Supporting data, code and simulation materials for Simulation-Based Optimisation of Reflux and Side-Draw Flows in an Industrial Benzene Column

Supporting research materials for "Metaheuristic-Based Optimization of Benzene Distillation for Energy Savings and Improved Operational Performance: A Case Study of Column 200-C-54 at Skikda Refinery, Algeria" by Khalida Abdelali, Riad BENDIB and EL-arkam MECHHOUD. Manuscript-title update: The manuscript was previously titled "Simulation-Based Optimisation of Reflux and Side-Draw Flows in an Industrial Benzene Column". The record title and archived documentation retain that earlier title. This description links the same deposited research materials to the current manuscript title given above; the deposited files, version 1.0.0 and DOI are unchanged. The study couples an Aspen Plus RadFrac model of benzene column 200-C-54 with MATLAB particle swarm optimisation (PSO) and the Grey Wolf Optimiser (GWO). Reflux and liquid side-draw mass flows are varied subject to benzene mass-fraction constraints SIDE >= 0.9998 and BOTTOM <= 0.02. Fixed-feed experiments, feed-tray screening and expanded operating-domain searches are documented separately. The uploaded archive contains 110 unique, byte-preserved payloads from 123 supplied research attachments, with a source/duplicate mapping, plus the supplementary verification source, labelled derived population CSVs, documentation and a read-only Python checking utility. Contents include the original Aspen model backup/companions, MATLAB evaluators and drivers, result CSV/MAT files, candidate/attempt/history records, final-point verification, the six new author-executed fixed-coordinate checks for historical PSO labels 202 and 303, and fixed-feed convergence figure exports. Historical/incomplete records are isolated and are not additional completed stochastic runs. Supplied PilotGrid data are identical aliases of the documented 66-point FineGrid; no separate 25-point pilot data are claimed. Read README.md, ENVIRONMENT.md, RUN_GUIDE.md and PROVENANCE.md in the archive before reuse. Run validation/validate_package.py for read-only hash and selected numeric consistency checks; it does not execute Aspen or MATLAB. The new supplementary verification source was drafted with ChatGPT assistance and executed by the authors; supplied duty and composition outputs are author-executed records, not AI-generated simulator results. Deposit documentation and checking utilities were also prepared with AI assistance. Limitations: Windows and licensed Aspen Plus/MATLAB are needed for COM-based simulations; particleswarm requires Global Optimization Toolbox. Available supplementary/tray-33 metadata identify MATLAB R2022b, but the exact historical Aspen Plus release and full software environment remain unconfirmed. Evaluator success is not independent certification of official RadFrac convergence, units, residuals or material/energy-balance closure. Assembly checks did not rerun the simulators. Reported reductions are model predictions, not demonstrated plant savings, global-optimality proofs or robust industrial operating recommendations. Raw reconciled plant historian data and separate plant-document sources are not supplied. Release 1.0.0. Licensing is path-specific: CC BY 4.0 applies to author-owned research data, documentation, figures and model inputs/configuration; MIT applies to author-owned source code under code/ and validation/. See LICENSE.md and the complete texts in LICENSES/ inside the archive for scope, attribution and third-party exclusions. Original payload bytes are preserved. Neither licence grants rights to commercial Aspen Plus or MATLAB software, vendor libraries or other third-party materials. The authors approved this public release.