How pervasive digital twins are x-tending enterprise intelligence

We have discovered one of the biggest advantages of having worked as management advisors with client companies across a wide range of the global industry spectrum has been the opportunity to connect the “strategic-dots” that aren’t obvious, yet are ones that have been hiding in plain sight all along. All it required was a willingness to take off the colored glasses of viewpoints emanating from within the silo of one industry and use them to cross-pollinate those emanating from other ones. The resulting insights have often surprised us.

The industry uses many forms and definitions for digital twin depictions. However, in our experience often the real point gets lost in the din of growing addressable markets and technology deployment discussions

A Digital Twin is an organised collection of physics-based methods and advanced analytics that is used to model the present state of every asset in a Digital Industrial world. Delivering on the vision of autonomous connected ecosystems, where data is connected seamlessly through the convergence of the physical and digital world, and intelligence is built into all processes.

At its core, the digital twin consists of sophisticated models or a system of models based on deep domain knowledge of specific industrial assets. The twin is informed by a massive amount of design, manufacturing, inspection, repair, online sensor, and operational data. It employs a collection of high-fidelity, computational physics-based models and advanced analytics to forecast the health and performance of operating assets over their lifetime.

By incorporating multi-physics simulation, data analytics, and machine learning capabilities, digital twins are able to demonstrate the impact of design changes, usage scenarios, environmental conditions, and other endless variables – eliminating the need for physical prototypes, reducing development time, and improving quality of the finalised product or process.

Digital twins take different names based on specific use-cases they were created to serve, and often take on vendor marketed brands.

A Product Digital Twin is created as early as the definition and design stage of a planned product. This allows engineers to simulate and validate product properties depending on the respective requirements. This “product digital twin” provides a virtual-physical connection that lets one analyse how a product performs under various conditions, and make adjustments in the virtual world to ensure that the next physical product will perform exactly as planned in the field.

The Production Digital Twin involves every aspect, from the machines and plant controllers to entire production lines in the virtual environment. This simulation process can be used to optimise production in advance with controller code generation and virtual commissioning. As a result, sources of error or failure can be identified and prevented before actual operation begins.

The Performance Digital Twin is constantly fed with operational data from products or the production plant. This allows information like status data from machines and energy consumption data from manufacturing systems to be constantly monitored. The performance digital twin captures this data from products and plants in operation and analyses it to provide actionable insight for informed decision-making. In turn, this makes it possible to perform predictive maintenance to prevent downtime and optimise energy consumption.

An Operational Twin is the plant’s as-is, transactional, and time-series information. In other words, it is the pieces of a digital twin specifically needed for operations — it provides the right workers with the data they need, in the right context, when they need it.

The Lifing Digital Twin is able to assess each asset within the plant and how that asset will age relative to its operation and exposure. Fatigue, stress, oxidation, and other phenomena are predictable using this digital twin and help optimise the maintenance vs. mission reliability of each asset as well as the entire operating system.

The Anomaly Digital Twin uses physics and data-based prognostic models to detect faults for improved asset failure mode management and reduced unplanned downtime. Using fusion techniques with Lifing models, anomaly models can increase the accuracy of production machine life curves and further personalise maintenance needs.

The Thermal Digital Twin determines thermal efficiency, plant capacity, and emissions prediction as well as simulation of all of parameters that can affect these outcomes. The technology can be used on all thermal cycles, from combined cycle and fossil to cogeneration and district heating. This twin often delivers the best value once it is integrated to the vendor’s own asset models.

The Transient Digital Twin can simulate the plant’s ability to react to changes in environmental or control shifts. This would include startup, ramp rate, minimum generation, and regulation performance. Deployed fully, this twin will provide insights into speed, stability, emissions, and stresses as well as predict the limitations of the plant with configuration and operational changes.

To ensure accurate modelling over the entire lifetime of a product or its production, digital twins use data from sensors installed on physical objects to determine the objects’ real-time performance, operating conditions, and changes over time. Using this data, the digital twin evolves and continuously updates to reflect any change to the physical counterpart throughout the product lifecycle, creating a closed-loop of feedback in a virtual environment that enables companies to continuously optimise their products, production, and performance at minimal cost.

