The essential guide to digital twins in manufacturing

Getting started with digital twins in manufacturing

When a major storm is blowing into your region, you can track its progress—real and projected—on your favorite weather-tracking app. What if you had the same visibility into your facility’s production process? What if you could look at a screen and see real-time status of the factory floor, along with highly accurate forecasts?

This isn’t a sci-fi thought experiment. That level of visibility is possible today, thanks to the capabilities of digital twins. Broadly, digital twins make it possible for manufacturers to use (rather than simply wrangle) the data produced by the organization.

When data is visualized in a meaningful way, manufacturing companies can glean data-driven insights, make informed decisions faster, and seize opportunities to stay ahead of their competition.

In this guide, we’ll explore what digital twins are, how they can advance manufacturing operations, and what you need to do to get your organization on the path to building and using a digital twin.

What are digital twins?

Digital twins are virtual replicas of physical objects or systems. Fueled by data, they’re a digital version of the physical system or thing they represent and they make it possible to simulate, optimize, analyze, and operate beyond what's possible in the current, purely physical environment.

Your weather app, for example, pulls temperature, humidity, and air quality data from sensors. It pulls in data from satellites. It taps historic data to offer context for real-time data and uses algorithms to predict what will happen next—and offer handy visualizations of that forecast—based on all of these data sources and more.

In a manufacturing context, digital twins operate the same way, pulling in data from various sources, sending that data through a variety of algorithms, and outputting a digital version of whatever they’re built to recreate.

Start building a digital twin in 4 steps.

Keep in mind: Your data doesn’t need to be fully prepared to begin building your digital twin (we’ll explain why shortly).

Getting started

1. Define the problem & outcome

While your data is being prepared, you can begin to identify the best problems to solve. We'll discuss in more detail later. We like to identify potential problem candidates via a ride-along with our clients. In a typical engagement, we might spend 48 hours with our client organization, following various employees and leaders through their normal day, observing their work and how they do it. Then we compare notes and discuss systems and structures that might benefit from digital twin technology.

We'll also identify likely optimizations you’ll be able to achieve. Here, we aim for realistic estimation. In our work with a provider of services for the railway industry, for example, we estimated that using digital twins to assess equipment and replicate the company’s physical premises could cut total service time by about a quarter. From there, it’s easy to extrapolate to the savings the digital twins would yield after a year, five years, and 10 years.

This is a great way to prioritize digital twin efforts: you might identify three or four opportunities to build digital twins; the one that can deliver the greatest ROI is likely the one to build first.

2. Gather your data

This is one reason data preparation work can happen in parallel to digital twin development. Once you’ve identified the best digital twin to build, you may find that you need new data sources. E.g., if you determine that temperature is a key metric to analyze in the machines in your factories but you don’t yet have temperature-tracking sensors on your machines or in your facility, you’ll need to invest in temperature sensors as step one.

3. Start small & iterate

This is a crucial step for ensuring speed to value. By starting with the smallest possible functioning version of a digital twin, you invest the minimum amount of money possible to get to a functioning model. Once you have that, you can start observing its performance in real-world conditions, learning its strengths and weaknesses, and modifying it to improve.

Note, however, that there is a key difference here from the typical “start small and iterate” step for most digital products.

With digital twins, accuracy is everything—that’s where the “twin” comes from. There is no value in building a “quick and dirty” version of a digital replica that does not incorporate accurate data because you cannot assess the value of a tool that doesn’t deliver accurate outputs.

In building digital twins, then, the goal is to build the smallest possible version of a digital twin that includes accurate data so you can start seeing its performance in the real world and learn from that as quickly as possible.

4. Analyze performance data

This is the fun part! As your digital twin makes your facility or a system within it more visible, you can start identifying areas of improvement—inefficiencies you never spotted, hidden redundancies, opportunities for sustainability, etc.

What kinds of problems can digital twin technology solve?

Digital twins are powerful instruments; as such, they’re not the right solution in every scenario, as we hinted at above. They tend to work well for systems that have the following characteristics:

High stakes

Digital twins require significant upfront investment to build, so they’re not worth building unless there's a lot of revenue or savings on the line.

Complex

For simple and / or static problems, there are available solutions that require fewer resources to build and maintain than a digital twin.

Measurable

Data is the fuel of the digital twin, which means that the system you recreate with the digital twin must have measurable components and you must have the infrastructure to capture those measurements.

Recurring

Again, this is about return on investment: building a digital twin for a one-time event will not likely yield adequate ROI.

Pertinent

Ideally, the product or system you build a digital twin of should have some relevance to your entire organization. I.e., improving that system will benefit the entire organization.

Sufficient & complete

Sufficient data — from software, databases, and IoT (internet of things) sensors — is crucial to create an accurate digital version that reflects its real-world counterpoint. The system you recreate digitally should be self-contained. If it isn’t, it will be much more difficult to build an accurate digital representation.

Keep in mind: Some manufacturing problems aren't suited for digital twins. One-offs, lower stakes issues, simpler or static problems, and those without measurable parameters can often be solved with less resource-intensive solutions.

