The Digital Thread connecting it all together
The Digital Thread connecting it all together: Data-Driven Architectures
The term Digital Thread made its first appearance in a 2013 Global Horizon report, but wasn't officially defined until 2018 when Singh & Willcox at MIT wrote about Digital Threads in a white paper titled "Engineering with a Digital Thread". In this academic essay the term Digital Thread is described as "a data-driven architecture that links together information generated from across the product lifecycle and is envisioned to be the primary or authoritative data and communication platform for a company’s products at any instance of time."
In essence, Digital Threads are data pipelines supported by modern enterprise architectures that attempt to weave data sources together, such as, sensors capturing changes in temperature or humidity, vibratory data generated by a physical asset like a pump, or instrumentation data sitting in an Enterprise Asset Management database. With most data being stored in fragmented databases today, that is later connected by way of code heavy point to point data integrations, Digital Threads aim to simplify and modernize legacy approaches to democratizing, governing, controlling, securing, analyzing, and integrating information.
The end goal for many organizations is to create, and leverage autonomous and intelligent data architectures and technologies that automate how information is collected, connected, and analyzed, analogous to how the human-mind processes, synthesizes and contextualizes the influx of thousands of sensory inputs ie. sounds, words, tastes, and visual variables to make data and physics driven risk-based survival decisions in real-time. Such technologies that mimic the brains networked, data-centric design enabling schema evolution without breaking code, and decoupling data from applications etc. exists today through complex data fabric, data mesh, and graph-like data architectures that are designed to accelerate data integration & application development against any number of business goals. However, many of these platforms were built with the primary purpose to accelerate the delivery of operational outcomes, and often lack an analytics layer, another aspect which Digital Threads aim to heighten beyond the rapid acquisition and contextualization of information across OT, IT, and IoT data sources.
In many cases analytics is typically the final layer for industrial companies to tackle – a topic that generally ties into some asset maintenance or planning business process related use case. For example, some firms are trying to optimize the performance of various assets, such as, pumps, blowers or transformers, depending on the industry, in order to prolong the lives of these degrading devices or predict why assets might break down while tying in prescriptive remediation steps for maintenance crews to action on remotely from a mobile device, including: parts ordering in SAP or asset health monitoring in real-time.
Historical performance information collected from sensors or SCADA data sources might also be analyzed to better understand flaws in current infrastructure or component designs -- information that can be used to enhance future asset construction parameters and deliverables ie. designing and building future assets that have a smaller overall GHG footprint that satisfies specific IPCC regulatory mandates. In other scenarios, analytics might be used to optimize production floor processes from. In any event, all of these use cases require companies to go beyond threading standard AI/ML models on top of data driven industrial processes, which are sufficient for maintaining and monitoring assets during steady state scenarios, but aren’t so effective amid unpredictable or black swan events. Modern organizations are looking to digitally thread AI/ML or data driven approaches for asset monitoring and maintenance activities with physics informed simulation and analytics/neural network capabilities, which help enhance predictive outcomes during non-steady state scenarios.
Physics based modelling and simulation allows organizations to seamlessly create 1D, 2D, 3D and dynamic virtual and physical representations of assets like hydro turbines, or complex connected systems like substations with the ability to superimpose physical forces, such as, electromagnetism, CFD etc. and material properties on top of these asset renderings. Once an asset or system representation has been created using a simple drag and drop GUI or developed by a data scientist in FORTRAN or C++, end users can run thousands of high fidelity system simulations approximated by Reduced Order Models (ROMs) in near-real time against any number of what if scenarios that might involve trying to understand, for example, why a transformer might trip or why, and when a specific substation component might fail in the future. Modern companies are combining data driven models with physics informed numerical methods and neural networks to become more effective at predicting failures, optimizing system performance, or enhancing the future design of assets.
Submit Name & Email to download the PDF White Paper
