pith. sign in

arxiv: 2601.11650 · v2 · pith:WULVMVBDnew · submitted 2026-01-15 · ⚛️ physics.chem-ph · cs.AI

Large Language Model Agent for User-friendly Chemical Process Simulations

Jingkang Liang , Niklas Groll , G\"urkan Sin This is my paper

Pith reviewed 2026-05-22 11:42 UTC · model grok-4.3

classification ⚛️ physics.chem-ph cs.AI
keywords large language model agentchemical process simulationnatural language interfaceflowsheet optimizationprocess flowsheet synthesisAVEVA Process SimulationModel Context Protocol
0
0 comments X

The pith

An LLM agent connected to process simulation software lets users analyze, optimize, and build chemical flowsheets using plain-language instructions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper integrates a large language model agent with AVEVA Process Simulation through the Model Context Protocol so that the agent can receive natural-language commands and translate them into programmatic actions that run rigorous simulations. In water-methanol separation case studies the agent autonomously examines existing flowsheets, identifies improvement opportunities, performs iterative optimization, extracts results, and presents them in readable form. A second study shows the same system can construct new flowsheets either through guided dialogue or a single prompt. The authors position the tool as useful for teaching concepts to novices and for accelerating routine work by practitioners, while noting that expert review remains necessary to catch errors.

Core claim

Linking an LLM agent to a commercial process simulator via a protocol that permits Python-based control allows the agent to execute complex simulation tasks, including flowsheet analysis, opportunity identification, iterative optimization, data extraction, and clear reporting, directly from ordinary user instructions, as demonstrated in separation-process case studies that test both analysis-optimization and autonomous synthesis modes.

What carries the argument

The Model Context Protocol server that converts LLM-generated natural-language instructions into executable Python code for direct interaction with AVEVA Process Simulation.

If this is right

  • Step-by-step dialogue mode supports reliable guided flowsheet construction suitable for educational settings.
  • Single-prompt mode generates quick baseline flowsheets that experienced users can refine later.
  • The agent automates data extraction and speeds up routine analysis tasks for practitioners.
  • Natural-language interaction lowers the expertise barrier for early-stage process exploration by non-specialists.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar integrations could extend the same natural-language interface to other commercial simulators beyond the one tested here.
  • If reliability improves, the approach might shorten the time from initial idea to first working simulation in industrial design workflows.
  • The combination of guided and autonomous modes offers a practical spectrum for users with varying levels of experience.

Load-bearing premise

The language model will produce correct programmatic instructions that yield accurate simulation outcomes without repeated oversimplification or calculation mistakes that demand expert intervention.

What would settle it

Execute the agent on the water-methanol separation optimization task and check whether the produced operating points and performance metrics match those obtained by an expert user without any manual corrections to the agent's code or results.

Figures

Figures reproduced from arXiv: 2601.11650 by G\"urkan Sin, Jingkang Liang, Niklas Groll.

Figure 1
Figure 1. Figure 1: Schematic overview of the proposed framework including MCP. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Simulation flowsheet of APS example "C1 - Water Methanol Separation". [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Constructed simulation flowsheet via step-by-step dialogue with LLM agent. [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Schematic overview of an extended framework including MAS and RAG. [PITH_FULL_IMAGE:figures/full_fig_p025_4.png] view at source ↗
read the original abstract

Modern process simulators enable detailed process design, simulation, and optimization; however, constructing and interpreting simulations is time-consuming and requires expert knowledge. This limits early exploration by inexperienced users. To address this, a large language model (LLM) agent is integrated with AVEVA Process Simulation (APS) via Model Context Protocol (MCP), allowing natural language interaction with rigorous process simulations. An MCP server toolset enables the LLM to communicate programmatically with APS using Python, allowing it to execute complex simulation tasks from plain-language instructions. Two water-methanol separation case studies assess the framework across different task complexities and interaction modes. The first shows the agent autonomously analyzing flowsheets, finding improvement opportunities, and iteratively optimizing, extracting data, and presenting results clearly. The framework benefits both educational purposes, by translating technical concepts and demonstrating workflows, and experienced practitioners by automating data extraction, speeding routine tasks, and supporting brainstorming. The second case study assesses autonomous flowsheet synthesis through both a step-by-step dialogue and a single prompt, demonstrating its potential for novices and experts alike. The step-by-step mode gives reliable, guided construction suitable for educational contexts; the single-prompt mode constructs fast baseline flowsheets for later refinement. While current limitations such as oversimplification, calculation errors, and technical hiccups mean expert oversight is still needed, the framework's capabilities in analysis, optimization, and guided construction suggest LLM-based agents can become valuable collaborators.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper presents an LLM agent integrated with AVEVA Process Simulation (APS) via the Model Context Protocol (MCP) to enable natural-language control of rigorous chemical process simulations. It demonstrates the approach with two water-methanol separation case studies: one showing autonomous flowsheet analysis, opportunity identification, iterative optimization, data extraction, and result presentation; the other showing step-by-step and single-prompt autonomous flowsheet synthesis. The work notes current limitations including oversimplification, calculation errors, and technical issues that still require expert oversight, while highlighting potential benefits for education and routine engineering tasks.