Included in the digital twin models are all necessary aspects of the physical asset or larger system including thermal, mechanical, electrical, chemical, fluid dynamic, material, lifetime, economic, and statistical. These models also accurately represent the plant or fleet under a large number of variations related to operation — fuel mix, ambient temperature, air quality, moisture, load, weather forecast models, and market pricing. Using these digital twin models and state-of-the-art techniques of optimisation, control, and forecasting, applications can more accurately predict outcomes along different axes of availability, performance, reliability, wear and tear, flexibility, and maintainability. The models in conjunction with the sensor data give the ability to predict the plant’s performance, evaluate different scenarios, understand trade-offs, and enhance efficiency.

We Expect Digital Twins To Become Pervasive Tools For Industrial-IoT

The most obvious advantage appears to accrue for an operator of a production process.  A digital twin provides capability to balance and optimise trade-offs between important factors over which they prior had minimal visibility or control. They also allow “what-if” scenarios to be tested against business objectives creating the most informed decisions possible. From an Asset Performance Management use-case, lifing models give the customer insight into when particular modes of operation are unwittingly damaging hardware, allowing the operator to avoid these modes. In the cases where operation in that mode is required, additional sensitivity to anomaly detection models is given and recommendations are made to maximise remaining life. This additional insight into the damage mechanism allows for a transition from traditional maintenance (manual-based scheduling) to true condition-based, predictive maintenance. Additionally, this environment is integrated with business applications designed to allow plant executives, plant managers, and workers to interact with the twin in real time.

Digital Twins Are A Monetisable White Space Opportunity For Many Adjacent Industries

Beyond the obvious services revenue streams vendors expect to create from delivering digital twins, we believe many other entities stand to generate significant new monetisable value from its use.

For Plant Operators: The frequent lack of visibility into asset performance, beyond monitoring data on operating and process parameters, compels them to rely on standardised and normalised charts of equipment behavior as a guide for planning maintenance for various equipment comprising the process. The result: Unforeseen breakdowns take a toll on the process and maintenance engineering team’s time and availability for more pressing productivity and quality related goals. Higher costs for manpower, overtime, and parts ordering become the norm.

An Equipment Vendor today is tasked not only with installation and annual maintenance of their delivered product, they are being increasingly tasked with sharing the risk of operational disruptions caused by faults and suboptimal operation of them. The practice of scheduled maintenance therefore loses its relevance.

A Parts Manufacturer who bases their entire production planning on archaic wear-and-tear models built on past, historically annualised data would be tempted to use the lowest common factors to maintain inventory. Ignoring actual usage conditions, ageing parameters, and locations of specific equipment that might have a higher predilection to failure.

A Software Developer who designs process simulation software has an opportunity to incorporate the advantage of piping in real time sensor data into their static computational models. Doing so bridges the insight gap between detection (“what went wrong”), prediction (“what could lead to future anomaly/failure”), and prescription (“what’s the right course of action in the most efficient manner”).

An Industrial Equipment Insurance Provider who bases annual premiums on the strength of an annual or six-monthly equipment inspection would have a dependence on risk assessment models that do not reflect gradual, imperceptible deterioration in performance, and by implication those early warning indicators of catastrophic failure that drive the maximum payouts on claims. Having access to digital twins of their covered assets allows them to update their internal risk models and insurance pricing in near real-time, benefiting bottom lines both for themselves and their clients.

For more information on how artificial intelligence, machine learning, and process simulation capabilities are being incorporated into industrial management, read other insightful pieces by ArcInsight Partners, such as “The A.I. Frontier Of Industrial-IoT Analytics.”

Praas Chaudhuri

Praas Chaudhuri is a Silicon Valley based strategy analyst & co-founder of ArcInsight Research Partners, a strategic research & advisory group focused on digital transformation driven by analytics, algorithms & pervasive connectivity. Broad areas of research interest at ArcInsight Partners include - (1) Industrial-IoT enabled manufacturing & production process transformations. (2) Evolution of intelligent cities enabled by new mobility design, urban dynamics & renewables. (3) Emerging business trends enabled by SaaS models and new monetisation strategies. Praas Chaudhuri can be reached for analyst-briefings and for management advisory via email: [email protected] or LinkedIn: www.linkedin.com/in/praaschaudhuri

Authors Posts