摘要：AbstractDigital Twins (DT) and their applications in prognostics and diagnostics have been rapidly growing and becoming highly valuable for various industries. They possess the ability to predict the Remaining Useful Life (RUL) of a device and avoid costly maintenance or failure. Digital Twins can be defined as a real-time bidirectional data exchange of a physical entity and its cyber manifestation. This technology relies on sophisticated development and effective sensor placement to excel in smart maintenance related fields. However, too many sensors can jeopardize the ability of the Digital Twin to perform effectively. A surplus of sensors corresponds with large data sets which increase the computational expense. LIVE Digital Twin is a comprehensive model-based solution to develop Digital Twins for asset management enabled through sensor communication. The LIVE structure is composed of four principle phases, Learn, Identify, Verify, and Extend, with the purpose of designing and optimizing the digital twin using multi-physics simulations. The goal of this research is to present the LIVE architecture to prescribe effective sensor locations for Digital Twins. Furthermore, without loss of generality, the Learn and Identify phases are applied in a case study to develop a maintenance solution for a Light Rail Transit system. To achieve this, various dynamic simulation data was collected from a high-fidelity model to calibrate the performance of a low fidelity model. The advantageous computational speed of low fidelity models makes it possible to quickly analyze various faulty conditions for the system. As the system degrades, the natural frequencies of the vibration deviate thus leading to potential resonance and system failure. In comparing the faulty system’s response to the healthy case, an optimal sensor position is located.
