Date of Award
2026
Degree Name
Computer Science
College
College of Engineering and Computer Sciences
Type of Degree
M.S.
Document Type
Thesis
First Advisor
Dr. Paulus Wahjudi
Second Advisor
Dr. Husnu Narman
Third Advisor
Dr. Sudipta Chowdhury
Fourth Advisor
Dr. Ammar Alzarrad
Abstract
This thesis presents a systems-based virtual reality training framework for safety-critical industrial maintenance tasks, using hydroelectric facility maintenance as a case study. The system integrates finite-state task modeling, event-level telemetry, and state-aware instructional feedback to guide participants through procedural tasks involving reservoir operation, rotor replacement, fan replacement, clog removal, and leak repair. Rather than relying solely on final task completion, the framework captures sequential user behavior through logged actions, wrong-order events, safety violations, hesitation events, completion time, and system disruption outcomes.
A user study was conducted with 23 participants to evaluate how state-aware feedback influenced procedural performance, safety behavior, and task progression across repeated simulation runs. A total of 23 participants were recruited, of whom 19 completed the full procedure and were included in the primary analysis. Participants completed multiple hydroelectric maintenance scenarios under either a feedback-supported condition or a no feedback condition. The resulting data were analyzed to compare error counts, safety violations, completion duration, hesitation behavior, and disruption-related outcomes between conditions and across task repetitions.
The findings suggest that state-aware feedback was associated with fewer procedural and safety-related mistakes, although the observed effects were limited by sample size and should be interpreted as exploratory rather than confirmatory. The core contribution of this work is a VR training architecture designed for interpretability and reliable detection of unsafe actions. The system models procedural state, flags unsafe or incorrect actions as they occur, and generates performance data in which each assessment traces back to an observable participant action. The results demonstrate that this design provides a reliable and auditable basis for procedural skill assessment in safety-critical industrial training environments.
Subject(s)
Computer science.
Virtual reality.
Data transmission systems.
Machine theory.
Maintenance.
Water-power.
Industrial safety.
System safety.
Feedback control systems.
Recommended Citation
Hatfield, Joshua, "State-aware adaptive feedback and performance telemetry in virtual reality industrial maintenance training" (2026). Theses, Dissertations and Capstones. 2106.
https://mds.marshall.edu/etd/2106
