool spreading, evaporation, inventory, and mass-balance model | Restricted Stage C component scope for declared ground inflow on horizontal surfaces |
| Project | Purpose | Reproducibility focus |
|---|---|---|
| pifira | Pressure-based inverse fire heat-flux and effective-area analysis, plus evidence-gated LH₂ thermodynamic utilities | Traceable LPG tank reconstruction and domain-checked LH₂ calculations |
| hydrogen-local-energetics | Data and code for local hydrogen energetics and lattice-controlled mobility | Curated DFT records, archived ML-potential results, manifests, and figure-generation code |
| qsafety-qae | Quantum amplitude-estimation research for safety and reliability probability models | Separates the probability model, classical baseline, circuit validation, and opt-in QPU execution |
standards · law · HAZOP evidence
│
▼
SAGA-PY ◄────────► hydrogen-station-sim
advisory layer dynamic process + sensors
declared LH₂ release
│
├── quasi-steady source ──► LH2PoolX
└── dynamic ground pool ──► DynamicLH2PoolX
resolved source state
│
├── SLAB family ──────────► slabx ──► slabx-lh2
└── DEGADIS family ───────► degadisx / degali
fire exposure ─────────────────► pifira
material-scale behaviour ──────► hydrogen-local-energetics
risk-estimation methods ───────► qsafety-qae
The arrows describe intended information flow and research relationships, not universal model interchangeability or certification. Each repository documents its own coupling rules, validation domain, and conditions under which a result must be withheld.
physical model → implementation → verification → validation
↑ │
└──────── limitations and reproducibility ◄─────────┘
I aim to make every important numerical result traceable: what assumptions produced it, where the model is valid, how it was checked, and how another researcher can reproduce it.
Python · FastAPI · NumPy · SciPy · pytest · CoolProp · Qiskit · numerical modelling · scientific validation
The repositories are open for inspection and reuse under their stated licences. For research use, consult each project's CITATION.cff, documentation, validation record, data policy, and limitations before applying its results. Safety-related outputs are research or decision-support artefacts unless a repository explicitly states otherwise.