Digital Twins for Underground Infrastructure

We are presenting a plot that illustrates deflections on the X-axis as a function of depth on the Y-axis, accompanied by measured data points. On the left side, a dashed green line represents the most probable prediction, while a red line indicates the characteristic design scenario. The results demonstrate that our designs are significantly conservative. We aim to adopt methods that reduce this conservatism while ensuring safety and efficiency.

This brings us to the concept of digital twins in underground spaces. Digital twins represent a dual system, requiring both a digital and a physical counterpart. We utilize labeled data acquisition technologies to monitor the physical twin, allowing us to update the digital replica. This digital model incorporates multiple interoperable information models through data processing technologies. The defining feature of this twin system is the two-way connection: the digital twin is updated based on the physical twin, and conversely, the digital twin informs and influences the operation of the physical twin.

We often refer to the "HOUSE of Digital Twin" framework by Khajavi et al. This framework illustrates that digital twinning is not a singular concept; rather, it is a comprehensive process built upon various enabling technologies, including Building Information Modeling and wireless sensor networks. As we transition from data to information, machine learning, data analytics, and cloud computing play significant roles. Digital twinning focuses on integrating these technologies to achieve desired outcomes.

In the context of underground spaces, there is often ambiguity regarding the definition of a digital twin and its levels of maturity. To address this, we have collaborated with Elena and my PhD student Daish Bahanagar at Cambridge to clarify the dimensions of digital twins, their maturity levels, and assessment criteria.

We categorize digital twins into four dimensions:

1. Descriptive Twin: What is it?
2. Reflective Twin: What is happening?
3. Predictive Twin: What will happen?
4. Prescriptive Twin: What should be done?

For each dimension, we have developed a maturity rubric to assess the level of maturity, which includes criteria such as model quality and completeness for descriptive twins, prediction capabilities for predictive twins, and decision-making capabilities for prescriptive twins. The assessment metrics range from low to high, manual to autonomous, and single mode to multi-mode across the various levels of twins.