The term Digital Twin emerged in 2017 as one of the main technological trends according to Gartner consultancy and its results have been demonstrated in multiple areas. This article aims to introduce the concept of Digital Twin technology, reviewing some of its main benefits and presenting a practical example. In order to develop a solution to optimize its warehouse, we co-innovate in collaboration with Frit Ravich, the leading manufacturer and distributor of external brands from well-established companies such as Mars Spain, Nestlé and Ferrero (3,000 references) that serve around 50,000 points of sale every week.

Introduction to digital twins

Broadly speaking, Digital Twins pursue the convergence between the real world and the virtual world, that is, where the processes involved in the life cycle of a good or service find their digital representation. In this virtual model everything is interconnected, and this can be understood as a source of valuable information which serves as a support for:

Generating detailed and realistic analyses
Early prediction of problems, preventing downtimes
Development of new business opportunities
Deliver better products for the future at lower costs through simulations.

Thanks to the increasingly widespread use of the Internet of Things (IoT), the costs associated with making these models have been reduced and are therefore more accessible to organizations of all sizes.

Benefits of digital twins

Digital Twin-based technology offers a wide range of benefits, the following are some of the most relevant:

Visibility: enables a better visibility into device and machine operations, as well as larger and more complex interconnected systems.
Predictive analytics: allows to interact with the model through interfaces and run queries to see what the model's behavior would be under different scenarios.
It helps to understand and explain behaviors: the model can be used as a documentation and communication tool to understand and explain the behaviors of machine collections.
System-to-system connection: the model can be used to connect with other business applications (such as ERP's), achieving improvements in the context of supply chain operations.

Application to a real case

In this section we present a fictional scenario, but similar to the one we have worked together with Frit Ravich, and that will serve to articulate the main steps and considerations that must be taken in the implementation of models based on Digital Twin.

Case presentation

Let us suppose an F&B product distributor that has a conventional warehouse to prepare and deliver orders. We want to know if it would be possible to implement a Digital Twin-based model to run simulations on the virtual design and determine which product distributions through the warehouse would be those that would minimize the number of meters required, which would result in a decrease in order preparation time.

Data collection

Considering the above context, we first proceed to the study of physical positioning and the structuring of corridors within the warehouse. The result of this first step will be an accurate map with coordinates, where each point will show a physical space available to store products. From here, the study of movement restrictions should be addressed.

Exploratory data analysis

The data obtained in the previous section are subject to a review process (volume inspection, data quality, etc.) in order to subsequently apply exploratory methods to understand the interrelationships between products: volumetric control by campaign and orders, evolutions, product detail, association rules for patterns detection, etc.

Restrictions and model creation

Taking as input the coordinate system, the routing of the plant and the historical data of orders by campaign, the virtual model at this point integrates all the elements to allow the execution of simulations. Here are also defined the graphical parameters for the user's interaction with the model, as well as the incorporation of additional restrictions associated with the products, such as those of the products (volume and weight) or special storage conditions (temperatures or storage according to packaging, among others).

Simulation

In order to proceed with the optimization, it is required to evaluate the results achieved with the configurations originally provided. For this reason, taking directly the historical data of the orders placed, they are transferred to the model to obtain the distances that have been required to make the orders without any optimization. For each of the different orders, the required distance (or cost) between the different products is obtained. From here we get the results of the current situation, which will serve as the basis for the optimization.

Optimizing results

Taking the results from the previous simulation, we determine which products have a greater and lesser impact during campaign routes in aggregate form. From this point on, the algorithm will suggest reordering along the corridors. The most requested products will end up in the corridors closest to the standard route (minimum route to be carried out to reach the exit) and the least requested in more remote positions. The result is delivered as a file with the new item relationship by product and will serve as input for the following runs.

Results

From the resulting file, a new execution of the algorithm is initiated to verify that the optimization has been made effective. It should happen that, during the campaign analyzed, the quantities of meters travelled have been decreased to complete all orders. Thus, once the execution is complete, the output file containing the details of the routes with the accumulated weights is verified and the distance has been reduced, so the number of meters required to perform all orders is now below the original scenario. The following figure contains an example for an optimized route. It can be noticed that the model accurately represents each aisle as well as the route that should be taken to reduce the number of meters required to complete the preparation of the order.

Conclusions

Digital Twins are a key part of the fourth industrial revolution, which, as seen, encompasses automation, data exchange and manufacturing technologies. Indeed, the most significant thing is the change in the paradigm, since the traditional approach to production is replaced by a set of design processes based on much more efficient virtual systems that allow to evaluate, simulate and predict features, expected returns or potential problems and failures.

Undoubtedly, there is a promising future to develop in the world of manufacturing and engineering and Digital Twins are a significant step towards achieving it.