Blog

Preparing for a digital twin strategy: considerations for manufacturers

Read the story

The use of digital twins in manufacturing

While digital twins are incredibly versatile, manufacturers tend to use them in these six ways:

System prediction
System simulation
Physical asset interoperability
Maintenance
System visualization
Product design simulation

Below are two examples of how manufacturers use digital twins.

Example 1: Absolut Digital twin of production lines for better, cheaper batch runs

The Absolut Company, makers of Absolut vodka, has digitalized and automated many parts of its operations. Part of that digitalization includes a digital twin of its production lines. The company tests batch runs of physical prototypes of new products digitally before launching them in the real world.

The simulation is validating that everything works as intended—and, when something doesn’t, they’re able to make changes before firing up the production line. Doing this in a virtual environment first saves time and money and ultimately leads to improving product quality more quickly and more consistently.

Example 2: Industrial blade manufacturer Digital twins to spot efficiency opportunities

A mid-sized industrial manufacturer works with food producers around the world to service and maintain the equipment it provides. Its operations model is circular, meaning parts are constantly in transit between the client site, manufacturing facility, and maintenance centers due to the regular upkeep required by its equipment. To enhance operational efficiency and optimize the movement of these parts, the company began exploring new approaches

They wondered whether shipping parts in different batch sizes could improve or hinder efficiency and considered whether rearranging the factory floor could reduce energy consumption, as electric forklifts transported parts. By creating digital twins of both its shipping operations and factory floor, the manufacturer could test these scenarios and gather valuable insights for continuous improvement of operations and reinvest the gains into developing digital twins for other areas of the organization.

Preparing your data for a digital twin

Manufacturers produce vast amounts of data from various sources, including software platforms and digitized factory components. A digital twin is a model built from this data to recreate specific aspects of your organization. However, just as you can’t build physical equipment without the right components, you can’t build a digital twin solution without properly prepared data inputs.

To prepare your data, you need to establish a stable data infrastructure, develop a data ingestion policy, implement a data cleaning policy, and ensure robust data governance.

Stable data infrastructure

In a digitalized facility, data is generated from various sources like sensors and ERP systems. For a digital twin application, this data needs to be accessible through a single source of truth. This involves setting up a storage solution and maintaining it to ensure continuous access and accuracy.

Data ingestion policy

Data is constantly generated and may include external sources. All this data must flow into your data repository; if it doesn’t, your digital twin will be based on outdated data and will provide inaccurate information.

Data cleaning policy

Raw data often comes in various formats and may include duplicates. To build an accurate digital twin model, data must be cleaned — deduplicated, organized, and standardized — to ensure it integrates smoothly into the virtual model.

Data governance

Effective data governance ensures ongoing quality management. This includes ensuring data availability, consistency, and compliance, as well as setting access levels for users like viewers, editors, and administrators.

In essence, the work of preparing your data to fuel a digital twin is the work of transitioning from simply having data (created by past technology investments) to actively using it to make your organization more efficient, effective, and competitive.

Enhance your data readiness

Expert support for digital twins

If you do not currently have in-house staff with the capacity or experience to do the work of preparing your data, you can bring in an outside consultant.

TXI offers a Data Maturity Accelerator engagement, which is specifically designed to assess an organization’s current level of data maturity and provide a roadmap for getting to a place where data can be used to fuel Industry 4.0 data transformation solutions, like digital twins and AI.

Once you’ve done this foundational data work, it’s time to build your digital twin.

Evaluate your data maturity

Use case

Improve production visibility in a manufacturing organization with a digital twin

Production process visibility is among the top challenges manufacturers face today. Building a digital twin of production lines can help increase visibility, facilitate decision-making, and even make it possible to test alternative production mechanisms (as in the Absolut example above).

We’ll call the organization in question Manufacturer X. During the ride-along with the partners they hired, they identify four potential problems to solve with a digital twin:

Demand forecasting
Inventory management
Overall flexibility of the manufacturing process
Real-time visibility of the factory floor

Of these, they determined creating a digital twin to replicate the factory floor in real time would deliver the highest ROI, in part because much of their equipment already had digital components they tracked via software. In addition to existing data, however, they determined they needed sensors to track temperature and humidity in the facility as a whole.

They worked with their partner to build a system visualization digital twin of their factory floor, with the goal of also using it for system simulation to eventually achieve predictive maintenance.

On the first day the digital twin was operational, Manufacturer X saw that the factory floor had a handful of locations where the humidity was consistently higher than average. They pulled maintenance reports for facility equipment and saw that the asset performance of the machines in the high-humidity areas experienced equipment failures and disruptions—and therefore caused unplanned downtime—more frequently than average.

Right away, the team had a clear pathway to reduce maintenance costs and unexpected downtime—a win that wouldn’t have been possible without a digital twin.

Learn more

Stay competitive in a changing market with digital twins

Digital twins are among the technologies powering the leaders of Industry 4.0. Built on the data manufacturers produce, they facilitate decision making in high-stakes environments by providing real-time visibility into complex systems.

Ready to see what digital twins might make possible at your organization?

Let’s talk