LLM-Driven Human-AI Collaborative Decision Support System for Complex Industrial Processes: A Case Study in Metallurgy
First-author journal article on LLM-driven human-AI collaborative decision support for complex metallurgical processes.
About
I am a PhD student in Control Science and Engineering at Northeastern
University, China. My research focuses on AI and large-model applications in complex industrial scenarios, with interests in industrial process modeling, multimodal perception, analysis, and decision support. I am strong at abstracting tasks from real problems, connecting data with models, and turning research results into verifiable, deployable systems.
Education
PhD Candidate, Control Science and Engineering
2024-08
Northeastern University
Master of Engineering
2021-09
2024-06
Anhui University of Technology
Bachelor of Engineering in Automation
2017-09
2021-06
Anhui University of Technology
Research Interests
Industrial LLM agents
RAG and tool-augmented decision support
Semantic time-series modeling
Intelligent metallurgy
Multimodal medical AI
Research
My research develops large-model systems for complex industrial and clinical settings where decisions must be grounded in process data, domain knowledge, and deployable software.
- Industrial LLM agents and decision support. RAG, tool calling, Text2SQL, and human-in-the-loop workflows for production-facing reasoning and operational decisions.
- Semantic time-series modeling. Methods that connect sensor sequences, task descriptions, and process semantics for adaptive prediction under changing operating conditions.
- Application-grounded AI. Metallurgical process intelligence remains the main application base, with selected work extending multimodal perception and reasoning to oral mucosal medical analysis.
Experience
Project Lead, Multimodal LLM Virtual Doctor for Oral Mucosal Dermoscopy
Xiangya Second Hospital
Built a multimodal large-model agent workflow for real clinical scenarios. The workflow integrates dermoscopy images, patient reports, and medical knowledge to support oral mucosal disease analysis, reasoning, and virtual-doctor-assisted diagnosis.
Multimodal LLMs
Medical AI
Agent workflows
Project Lead, Digital-Twin Large Model for Thickening-Dewatering
Baowu Resources
Built a production-facing thickening-dewatering digital-twin large-model workflow. The system connects field signals, condition recognition, a RAG knowledge base, Text2SQL, and tool calling for perception and decision support.
Industrial LLMs
Digital twins
RAG
Project Member, National Key Research and Development Program
National Key R&D Program
Developed 5G-enabled endpoint-prediction modules for mineral-processing and metallurgical processes. The modules connect industrial data access, online state estimation, and web dashboards to support process monitoring and endpoint judgment.
5G industrial systems
Endpoint prediction
Process monitoring
Project Lead, LF-Furnace Visual Perception and Energy-Saving Control
Baowu Magang
Delivered LF-furnace visual perception and energy-saving control for field production. The work connected site perception, networking data, model prediction, control logic, and production software.
Visual perception
Energy-saving control
Metallurgy
Project Lead, Intelligent Transformation of the Iron-Tapping Process
Baowu Changgang Ironmaking Plant
Built an intelligent-transformation workflow for the iron-tapping process. The workflow connected molten-iron level perception, data communication, algorithm services, and software platforms, and completed digital-twin scenario construction and deployment.
Ironmaking
Digital twins
Industrial deployment
Education
PhD Candidate, Control Science and Engineering
Northeastern University
College of Information Science and Engineering.
Master of Engineering
Anhui University of Technology
School of Metallurgical Engineering
Bachelor of Engineering in Automation
Anhui University of Technology
School of Electrical and Information Engineering.
Publications
(2026).
LLM-Driven Human-AI Collaborative Decision Support System for Complex Industrial Processes: A Case Study in Metallurgy.
Neural Networks.
(2026).
Zero-Shot Capillary Segmentation in Dermoscopy Images via SAM2: A Case Study on Oral Mucosa.
JBHI.
(2025).
A meta-contrastive learning hybrid model for adaptive temperature trend prediction in variable ladle preheating.
EAAI.
Research Highlights
A meta-contrastive learning hybrid model for adaptive temperature trend prediction in variable ladle preheating
First-author journal article on adaptive temperature trend prediction in variable ladle preheating.
Hybrid Grid Search and Bayesian optimization-based random forest regression for predicting material compression pressure in manufacturing processes
First-author journal article on random-forest-based pressure prediction in manufacturing processes.
Iron-Tapping State Recognition of Blast Furnace Based on Bi-GRU Composite Model and Post-Processing Classifier
First-author journal article on blast furnace iron-tapping state recognition.