A Single-Device Environment-Adaptive Mixed Reality Framework for Real-Time Industrial Fault Diagnosis

Abstract

In industrial environments, monitoring and fault diagnosis of mechanical equipment face challenges such as spatial localization drift and delays in real-time data rendering, especially in complex settings with low illumination, weak textures, and strong interference. Traditional methods struggle to effectively integrate monitoring data with physical entities, increasing cognitive load and reducing diagnostic accuracy. To address these issues, we propose the Single-Device Mixed Reality (SEMR) framework, a novel solution that enhances industrial equipment monitoring and fault diagnosis. The framework integrates three key mechanisms: an environment-aware model that adjusts the confidence of Simultaneous Localization and Mapping (SLAM) to ensure precise spatial registration, a Kalman filter-based motion prediction to reduce rendering delays, and a fault-tolerant gaze interaction system for hands-free operation. Experimental results demonstrate that SEMR reduces the spatial registration error by 52.1%, from 14.2 cm to 6.8 cm, and decreases latency during dynamic inspections by 26.7%, improving diagnostic accuracy and real-time performance. The proposed method provides a cost-effective and reliable solution for enhancing industrial fault diagnosis and equipment monitoring, particularly in challenging environments.