© maxkabakov | iStock
How You Can Make Digital Twins Work in the Real World

How You Can Make Digital Twins Work in the Real World

Nov. 27, 2019
Manufacturers need to have a specific business case and production-based purpose to deliver the full potential of digital twins.

The concept of a digital twin has been around for nearly two decades. However, it's lower-cost enabling technologies that are putting this modern approach to manufacturing within reach of more companies.

Optimizing Digital Twins in Manufacturing

By definition, a digital twin is an exact representation of a physical production system in a digital environment. This makes it possible for manufacturers to see how every phase of production contributes to or detracts from product quality.

Most commonly, digital twins are used for quality control, improvement, system diagnostics, monitoring, optimization, and prediction of production outcomes and machinery performance. Optimizing digital twins requires investments both in the machinery being monitored and in software that turns data into actionable insights.

Some manufacturers are paying to upgrade or replace their fully depreciated legacy machinery, so they can deploy machines that are equipped with sensors, their own operating systems, and IP addresses. These smart, connected machines are capable of reporting yield rates on specific production runs, providing real-time monitoring and data that enable manufacturers to gain the advances of a digital twin.

The most advanced manufacturers rely on supervised and unsupervised machine-learning algorithms to discover how they can improve quality based on the terabytes of data generated daily in a typical production environment. Both types of algorithms, which are the foundation of digital twins’ ability to deliver valuable insights, need to have process- and product-based real-time monitoring data.

Supervised machine-learning algorithms take advantage of historical trending data extracted from real-time monitoring to seek out and find the optimized solution to a specific problem, such as production scheduling. Unsupervised machine-learning algorithms excel at finding patterns in unlabeled data that come from real-time monitoring, making these algorithms perfect for root cause analysis of product quality improvements.

When introducing a digital twin strategy, manufacturers need to have a specific business case and production-based purpose if they are going to deliver its full potential. Senior management also needs to “own” the digital twin outcomes and help remove barriers to its adoption, starting at the pilot level and progressing through production.

All digital twin pilots that make it into production have very clear, well-defined metrics and key performance indicators (KPIs) that keep them focused on delivering the results they need to scale and deliver greater value.

Lessons Learned from Digital Twin Early Adopters

Digital twin early adopters tend to focus on improving product quality while also seeking to increase maintenance optimization, including two well-known reliability metrics of mean time between failures (MTBF) and mean time to repair (MTTR).

Additionally, real-time monitoring can provide contextually rich data sets that enable machine-learning algorithms to provide insights into how to achieve product quality and maintenance. Digital twins are most accurate when they combine supervised machine-learning algorithms, which take historical trending data and build models across the data points to look for patterns, with unsupervised machine-learning algorithms optimized for finding new patterns of anomalies.

Early adopters are finding that digital twins deliver valuable data and contributions in four key areas of their business.

1. Improved Efficiency

Digital twins are delivering valuable new insights into how end-to-end production processes can be made more efficient. Creating a reliable, scalable digital twin strategy requires every phase or stage of the production process to be monitored in real time. This ensures that the digital twin is an exact representation of the physical system, which enables production and quality engineering teams to test scenarios aimed at improving end-to-end production efficiency, scale, and speed.

2. Quality Management

Digital twins help manufacturers improve quality management by analyzing real-time production and quality data to discover why quality anomalies are happening. Through the resulting insights, these businesses have been able to reduce scrap rates, increase machine and plant yields, and reduce return material authorizations (RMA).

Manufacturers are increasingly using digital twins to simulate quality and conduct failure mode and effects analysis (FMEA) of new products during the design stage. This makes it possible to identify all of the possible failures in a design or manufacturing process before a single part is created, and it is invaluable in reducing scrap levels of when entirely new parts are produced.

3. Optimized MRO

Digital twins are proving to be very effective at improving the maintenance, repair, and overhaul (MRO) and asset longevity of machinery manufacturers rely on daily to deliver products. Understanding how specific combinations of material components, machinery load levels, cycle times, operator team training levels, tool calibration, and preventative maintenance activity all contribute to improving MRO is the ultimate goal manufacturers have with digital twins.

Knowing which combination of factors most and least impact asset longevity and MRO accuracy and scale is invaluable for keeping a production center at peak efficiency.

4. Cross-Group Collaboration

Digital twins enable manufacturers to achieve greater visibility and collaboration, from initial product design to production, by bringing engineering, production, sales, and marketing together as a common team. By capitalizing on the ability to first simulate and then plan changes to the production process, digital twins contribute to more efficient new product development cycles, improve product quality, and increase yield rates across diverse production systems.

As a result, design to manufacturing is now a reality for many manufacturers pursuing build-to-order, configure-to-order, and engineer-to-order product strategies.

Conclusion

The influence of digital twins is already being seen in the development of every type of next-generation product, from supersonic aircraft to next-generation industrial machinery. However, for digital twin strategies to be effective, they must be anchored in a solid business case and linked to a production-driven purpose

Those responsible for advancing digital twins from pilot to production need to be held accountable for the metrics and KPIs that reflect their contribution to a company’s business outcomes. And it’s a must-do for any senior executive to the C-level to “own” the digital twin initiative to ensure its success. All digital twin pilots that make it from pilot to production had a C-level owner who removed the barriers and roadblocks to its success. 

Louis Columbus serves as a principal at IQMS, now Delmiaworks, part of the Dassault Systèmes portfolio.

About the Author

Louis Columbus | Principal, Delmiaworks (IQMS)

Louis Columbus serves as a principal at IQMS, now Delmiaworks, part of the Dassault Systèmes portfolio.

Previous positions include product management at Ingram Cloud, product marketing at iBASEt, Plex Systems, senior analyst at AMR Research (now Gartner), marketing and business development at Cincom Systems, Ingram Micro, a SaaS start-up, and at hardware companies. Columbus is also a member of the Enterprise Irregulars.

Professional experience includes marketing, product management, sales, and industry analyst roles in the enterprise software and IT industries. His academic background includes an MBA from Pepperdine University and completion of the Strategic Marketing Management and Digital Marketing programs at the Stanford University Graduate School of Business.

Columbus teaches MBA courses in international business, global competitive strategies, international market research, and capstone courses in strategic planning and market research. He has taught at California State University, Fullerton; University of California, Irvine; Marymount University, and Webster University.