Simulator-Guided Case Revision in a Chemical Process Network: Recovery and Evidence Limits
Systemic Analytics · 22.09.2026 · 4 (3) · 208-221
DOI: 10.31181/sa43202691 ↗Kurz gefasst
Fachartikel in der Zeitschrift Systemic Analytics, online erschienen am 22.09.2026. Verfasst von Sogol Motallebi und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Operational recovery requires choosing an intervention, assessing its likely outcome, and learning from the result. This study examines simulator-guided revision of retrieved cases in an eight-unit chemical-process surrogate. The reported campaign contains 600 balanced single-fault episodes and five methods. TWIN- Case-Based Reasoning (CBR) combines ridge-derived feature relevance, a severity rule, simulation of at most three candidate actions, and retention of corrected solutions. Its reported recovery rate is 74.8%, compared with 63.3% for the random forest, a difference of 11.5 percentage points. Three limits determine interpretation: the verification model is the episode-generating simulator; corrected test cases enter the case base for later test episodes; and component variants are reconstructed from logged internals. The action-counter description is also inconsistent with the reported mean charge and revision rate, requiring original implementation records. Thus the comparison describes an idealized, sequential simulation campaign, rather than field validation or ordinary frozen-model cross-validation. Recovery checks do not certify continuous safety, and explanation fidelity was not measured. The contribution is a bounded analysis of experience reuse, candidate revision, and the evidence needed for an operational decision-support study.
Zitieren
APA
Motallebi, S., & Ferreira, D. A. (2026). Simulator-Guided Case Revision in a Chemical Process Network: Recovery and Evidence Limits. Systemic Analytics, 4(3), 208-221. https://doi.org/10.31181/sa43202691
Vancouver
Motallebi S, Ferreira DA. Simulator-Guided Case Revision in a Chemical Process Network: Recovery and Evidence Limits. Systemic Analytics. 2026;4(3):208-221. doi:10.31181/sa43202691
BibTeX
@article{Motallebi2026,
author = {Motallebi, Sogol and Ferreira, Daniel A.},
title = {Simulator-Guided Case Revision in a Chemical Process Network: Recovery and Evidence Limits},
journal = {Systemic Analytics},
year = {2026},
volume = {4},
number = {3},
pages = {208--221},
doi = {10.31181/sa43202691}
}
EUVIMED