Significance. If the reliability concerns can be addressed, the framework could meaningfully lower barriers to process simulation for novices and accelerate data-handling tasks for experts. The explicit MCP-based integration with a commercial simulator and the dual interaction modes constitute a practical engineering contribution that could serve as a template for LLM agents in other domain-specific software.

major comments (2)
  1. [Abstract and Case Studies] Abstract and Case Studies sections: the central claim of autonomous operation (analysis, iterative optimization, and flowsheet synthesis from plain-language prompts) is undercut by the explicit admission of 'oversimplification, calculation errors, and technical hiccups' that require expert oversight, yet no quantitative metrics (success rates, error frequency, retry counts, or human-intervention statistics) are reported for the LLM-generated MCP/Python actions.
  2. [Case Studies] Case Studies (water-methanol examples): the reported successes in autonomous iteration and clear presentation are presented qualitatively without baselines, error logs, or comparison to manual workflows, making it impossible to judge whether the demonstrated capabilities are representative or depend on unstated corrections.
minor comments (2)
  1. [Methods] The description of the MCP server toolset would benefit from an enumerated list of the specific Python-callable tools exposed to the LLM to improve reproducibility.
  2. [Case Studies] Figure captions in the case-study sections should explicitly label flowsheet elements and result tables so that readers can follow the agent's actions without referring back to the main text.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and for acknowledging the practical value of the MCP-based integration with AVEVA Process Simulation. We address each major comment below, clarifying the scope of our claims and indicating where revisions will strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract and Case Studies] Abstract and Case Studies sections: the central claim of autonomous operation (analysis, iterative optimization, and flowsheet synthesis from plain-language prompts) is undercut by the explicit admission of 'oversimplification, calculation errors, and technical hiccups' that require expert oversight, yet no quantitative metrics (success rates, error frequency, retry counts, or human-intervention statistics) are reported for the LLM-generated MCP/Python actions.

    Authors: We agree that quantitative metrics would provide valuable context for assessing reliability. The manuscript explicitly states the limitations and frames the work as a demonstration of the integration framework rather than a claim of perfect autonomy; the central contribution is the MCP-enabled natural-language interface that supports analysis, optimization, and synthesis with expert oversight. The case studies illustrate representative workflows rather than exhaustive benchmarks. We will revise the Abstract and Case Studies sections to include a brief discussion of observed error patterns and intervention points from the presented runs, together with an outline of metrics that could be used in future evaluations. revision: partial

  2. Referee: [Case Studies] Case Studies (water-methanol examples): the reported successes in autonomous iteration and clear presentation are presented qualitatively without baselines, error logs, or comparison to manual workflows, making it impossible to judge whether the demonstrated capabilities are representative or depend on unstated corrections.

    Authors: The case studies are intended as illustrative demonstrations of the two interaction modes (step-by-step and single-prompt) across different task complexities, not as a comparative performance study. We acknowledge that the absence of baselines or systematic error logs limits quantitative judgment of representativeness. We will add a short paragraph to the Case Studies section that explicitly notes the qualitative nature of the examples, describes the types of interventions observed, and suggests directions for future controlled comparisons with manual workflows. revision: partial

Circularity Check

0 steps flagged

No significant circularity in applied demonstration paper

full rationale

The paper describes an LLM agent integration with AVEVA Process Simulation via Model Context Protocol and presents two water-methanol separation case studies as empirical demonstrations of autonomous analysis, optimization, and flowsheet construction. No mathematical derivations, equations, fitted parameters, or self-referential definitions appear in the provided text or abstract. Claims rest on described workflows and acknowledged limitations (oversimplification, calculation errors requiring expert oversight) rather than any reduction of outputs to inputs by construction. Self-citations are absent from the load-bearing elements, and the work is self-contained as an applied engineering demonstration without invoking uniqueness theorems or ansatzes from prior author work.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on the functional existence of the Model Context Protocol for programmatic control of AVEVA Process Simulation and on the LLM's capacity to map natural language to correct simulation actions. No free parameters or new physical entities are introduced.

axioms (1)
  • domain assumption The Model Context Protocol server toolset enables reliable programmatic communication between the LLM and AVEVA Process Simulation.
    This is invoked to allow the agent to execute simulation tasks from plain-language instructions.

pith-pipeline@v0.9.0 · 5786 in / 1282 out tokens · 55106 ms · 2026-05-22T11:42:38.486365+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

82 extracted references · 82 canonical work pages · 6 internal anchors

  1. [1]

    Simulation based process design: Modeling and applications.Advanced Engineering Informatics, 26(4):763–781, October 2012

    Yoichiro Suzuki, Majid Yahyaei, Yan Jin, Hideo Koyama, and Gahee Kang. Simulation based process design: Modeling and applications.Advanced Engineering Informatics, 26(4):763–781, October 2012. ISSN 14740346. doi:10.1016/j.aei.2012.06.002

    work page doi:10.1016/j.aei.2012.06.002 2012

  2. [2]

    A VEV A Process Simulation

    A VEV A Group plc. A VEV A Process Simulation. Commercial software product page, 2025. URL https: //www.aveva.com/en/products/process-simulation/

    work page 2025

  3. [3]

    Aspen Plus

    AspenTech. Aspen Plus. Commercial software product page, 2025. URL https://www.aspentech.com/en/ products/engineering/aspen-plus

    work page 2025

  4. [4]

    The Review of Software Application Relevant to Chemical Engineering

    Mmekutmfon Peter. The Review of Software Application Relevant to Chemical Engineering. January 2023

    work page 2023

  5. [5]

    Deep Research Agents: A Systematic Examination And Roadmap, September 2025

    Yuxuan Huang, Yihang Chen, Haozheng Zhang, Kang Li, Huichi Zhou, Meng Fang, Linyi Yang, Xiaoguang Li, Lifeng Shang, Songcen Xu, Jianye Hao, Kun Shao, and Jun Wang. Deep Research Agents: A Systematic Examination And Roadmap, September 2025. arXiv:2506.18096 [cs]

    work page arXiv 2025

  6. [6]

    Dahm, Robert P

    Kevin D. Dahm, Robert P. Hesketh, and Mariano J. Savelski. Is Process Simulation Used Effectively in ChE Courses?Chemical Engineering Education, 36(3):192–203, July 2002. ISSN 2165-6428

    work page 2002

  7. [7]

    From LLM Reasoning to Autonomous AI Agents: A Comprehensive Review

    Mohamed Amine Ferrag, Norbert Tihanyi, and Merouane Debbah. From LLM Reasoning to Autonomous AI Agents: A Comprehensive Review, April 2025. arXiv:2504.19678 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  8. [8]

    Llm agents for education: Advances and applications,

    Zhendong Chu, Shen Wang, Jian Xie, Tinghui Zhu, Yibo Yan, Jinheng Ye, Aoxiao Zhong, Xuming Hu, Jing Liang, Philip S. Yu, and Qingsong Wen. LLM Agents for Education: Advances and Applications, March 2025. arXiv:2503.11733 [cs]

    work page arXiv 2025

  9. [9]

    A Survey on Code Generation with LLM-based Agents

    Yihong Dong, Xue Jiang, Jiaru Qian, Tian Wang, Kechi Zhang, Zhi Jin, and Ge Li. A Survey on Code Generation with LLM-based Agents, September 2025. arXiv:2508.00083 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  10. [10]

    DrugMCTS: a drug repurposing framework combining multi-agent, RAG and Monte Carlo Tree Search, July 2025

    Zerui Yang, Yuwei Wan, Siyu Yan, Yudai Matsuda, Tong Xie, Bram Hoex, and Linqi Song. DrugMCTS: a drug repurposing framework combining multi-agent, RAG and Monte Carlo Tree Search, July 2025. arXiv:2507.07426 [cs]

    work page arXiv 2025

  11. [11]

    Tradingagents: Multi-agents llm financial trading framework

    Yijia Xiao, Edward Sun, Di Luo, and Wei Wang. Tradingagents: Multi-agents llm financial trading framework. arXiv preprint arXiv:2412.20138, 2024

    work page arXiv 2024

  12. [12]

    Multi-agent systems for chemical engineering: a review and perspective.Current Opinion in Chemical Engineering, 51:101209, March 2026

    Sophia Rupprecht, Qinghe Gao, Tanuj Karia, and Artur M Schweidtmann. Multi-agent systems for chemical engineering: a review and perspective.Current Opinion in Chemical Engineering, 51:101209, March 2026. ISSN 22113398. doi:10.1016/j.coche.2025.101209

    work page doi:10.1016/j.coche.2025.101209 2026

  13. [13]

    Attention Is All You Need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention Is All You Need, August 2023. arXiv:1706.03762 version: 7

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  14. [14]

