Digital twin applications are now firmly established in the energy industry, long valued for their predictive capacities. But rapid advances in artificial intelligence (AI) capabilities are now making digital twins even more useful by formulating and facilitating actions based on those predictions.
That ability to convert prediction into action is the driving force behind multi-agent driven Front Running Simulation (FRS), a concept explored in a recent paper by digital twin pioneer Dr Michael Grieves, Executive Director of the Digital Twin Institute.
The paper – Front Running Simulation: A Digital Twin Framework for Real-Time Replication, Prediction, and Goal Navigation – explains how FRS can
continuously synchronise with a live system, predict probable futures using simulation and AI, and identify navigable information actions that move a system toward its objectives with minimal wasted time, energy, and material.
The Digital Twin Consortium (DTC), an industry group, has announced member adoption of an initiative to move the framework from concept to real-world application. This centres initially on developing a working FRS implementation now in operation on a hydrocarbon processing plant digital twin, which aims to safely convert predictions into executable actions, while maintaining human oversight. The FRS concept is also applicable in a range of other industries where digital twins are used
“FRS answers the question every digital twin programme eventually hits: what do you do with a prediction?”, Dan Isaacs, DTC’s General Manager and Chief Technology Officer, said.
“For our members, it is a concrete, demonstrated path from foresight to guarded, auditable action. For the broader ecosystem, it is a reusable pattern — fusing causation and correlation, gated by safety and confidence — that applies across process, energy, water, and discrete manufacturing and several other applications,” he added.
The FRS framework is underpinned by a Simulation Integration Engine, which fuses model-driven causation from physics-based models with data-driven correlation from current and historical data, which is continuously updated.
Grieves and Pieter van Schalkwyk, CEO of agentic operations firm XMPro, explained how real-world applications would work in a recent webinar hosted by DTC.
Grieves said that, while the idea of front running simulation was not new, the simulation and AI computing capabilities needed to implement it effectively were lacking until recently.
“Simply put, FRS consists of replicating current reality, predicting the future outcomes, and navigating to reach our goals,” he told the webinar.
Paradigm shift
Grieves said this marked a break with the way traditional digital twins were often used to detect, diagnose, and respond to events that had already occurred, or used for predictive maintenance, which, although forward-looking, was typically focused on reacting before a component fails.
“Front-running simulation changes the operational paradigm by asking: What is likely to happen next, and what should we do to achieve our goals?” Grieves said.
Grieves has worked with XMPro to develop an FRS platform simulation linked to a digital twin for a depropanizer, which removes propane from the raw feedstock in a hydrocarbons plant. The simulation, demonstrated by van Schalkwyk in the webinar, is based on physics-based prediction, a causal model, a deterministic safety guardian, and a confidence tier. The result is a continuously updated list of options based on multiple variables, and with assigned confidence levels, for actions to rectify potential problems the digital twin has identified.
Although FRS could ultimately trigger such actions automatically without human intervention, that is unlikely to happen until there is more confidence in the guardrails still being put in place around AI.
“My tendency right now is to say we should have humans in the loop who get fed this [FRS output] and then make the final determination. Because quite frankly, we’re the people that have to live with the execution,” Grieves said.
(Photo: Shutterstock)
