Predictive Digital Twins and Operational Planning: A Best-Practice Methodology

Which continuous improvement initiatives should you prioritise? How many operators do you need to balance cost-per-part with maximum throughput? Answering questions like these involves synthesising complex data from dynamic cyber-physical systems, which is why predictive digital twins are an increasingly important part of effective operational planning.

But what’s the best way to implement predictive digital twins so you get answers to key business questions and maximise return on your digital investments? In this article, I outline Lanner’s best-practice approach, which has been honed over more than 35 years of working with companies across industries.

A best-practice predictive digital twin implementation methodology

Building an effective predictive digital twin is something of an art and a science. The way you create it determines the quality of the result you get – in terms of output and the ability to experiment with different ‘what-if’ scenarios.

This means it’s not just about using software to model a process. You need a methodology that combines the people, processes and technologies that are key to achieving a successful outcome.

This 5-phase methodology is proven to do that. Let’s look at each phase in more detail.

Phase 1: Scoping study

Starting with a scoping study is a critical success factor. It can be tempting to open your simulation software and start modeling. But this phase is designed to bring out all the business questions the predictive digital twin needs to answer. It involves interviews with key stakeholders to understand short- and long-term goals related to those questions, as well as the objectives for the predictive digital twin itself.

This then determines what processes need to be modelled, what level of detail the model needs to reflect, what data is required and what reports will look like. It also addresses future uses for the predictive digital twin, looking at how it should be reconfigurable and reusable for future requirements.

During this phase, we often get feedback saying that, through the detailed scoping exercise, people learned more about their operational processes than they ever did before – and this generates real excitement about the benefits the predictive digital twin will bring.

Phase 2: Model development

Once the scope is fully agreed, model development can begin. This involves gathering and preparing data as well as creating the predictive digital twin in software like WITNESS Horizon.

For example, we developed a predictive digital twin for French retail giant Carrefour, which was looking to optimise distribution centre operations so as to respond to demand fluctuation while ensuring high levels of service. The model drew in approximately 20,000 elements, including 17,000 variables, 260 kilometres of conveyors for carrying parts, over 100 people, over 50 vehicles and 350 tracks (representing several kilometres of cart circuits).

All this data fed into a simulation model built in WITNESS, with the model controlled by an Excel interface. That way, anyone familiar with Excel could use the model, because running scenarios was as simple as changing numbers in a spreadsheet. A new scenario only took a few seconds to create even though it contained several hundred items of data.

Phase 3: Test and validate the simulation model

Once the model is developed to a certain point, it’s time to test and validate it – to ensure the results are trustworthy and that it accurately answers the questions defined in the scoping study. Rigorous testing gives confidence in the model itself, and that the data is feeding it effectively.

There are many different ways of validating the model. One involves using established data sources to compare results. North West Air Ambulance used a predictive digital twin built in WITNESS to understand operational requirements for providing emergency healthcare at night. They validated the simulation model using data from the Helicopter Emergency Medical Service database.

Phase 4: Plan and execute experimentation

Once the model is fully tested and validated, the experimentation begins. This is the phase where you run ‘what-if’ scenarios to understand the effect of different decisions and trade-offs. WITNESS, for instance, has advanced experimentation and optimisation functionality that simplifies this process and the process of interpreting results, so you gain the deep insight you need.

TRW Automotive’s experience is a great example. They used WITNESS to create a predictive digital twin of supply chain operations, giving them a dynamic simulation of material flows, storage capacity, handling equipment and personnel. It mapped different equipment, all transfer points, truck capacity and constraints, with a view to eliminating delivery points across the supply chain and reducing the number of forklifts used in production. The model also simulated handlers’ work on production lines to identify inefficiencies.

By experimenting with different scenarios, TRW removed bottlenecks and optimised warehouse and supply chain operations, leading to annual savings of nearly £100,000.

Phase 5: Deliverable handover and training

Predictive digital twins created using this methodology are unique because they have longevity. As a result, they don’t just enable experimentation with immediate business questions, they also provide a foundation for ongoing evidence-based decision making that maximises efficiency and capex ROI. Once the predictive digital twin is handed over to operations, it can become embedded in operational planning processes.

This was the experience at Nissan, where they built a culture of using predictive digital twins to answer both capex and opex questions. A few examples are:

• Successfully identifying process improvements to double powertrain production rates
• De-risking the leak testing process for battery trees
• Minimising pallet requirements without affecting throughput, saving £22,000
• Reducing bottlenecks in the bumper paint facility by determining the optimum storage scenario
• Determining whether a new paint guard films station was required to meet planned demand, resulting in £25,000 savings on capital expenditure

As Martin Perkins, Industrial Engineer at Nissan Motor Manufacturing UK, said: “We’re quite a lean department, and working with Lanner and WITNESS has helped us develop models which are now used regularly to make business decisions across Nissan in the UK. Thanks to the modelling, we’ve been able to implement steady improvements in our processes, and simulation has become a key part of Nissan’s adoption of Industrial IoT and smart technology.”

Want to learn more?
Contact us today to discuss, or learn more about how predictive simulation and digital twins can help you by registering for our upcoming free webinar:
Operational Planning Using Predictive Simulation