    Emergent Abilities of Large Language Models

    Jason Wei, Yi Tay, Rishi Bommasani, Colin Raffel, Barret Zoph, Sebastian Borgeaud, Dani Yogatama, Maarten Bosma, Denny Zhou, Donald Metzler, Ed H. Chi, Tatsunori Hashimoto, Oriol Vinyals, Percy Liang, Jeff Dean, and William Fedus. Emergent Abilities of Large Language Models, October 2022. arXiv:2206.07682 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  15. [15]

    Instruction Tuning for Large Language Models: A Survey.ACM Computing Surveys, page 3777411, November 2025

    Shengyu Zhang, Linfeng Dong, Xiaoya Li, Sen Zhang, Xiaofei Sun, Shuhe Wang, Jiwei Li, Runyi Hu, Tianwei Zhang, Guoyin Wang, and Fei Wu. Instruction Tuning for Large Language Models: A Survey.ACM Computing Surveys, page 3777411, November 2025. ISSN 0360-0300, 1557-7341. doi:10.1145/3777411

    work page doi:10.1145/3777411 2025

  16. [16]

    Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback

    Yuntao Bai, Andy Jones, Kamal Ndousse, Amanda Askell, Anna Chen, Nova DasSarma, Dawn Drain, Stanislav Fort, Deep Ganguli, Tom Henighan, Nicholas Joseph, Saurav Kadavath, Jackson Kernion, Tom Conerly, Sheer El-Showk, Nelson Elhage, Zac Hatfield-Dodds, Danny Hernandez, Tristan Hume, Scott Johnston, Shauna Kravec, Liane Lovitt, Neel Nanda, Catherine Olsson, ...

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  17. [17]

    Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, brian ichter, Fei Xia, Ed Chi, Quoc V Le, and Denny Zhou. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, editors,Advances in Neural Information Processing Systems, volume 35, pages 24824–24837. Curran Associ...

    work page 2022

  18. [18]

    A Survey on Large Language Models with some Insights on their Capabilities and Limitations, January 2025

    Andrea Matarazzo and Riccardo Torlone. A Survey on Large Language Models with some Insights on their Capabilities and Limitations, January 2025. arXiv:2501.04040 [cs]

    work page arXiv 2025

  19. [19]

    The Evolving Role of Large Language Models in Scientific Innovation: Evaluator, Collaborator, and Scientist, July

    Haoxuan Zhang, Ruochi Li, Yang Zhang, Ting Xiao, Jiangping Chen, Junhua Ding, and Haihua Chen. The Evolving Role of Large Language Models in Scientific Innovation: Evaluator, Collaborator, and Scientist, July

    work page

  20. [20]

    Evolving Roles of LLMs in Scientific Innovation: Assistant, Collaborator, Scientist, and Evaluator

    arXiv:2507.11810 [cs]. 26 LLM Agent for Chemical Process SimulationsA PREPRINT

    work page internal anchor Pith review Pith/arXiv arXiv

  21. [21]

    A survey on large language model based autonomous agents.Frontiers of Computer Science, 18(6):186345, March 2024

    Lei Wang, Chen Ma, Xueyang Feng, Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Jiakai Tang, Xu Chen, Yankai Lin, Wayne Xin Zhao, Zhewei Wei, and Jirong Wen. A survey on large language model based autonomous agents.Frontiers of Computer Science, 18(6):186345, March 2024. ISSN 2095-2236. doi:10.1007/s11704-024- 40231-1

    work page doi:10.1007/s11704-024- 2024

  22. [22]

    Agentic AI Frameworks: Architectures, Protocols, and Design Challenges, August 2025

    Hana Derouiche, Zaki Brahmi, and Haithem Mazeni. Agentic AI Frameworks: Architectures, Protocols, and Design Challenges, August 2025. arXiv:2508.10146 [cs]

    work page arXiv 2025

  23. [23]

    React: Synergizing reasoning and acting in language models

    Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik R Narasimhan, and Yuan Cao. React: Synergizing reasoning and acting in language models. InThe Eleventh International Conference on Learning Representations, 2023

    work page 2023

  24. [24]

    Toolformer: Language models can teach themselves to use tools

    Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu, Maria Lomeli, Eric Hambro, Luke Zettlemoyer, Nicola Cancedda, and Thomas Scialom. Toolformer: Language models can teach themselves to use tools. Advances in Neural Information Processing Systems, 36:68539–68551, 2023

    work page 2023

  25. [25]

    What is the Model Context Protocol (MCP)? Technical documentation, 2025

    Anthropic. What is the Model Context Protocol (MCP)? Technical documentation, 2025. URL https:// modelcontextprotocol.io/docs/getting-started/intro

    work page 2025

  26. [26]

    Tenenbaum, and Igor Mordatch

    Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, and Igor Mordatch. Improving factuality and reasoning in language models through multiagent debate. InForty-first International Conference on Machine Learning, 2024

    work page 2024

  27. [27]

