Digital Twin-Based Predictive Fault Detection and Diagnosis Architecture

Identifies and Prioritizes Potential Faults in Constructed Facilities for Early Intervention

This digital twin-based architecture predicts, detects, and prioritizes faults in buildings. Facilities management systems help ensure the efficient and safe operation of constructed facilities. When a fault eventually arises, it requires time, money, and resources to diagnose and fix. Being able to predict necessary maintenance saves more money than standard preventative or corrective maintenance1. Current facility management systems are constrained by high implementation failure rates, difficulties in data and system integration, insufficient support for predictive maintenance, and basic data visualization features.

 

Researchers at the University of Florida have developed a fault detection framework that integrates criticality analysis with digital replicates of the facility, known as digital twins, to predict and diagnose necessary maintenance. Identifying the fault at early stages reduces downtime and streamlines the process of diagnosing and correcting. This process enhances existing fault detection and diagnosis systems by monitoring the output of assets. When the most critical faults are first identified the system prioritizes them, considering the cost and impact of downtime and loss of functionality and visually displays the list and necessary measures in a convenient user interface.

 

Application

Preemptive detection, diagnosis, and prioritization of structural faults in buildings and constructed facilities

 

Advantages

  • The model prioritizes predicting faults, reducing cost and impact of downtime
  • Standardizes the prioritization scheme, ensuring a consistent response
  • Visually prioritizes output, making interpretation easier
  • Digital twin architecture is highly flexible, enabling its integration into existing software

 

Technology

This framework consists of a three-layer system: the information, application, and service and visualization layers, where the output from each layer feeds into the next. The first layer collects data from various sources, including machine outputs and sensor outputs, containing data related to the managed asset. The contents of the first layer serve as the input for the second layer, which processes the data. The second layer processes the collected data from the first layer to perform fault detection, diagnosing issues, predicting faults, and generating an alarm if necessary. The output from the second layer is then interpreted and visualized by the third layer, which displays the results in a prioritized manner and can predict potential failures over time.

Patent Information: