Code & Frameworks.

Python tooling that powers the simulation studies — each framework links to its GitHub repository.

01 / framework

Aspen Automation Framework

Python automation pipeline for Aspen Plus — DOE generation, parallel simulation execution, surrogate model training, and optimization. Enabled a 1,575-case study completed in hours instead of weeks.

Problem

Aspen Plus is designed for interactive use. Running thousands of simulations for DOE-based dataset generation manually requires opening the simulation file, setting each input, running, recording results — for every case. At ~2–5 minutes per run and 1,000+ cases required for surrogate model training, manual operation is not feasible.

Solution

Full pipeline automation from simulation to optimization:

Stage	Module	Description
Simulation	`aspen_simulation_engine`	Python COM interface to Aspen Plus
DOE	`model_doe_generator` (GUI)	Grid / LHS design of experiments
Data collection	`simulation_worker`	Parallel simulation execution
Preprocessing	`preprocessing_engine`	Outlier removal, normalization
Surrogate training	`SurrogateTrainingPipeline`	RF, GB, NN, GP model comparison
Optimization	`BayesianOptimizer`, `MultiObjectiveOptimizer`	Bayesian + NSGA-II

Impact

1,575 simulations completed in hours (vs. weeks manually)
Eliminated human transcription errors in parameter setting and result recording
Reusable pipeline applicable to any Aspen Plus model with minimal reconfiguration
Powers the CCPP-LOHC optimization study end-to-end

Architecture Highlights

Document-level Reinit pattern eliminates convergence propagation contamination across runs
Modular separation: DOE / simulation / preprocessing / training / optimization
Configuration-driven scenarios — new study = new YAML config, no code change
Failure isolation: one simulation crash does not abort the run; failed cases are flagged and retried

Repository

→ github.com/LEE-Inseok/aspen-automation

PythonCOM APIDOESurrogate ModelingNSGA-IIBayesian Optimization

02 / framework

SBH Thermal Reactor Framework

ODE-based thermal simulation framework for reactor configurations — separates model definition, scenario specification, and execution. Powers the 5-scenario heat recovery comparison study.

Problem

The thermal simulation study required running the same energy balance equations across 5 reactor configurations × multiple parameter variations. Hardcoding each scenario separately leads to duplicated equation logic, manual parameter management, and difficulty adding new scenarios.

Solution

Three-layer separation:

Stage	Description
Model definition	Energy balance ODEs for reactor + heat recovery structures
Scenario configuration	Parameter sets for each system configuration (baseline, wrap coil, internal coil, embedded purification, hybrid)
Simulation execution	ODE integration across all scenarios using `scipy.integrate.solve_ivp`
Post-processing	Temperature profile comparison, heat recovery quantification, scenario ranking

Before / After

Before: Individual scripts per scenario, copy-pasted ODE logic, manual result collection.

After: Single model definition reused across all scenarios. New scenario = add one parameter dictionary. Results automatically collected and compared.

Impact

5 scenarios compared systematically with identical model equations
Parameter sensitivity analysis enabled without code modification
Fully reproducible — any scenario can be re-run with exact parameters logged

Repository

→ Part of the broader Aspen Automation framework.

PythonSciPyODEReactor Modeling