    LangGraph

    LangChain Inc. LangGraph. Technical documentation, 2025. URL https://www.langchain.com/langgraph

    work page 2025

  28. [28]

    Microsoft. AutoGen. Technical documentation, 2025. URLhttps://microsoft.github.io/autogen

    work page 2025

  29. [29]

    Cognition

    Cognition AI Inc. Cognition. Artificial intelligence research laboratory and software platform, 2025. URL https://cognition.ai

    work page 2025

  30. [30]

    Announcing the agent2agent protocol (a2a)- google developers blog

    Google Developers. Announcing the agent2agent protocol (a2a)- google developers blog. Developer blog announcement, 2025. URL https://developers.googleblog.com/en/ a2a-a-new-era-of-agent-interoperability/

    work page 2025

  31. [31]

    DrugPilot: LLM-based Parameterized Reasoning Agent for Drug Discovery, July 2025

    Kun Li, Zhennan Wu, Shoupeng Wang, Jia Wu, Shirui Pan, and Wenbin Hu. DrugPilot: LLM-based Parameterized Reasoning Agent for Drug Discovery, July 2025. arXiv:2505.13940 [cs]

    work page arXiv 2025

  32. [32]

    Meng-Lin Tsai, Chong Wei Ong, and Cheng-Liang Chen. Exploring the use of large language models (LLMs) in chemical engineering education: Building core course problem models with Chat-GPT.Education for Chemical Engineers, 44:71–95, July 2023. ISSN 17497728. doi:10.1016/j.ece.2023.05.001

    work page doi:10.1016/j.ece.2023.05.001 2023

  33. [33]

    Gargalo, Julian Kager, Steen Larsen, Krist V

    Fiammetta Caccavale, Carina L. Gargalo, Julian Kager, Steen Larsen, Krist V . Gernaey, and Ulrich Krühne. ChatGMP: A case of AI chatbots in chemical engineering education towards the automation of repeti- tive tasks.Computers and Education: Artificial Intelligence, 8:100354, June 2025. ISSN 2666-920X. doi:10.1016/j.caeai.2024.100354

    work page doi:10.1016/j.caeai.2024.100354 2025

  34. [34]

    Matthew Keith, Eleanor Keiller, Christopher Windows-Yule, Iain Kings, and Phillip Robbins. Harnessing genera- tive AI in chemical engineering education: Implementation and evaluation of the large language model ChatGPT v3.5.Education for Chemical Engineers, 51:20–33, April 2025. ISSN 17497728. doi:10.1016/j.ece.2025.01.002

    work page doi:10.1016/j.ece.2025.01.002 2025

  35. [35]

    Gargalo, Krist V

    Fiammetta Caccavale, Carina L. Gargalo, Krist V . Gernaey, and Ulrich Krühne. Towards Education 4.0: The role of Large Language Models as virtual tutors in chemical engineering.Education for Chemical Engineers, 49:1–11, October 2024. ISSN 17497728. doi:10.1016/j.ece.2024.07.002

    work page doi:10.1016/j.ece.2024.07.002 2024

  36. [36]

    Complementary role of large language models in educating undergraduate design of distillation column: Methodology development.Digital Chemical Engineering, 9:100126, December 2023

    Zong Yang Kong, Vincentius Surya Kurnia Adi, Juan Gabriel Segovia-Hernández, and Jaka Sunarso. Complementary role of large language models in educating undergraduate design of distillation column: Methodology development.Digital Chemical Engineering, 9:100126, December 2023. ISSN 2772-5081. doi:10.1016/j.dche.2023.100126

    work page doi:10.1016/j.dche.2023.100126 2023

  37. [37]

    Gupta, Neereja Sundaresan, Thomas Alexander, Christopher J

    Daniil A. Boiko, Robert MacKnight, Ben Kline, and Gabe Gomes. Autonomous chemical research with large language models.Nature, 624(7992):570–578, December 2023. ISSN 0028-0836, 1476-4687. doi:10.1038/s41586- 023-06792-0

    work page doi:10.1038/s41586- 2023

  38. [38]

    Bran, Sam Cox, Oliver Schilter, Carlo Baldassari, Andrew D

    Andres M. Bran, Sam Cox, Oliver Schilter, Carlo Baldassari, Andrew D. White, and Philippe Schwaller. Aug- menting large language models with chemistry tools.Nature Machine Intelligence, 6(5):525–535, May 2024. ISSN 2522-5839. doi:10.1038/s42256-024-00832-8

    work page doi:10.1038/s42256-024-00832-8 2024

  39. [39]

