Lightweight Transformer Models for On-Device Fault Detection: A Benchmark Study on Resource-Constrained Deployment
arXiv:2606.24173v1 Announce Type: cross Abstract: On-device fault detection enables real-time diagnostics without cloud dependency, but deploying machine learning models on resource-constrained hardware demands careful tradeoffs between accuracy, latency, and model size. We present a benchmark comparing traditional ML methods (Random Forest, XGBoost, SVM, Logistic Regression) against lightweight transformer architectures (DistilBERT, TinyBERT-6L, TinyBERT-4L, MobileBERT) for binary fault detection across three public datasets: NASA C-MAPSS turbofan degradation, SECOM semiconductor manufacturin
The proliferation of edge computing devices and the need for immediate, localized diagnostics are driving research into efficient on-device machine learning models.
This development improves real-time operational efficiency and reduces reliance on cloud infrastructure for critical applications like fault detection, enhancing system resilience and data privacy.
The feasibility of deploying sophisticated AI models, specifically transformer architectures, directly onto resource-constrained edge devices for immediate decision-making is increasing.
- · Edge computing hardware manufacturers
- · Industrial IoT sectors
- · AI model optimization companies
- · Companies with remote assets
- · Cloud-dependent diagnostic solutions
- · Legacy fault detection hardware
- · High-latency systems
Increased adoption of AI for proactive maintenance and operational monitoring in diverse industrial and consumer settings.
Reduced operational costs and improved uptime for machinery and systems across various industries due to faster, localized fault detection.
Potential for new business models centered around 'AI as a sensor' or 'AI as a maintenance layer' directly integrated into products.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI