TWINSENSE - AI-based Virtual Sensing for Enhanced Real-time Understanding and Learning Systems Enhancement

AI-powered virtual sensing technology for industrial asset monitoring with improved proactive maintenance accuracy.

The TWINSENSE testbed improves industrial monitoring by validating digital twin technology's capability to perform real-time virtual measurements of critical variables across diverse industrial assets. The testbed addresses the persistent challenge of measuring inaccessible or costly-to-monitor variables by leveraging advanced digital twins for continuous virtual sensing, enabling AI-driven proactive maintenance strategies that transform how industries monitor their most critical equipment.

The testbed integrates high-fidelity physics-based digital twin models with AI-powered novelty detection algorithms, utilizing transfer learning techniques to calibrate systems using both virtual and real-world data. The implementation employs a multi-asset comparative validation approach, conducting blind testing protocols to compare virtual versus physical measurements where possible. The testbed validates transfer learning effectiveness through domain adaptation experiments between digital and physical environments, ensuring AI model calibration accuracy while addressing real-time computational constraints through optimized edge and cloud computing architectures.

Expected outcomes include achieving a certain level of virtual sensing accuracy deviation from physical measurements, fault detection accuracy with low false positive rates, and improvement in AI model performance when combining virtual and real data. These results will enable continuous monitoring of previously inaccessible parameters and deliver targeted improvement in proactive maintenance accuracy with enhanced remaining useful life estimation for strategic maintenance planning

The TWINSENSE testbed contributes to industry advancement in the following ways:

1. Virtual Sensing Standards: Establishing accuracy benchmarks and validation protocols for virtual measurements across industrial applications, providing industry-wide reference frameworks for digital twin-based monitoring systems.
2. Transfer Learning Methodologies: Developing and validating methodologies for digital-physical domain adaptation in industrial contexts, enabling scalable AI deployment across diverse equipment types and operational environments.
3. Proactive Maintenance Guidelines: Creating comprehensive best practices for AI-driven maintenance using virtual sensing data, advancing predictive maintenance capabilities beyond traditional sensor-dependent approaches.
4. Virtual Instrumentation Frameworks: Advancing virtual instrumentation standards for industrial applications, demonstrating reliable measurement capabilities for inaccessible or high-cost monitoring scenarios.
5. AI Model Calibration Techniques: Establishing validated approaches for combining virtual and physical data to enhance AI model performance, contributing to improved fault detection and diagnostic accuracy across industrial sectors.
6. Implementation Roadmaps: Providing validated deployment strategies for virtual sensing across manufacturing, process industries, and critical infrastructure, enabling systematic adoption of digital twin monitoring technologies.

Lead Developer

Co-Developer