    Chemllm: A chemical large language model

    Di Zhang, Wei Liu, Qian Tan, Jingdan Chen, Hang Yan, Yuliang Yan, Jiatong Li, Weiran Huang, Xiangyu Yue, Wanli Ouyang, Dongzhan Zhou, Shufei Zhang, Mao Su, Han-Sen Zhong, and Yuqiang Li. ChemLLM: A Chemical Large Language Model, April 2024. arXiv:2402.06852 [cs]. 27 LLM Agent for Chemical Process SimulationsA PREPRINT

    work page arXiv 2024

  40. [40]

    Batgpt-chem: A foundation large model for retrosynthesis prediction

    Yifei Yang, Runhan Shi, Zuchao Li, Shu Jiang, Bao-Liang Lu, Yang Yang, and Hai Zhao. BatGPT-Chem: A Foundation Large Model For Retrosynthesis Prediction, August 2024. arXiv:2408.10285 [cs]

    work page arXiv 2024

  41. [41]

    Constante Flores, and Can Li

    Abdullah Khan, Rahul Nahar, Hao Chen, Gonzalo E. Constante Flores, and Can Li. FaultExplainer: Leveraging large language models for interpretable fault detection and diagnosis.Computers & Chemical Engineering, 199: 109152, August 2025. ISSN 00981354. doi:10.1016/j.compchemeng.2025.109152

    work page doi:10.1016/j.compchemeng.2025.109152 2025

  42. [42]

    Wenli Du and Shaoyi Yang. The potential and challenges of large language model agent systems in chemical process simulation: from automated modeling to intelligent design.Frontiers of Chemical Science and Engineering, 19(10):99, October 2025. ISSN 2095-0179, 2095-0187. doi:10.1007/s11705-025-2587-5

    work page doi:10.1007/s11705-025-2587-5 2025

  43. [43]

    GPT Prompt Engineering for a Large Language Model- Based Process Improvement Generation System.Korean Journal of Chemical Engineering, 41(12):3263–3286, November 2024

    Donghyeon Lee, Jaewook Lee, and Dongil Shin. GPT Prompt Engineering for a Large Language Model- Based Process Improvement Generation System.Korean Journal of Chemical Engineering, 41(12):3263–3286, November 2024. ISSN 1975-7220. doi:10.1007/s11814-024-00276-1

    work page doi:10.1007/s11814-024-00276-1 2024

  44. [44]

    LLM-guided Chemical Process Optimization with a Multi-Agent Approach, 2025

    Tong Zeng, Srivathsan Badrinarayanan, Janghoon Ock, Cheng-Kai Lai, and Amir Barati Farimani. LLM-guided Chemical Process Optimization with a Multi-Agent Approach, 2025. arXiv:2506.20921 [cs]

    work page arXiv 2025

  45. [45]

    Toward Automated Simulation Research Workflow through LLM Prompt Engineering Design.Journal of Chemical Information and Modeling, 65(1):114–124, January 2025

    Zhihan Liu, Yubo Chai, and Jianfeng Li. Toward Automated Simulation Research Workflow through LLM Prompt Engineering Design.Journal of Chemical Information and Modeling, 65(1):114–124, January 2025. ISSN 1549-9596. doi:10.1021/acs.jcim.4c01653

    work page doi:10.1021/acs.jcim.4c01653 2025

  46. [46]

    Application of Artificial Intelligence in process simulation tool

    Nikhil Rajeev, Suresh Jayaraman, Prajnan Das, and Srividya Varada. Application of Artificial Intelligence in process simulation tool. pages 1769–1774, Ghent, Belgium, July 2025. doi:10.69997/sct.126215

    work page doi:10.69997/sct.126215 2025

  47. [47]

    Seider, Daniel R

    Warren D. Seider, Daniel R. Lewin, J. D. Seader, Soemantri Widagdo, Rafiqul Gani, and Ka Ming Ng.Product and Process Design Principles: Synthesis, Analysis and Evaluation. John Wiley & Sons, May 2016. ISBN 978-1-119-28263-1

    work page 2016

  48. [48]

    Introducing Claude 4

    Anthropic. Introducing Claude 4. Product announcement, 2025. URL https://www.anthropic.com/news/ claude-4

    work page 2025

  49. [49]

    Text Arena | LMArena

    LMArena Team. Text Arena | LMArena. Online benchmark leaderboard, 2025. URL https://lmarena.ai/ leaderboard/text

    work page 2025

  50. [50]

    Jeremiah Lowin. FastMCP. Open-source MCP server and client framework, 2025. URL https://gofastmcp. com

    work page 2025

  51. [51]

    Efficient methanol synthesis: Perspectives, technologies and optimiza- tion strategies.Progress in Energy and Combustion Science, 56:71–105, September 2016

    Giulia Bozzano and Flavio Manenti. Efficient methanol synthesis: Perspectives, technologies and optimiza- tion strategies.Progress in Energy and Combustion Science, 56:71–105, September 2016. ISSN 03601285. doi:10.1016/j.pecs.2016.06.001

    work page doi:10.1016/j.pecs.2016.06.001 2016

  52. [52]

    Anggara Dwita Burmana, Yacine Benguerba, Filicia Wicaksana, Vikram Alexander, Silvia Nova, Taslim Taslim, and Iriany Iriany. Improving efficiency and product quality by optimizing methanol purification in biodiesel plant: Addressing abnormalities in distillation column performance.International Journal of Thermofluids, 25:101019, January 2025. ISSN 266620...

    work page doi:10.1016/j.ijft.2024.101019 2025

  53. [53]

    Isobaric vapor-liquid equilibria for methanol + ethanol + water and the three constituent binary systems.Journal of Chemical & Engineering Data, 38(3): 446–449, July 1993

    Kiyofumi Kurihara, Mikiyoshi Nakamichi, and Kazuo Kojima. Isobaric vapor-liquid equilibria for methanol + ethanol + water and the three constituent binary systems.Journal of Chemical & Engineering Data, 38(3): 446–449, July 1993. ISSN 0021-9568, 1520-5134. doi:10.1021/je00011a031

    work page doi:10.1021/je00011a031 1993

  54. [54]

    Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell

    Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? InProceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, FAccT ’21, pages 610–623, New York, NY , USA, March 2021. Association for Computing Machinery. ISBN 978-1-4503-8309-...

    work page doi:10.1145/3442188.3445922 2021

  55. [55]

    A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions,

    Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qianglong Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, and Ting Liu. A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions.ACM Trans. Inf. Syst., 43(2):42:1–42:55, January 2025. ISSN 1046-8188. doi:10.1145/3703155

    work page doi:10.1145/3703155 2025

  56. [56]

    Gargalo, Siwanat Chairakwongsa, Gürkan Sin, and Rafiqul Gani

    Alberto Quaglia, Carina L. Gargalo, Siwanat Chairakwongsa, Gürkan Sin, and Rafiqul Gani. Systematic network synthesis and design: Problem formulation, superstructure generation, data management and solution.Computers & Chemical Engineering, 72:68–86. ISSN 00981354. doi:10.1016/j.compchemeng.2014.03.007

    work page doi:10.1016/j.compchemeng.2014.03.007 2014

  57. [57]

    V ollmer, Resul Al, Krist V

    Nikolaus I. V ollmer, Resul Al, Krist V . Gernaey, and Gürkan Sin. Synergistic optimization framework for the process synthesis and design of biorefineries.Frontiers of Chemical Science and Engineering, 16(2):251–273. ISSN 2095-0179, 2095-0187. doi:10.1007/s11705-021-2071-9

    work page doi:10.1007/s11705-021-2071-9 2095

  58. [58]

    Gernaey, and Gürkan Sin

    Resul Al, Chitta Ranjan Behera, Krist V . Gernaey, and Gürkan Sin. Stochastic simulation-based superstructure optimization framework for process synthesis and design under uncertainty.Computers & Chemical Engineering, 143:107118. ISSN 00981354. doi:10.1016/j.compchemeng.2020.107118. 28 LLM Agent for Chemical Process SimulationsA PREPRINT

    work page doi:10.1016/j.compchemeng.2020.107118 2020

  59. [59]

    Mathematical exploration and discovery at scale

    Terence Tao. Mathematical exploration and discovery at scale. Blog post, November 2025. URL https: //terrytao.wordpress.com/2025/11/05/mathematical-exploration-and-discovery-at-scale/

    work page 2025

  60. [60]

    Hallucination to Truth: A Review of Fact-Checking and Factuality Evaluation in Large Language Models.Artificial Intelligence Review, 59(2):70, January 2026

    Subhey Sadi Rahman, Md Adnanul Islam, Md Mahbub Alam, Musarrat Zeba, Md Abdur Rahman, Sadia Sultana Chowa, Mohaimenul Azam Khan Raiaan, and Sami Azam. Hallucination to Truth: A Review of Fact-Checking and Factuality Evaluation in Large Language Models.Artificial Intelligence Review, 59(2):70, January 2026. ISSN 1573-7462. doi:10.1007/s10462-025-11454-w

    work page doi:10.1007/s10462-025-11454-w 2026

  61. [61]

    C1 - Water Methanol Separation

    Fengli Xu, Qianyue Hao, Chenyang Shao, Zefang Zong, Yu Li, Jingwei Wang, Yunke Zhang, Jingyi Wang, Xiaochong Lan, Jiahui Gong, Tianjian Ouyang, Fanjin Meng, Yuwei Yan, Qinglong Yang, Yiwen Song, Si- jian Ren, Xinyuan Hu, Jie Feng, Chen Gao, and Yong Li. Toward large reasoning models: A survey of reinforced reasoning with large language models.Patterns, 6(...

