Siyuan Chen / Research library
Research on maintenance decisions, industrial AI and robot programming. Each publication has an abstract or article overview, key points and a ready-to-use citation.
Download all citations ↗Journal of Manufacturing Systems
This review combines published research with industry interviews to examine AI-enhanced digital twins in maintenance. It identifies gaps between research and industrial practice and proposes a five-layer implementation framework.
Engineering Applications of Artificial Intelligence
False alarms in glue-line inspection lead to unnecessary interventions and downtime. We compare 17 unsupervised models and 28 backbones on industrial images, examining detection accuracy, robustness and inference time with limited training data.
IOP Conference Series: Materials Science and Engineering
Equipment lifetime prediction and fault diagnosis are often modeled separately. This Transformer shares a sensor-data encoder across both tasks and is evaluated on data from 100 industrial machines.
arXiv:2608.21417 · Accepted for CIE53 2026
This workflow connects retrieval-grounded ABB RAPID generation to RobotStudio through MCP. Simulation feedback reveals execution errors that static code checks miss and supports iterative correction of robot programs.
Proceedings - Winter Simulation Conference
Maintenance priorities change with the state of a production system. This study combines a digital twin with proximal policy optimization to prioritize maintenance tasks in a lab-scale drone factory.
SSRN 6681288
Maintenance logs contain diagnostic knowledge that numerical-only agents cannot use directly. This preprint combines local LLMs with hierarchical multi-agent reinforcement learning and a production-cost reward in a lab-scale drone assembly study.
Licentiate thesis · Chalmers University of Technology
This licentiate thesis examines the transition from failure prediction to actionable maintenance decisions. It brings together five studies on industrial requirements, digital twins, reinforcement learning and anomaly detection.
Proceedings - Winter Simulation Conference
Interviews and a workshop identify what industrial stakeholders need from a maintenance digital twin. The study maps implementation requirements, decision-support needs and barriers to adoption.
IOP Conference Series: Materials Science and Engineering
Robot code generation needs reliable, domain-specific engineering knowledge. This paper explores RAG architectures that ground language models in controlled sources while retaining human oversight and traceability.
Proceedings - Winter Simulation Conference
A drone-factory digital twin supports maintenance comparisons without interrupting real production. The demonstrator evaluates what-if scenarios and combines performance indicators with structured decision analysis.
IOP Conference Series: Materials Science and Engineering
Adaptive robot instruction can change the demands placed on industrial trainees. This study compares standard and personalized humanoid-robot instructions, examining mental workload and task performance.
Proceedings - Winter Simulation Conference
This PyTorch-based simulation generates synthetic data for discrete manufacturing systems. Timed Petri nets and a directed graph represent production flow, creating an expandable environment for manufacturing AI experiments.
World Manufacturing Report 2024
This award-winning essay examines opportunities and barriers for generative AI in industrial maintenance. It discusses synthetic data, workforce learning, implementation costs, data quality and governance.