We hear the term Digital Twin quite frequently nowadays. The term, rather the concept, was coined long back and has since evolved while staying true to its core ideology. In general terms, a Digital Twin is a cyber replica of an entity such as a machine, material, process, or person (role).
In this short article, we shall investigate the two main types of "digital twins" that we come across over the life cycle of the typical plant - the Engineering (EPC – Engineering, Procurement & Construction) phase and the Operations phase.
While the type classification applies in general to both process manufacturing and discrete manufacturing plants, our specific context is process manufacturing.
The above graphics gives an overview of the purpose, the two main versions of the digital twins, and software tools involved over the two major phases of the plant lifecycle.
Before any plant is built, engineers need to design every small component, every machine, every system, and then put it all together to define the process flow of the complete plant itself. Engineers need to simulate the performance of everything they have designed under various operating conditions to ensure functional adherence, safety, etc. This is done using various 2D, 3D CAD, CAM, CAE, and various simulations (software) tools. Collectively, these tools fall under a category commonly known as PLM (Product Life Cycle Management) suite.
The PLM suite allows engineers to design, engineer, and test a manufacturing/process plant or product before the physical prototyping/manufacturing or procurement & construction starts. There is no operations data involved as this is all done before an asset or a plant goes into its operating phase. The purpose of the PLM digital twin is to ensure that a physical asset will serve its intended function.
In the operations stage (i.e., when a plant is functioning in the real-world), the story is a little different. The concept of a digital twin takes on a whole new meaning. We can try to understand it by categorizing it into two broad classes, based on its purpose & accompanying business model.
Advancement of technology has allowed a new business model where the OEM (Original Equipment Manufacturer) can "sell" a physical product as a "service." For example, one can sell a "compression" service instead of selling the compressor. In this business model, the physical asset remains the property of the OEM or an intermediate service provider. The OEM/service provider also remains responsible for ensuring "availability" (and few other performance KPI). In short, they remain responsible for the upkeep of the asset. The user pays for the usage as per the SLA (service level agreement).
In this kind of industrial application, IoT (IIoT) plays a significant role. Usually, the OEM would embed a large array of sensors and invest in broadband connectivity to be able to monitor a large fleet of machines (sometimes small plants as well), primarily to ensure availability.
The data stream from these IoT sensors does not require to be part of the control network (process control LAN) and need not be connected to DCS/PLC. It reduces the cost per sensor up to 90% (no more cabling, junction box, marshaling panel, I/O, related engineering and lifecycle cost, etc.). This is a very major advantage and is commonly referred to as "Industrial IoT (IIoT)."
In this type of application, the digital twins are built for blending large streams of time-series data (vibration, temperature, pressure, etc.) and design data and maintenance related data, to optimize "machine" performance. This is essentially a specialized "operations decision support" application. It is common for assets that are geographically distributed and large in numbers (e.g., Wind turbines, etc.). There are large LNG plants (many compressors) and other types of plants where the operator just pays for "compression services."
Note that in industries that already have a large number of sensors for control purposes, IoT sensors are being added to replace periodic manual inspections (for corrosion, steam-trap monitoring, etc.), and manual monitoring processes. These IoT sensors may or may not be coupled with a "Servitization" business model offered by a vendor.
The day to day operation of a manufacturing plant (petroleum refinery, chemical/petrochemical plant, steel plant, cement plant, breweries, etc.) is very complex. It involves continuous coordination of various functions and activities such as planning, scheduling, inventory, production, process (physical), quality, maintenance, procurement, dispatch, etc. Consequently, it involves data from a myriad of IT and OT applications, traditionally grouped in five hierarchical layers as per "Purdue Enterprise Reference Architecture (PERA)", as shown below.
Applications in these five layers and the data thereof are mission-critical in nature and fuel the core manufacturing, supply chain, and business operations.
Data about every entity, be it a machine, material, work process, or even a person, is scattered across various applications residing in these PERA layers.
For example, even a typical motor driving a pump has associated data such as:
All of this data is spread across 5 - 7 different IT & OT applications at the very least. A typical manufacturing business has its data spread across 30 to 50 different applications. However, every entity, incident, and process are interrelated and have implications that are cross-functional in nature, impacting both upstream and downstream operations as well.
An "Operations Digital Twin (ODT)," is about converging and contextualizing these myriad data streams and their analytical derivations (calculations, rule, logic, ML, etc.). The "context" is essentially the "mental models" or paradigms of the stakeholders. In short, ODT is essentially a blend of
The key takeaways are that there are different kinds of "Digital twins." The word and the concept originated from the PLM software domain. Sometimes, the various "Digital Twins" overlap to some extent in features and functionalities. For example, one may overlay operations data and analytics on a 3D plant model (however, it has performance implications).
The graphic below shows a "Process Operations" dashboard driven by the "Operations Digital Twin" of a petrochemical process unit.
"Digital Twin" is a concept that will continue to evolve and will have huge implications on how data leveraged for driving excellence.