    work page doi:10.1016/j.patter.2025.101370 2025

  62. [62]

    Increase Number of Stages - Current: 13 theoretical stages - Option: Add 2-5 more stages (15-18 total) to improve separation efficiency - Benefit: Higher purity products with potentially lower reflux ratio

    work page

  63. [63]

    Optimize Feed Stage Location - Current: Stage 7 (middle of column) - Option: Test feed stages 6-8 to find optimal thermal efficiency - Benefit: Minimize reboiler duty while maintaining product quality

    work page

  64. [64]

    Add Side Streams - Current: Simple distillation with overhead/bottoms - Option: Add side draws for intermediate compositions - Benefit: Better heat integration and product flexibility Operating Parameter Optimization

    work page

  65. [65]

    Reflux Ratio Adjustment - Current: RR = 1.0 (conservative) - Option: Reduce to 0.7-0.8 or increase to 1.2-1.5 depending on objectives - Lower RR: Reduces energy consumption but may compromise purity - Higher RR: Improves separation but increases energy costs

    work page

  66. [66]

    Pressure Optimization - Current: 101.3 kPa top pressure - Option: Operate at higher pressure (150-200 kPa) for better heat integration - Benefit: Higher condensing temperature enables heat recovery

    work page

  67. [67]

    Feed Conditioning - Current: Feed at 351K (subcooled liquid, VF = -0.33) 33 LLM Agent for Chemical Process SimulationsA PREPRINT - Option: Preheat feed closer to bubble point or partial vaporization - Benefit: Reduces reboiler duty requirements Advanced Process Configurations

    work page

  68. [68]

    Heat Integration - Install feed-effluent heat exchangers - Use hot bottoms product to preheat feed - Benefit: 15-25% energy savings typically achievable

    work page

  69. [69]

    Double-Effect Configuration - Operate two columns at different pressures - Use overhead vapor from high-pressure column to drive low-pressure reboiler - Benefit: ˜40% energy reduction possible

    work page

  70. [70]

    Reactive Distillation - If methanol purification is the goal, consider reactive separation - Benefit: Simultaneous reaction and separation Equipment Modifications

    work page

  71. [71]

    Enhanced Tray Design - Current: 100% efficiency assumption - Option: Install high-efficiency trays (structured packing) - Benefit: Better mass transfer, potentially fewer actual trays needed

    work page

  72. [72]

    Advanced Control Systems - Implement composition control on key streams - Add feedforward control for feed disturbances - Benefit: More consistent product quality and energy efficiency Immediate Quick Wins Most promising near-term improvements: - Optimize reflux ratio (test 0.8-1.2 range) - Preheat feed to reduce subcooling - Add 2-3 stages if column modi...

    work page

  73. [73]

    Sep_MeOHWater

    Create simulation "Sep_MeOHWater"

    work page

  74. [74]

    Sep_MeOHWater

    Open simulation "Sep_MeOHWater" Answer 2.0.0: 35 LLM Agent for Chemical Process SimulationsA PREPRINT I’ll help you construct a simulation model for the separation of water and methanol using AVEVA Process Simulation. Let me follow your steps exactly. Step 1: Connect to APS − →aps_connect()→(connection: true, ...) Great! Successfully connected to APS. I c...

    work page

  75. [75]

    Feed Source- for the feed stream

    work page

  76. [76]

    Distillation Column- for separation

    work page

  77. [77]

    Let me check the Feed model variables: − →model_all_vars(model_name: Feed)→(success: true, ...) Perfect! Now I can see the variable names

    Two Sinks- for distillate and bottoms products − →model_add(x: 100, y: 200, model_name: Feed, model_type: Source)→(success: true, ...) − →model_add(x: 300, y: 200, model_name: Column, model_type: Column)→(success: true, ...) − →model_add(x: 450, y: 150, model_name: Distillate, model_type: Sink)→(success: true, ...) − →model_add(x: 450, y: 250, model_name:...

    work page

  78. [78]

    Feed Source(Feed) - configured with: - 10 kg/s mass flow rate - 80°C (353.15 K) temperature - 1 bar (101.325 kPa) pressure - 50 mol% Water, 50 mol% Methanol - WaterMethanol_NRTL fluid package

    work page

  79. [79]

    Distillation Column(Column) - configured with: - 15 theoretical stages - Internal condenser - Internal reboiler - WaterMethanol_NRTL fluid package

    work page

  80. [80]

    Product Sinks- Distillate and Bottoms Now I need you to connect the models manually using the AVEVA interface:

    work page

Showing first 80 references.