Covid-19 created a lot of supply chain disruption, demanding fast countermeasures to the resulting supply shortages, delivery delays, or demand fluctuation triggered by the chaotic business environment, price fluctuations, and policy changes on the healthcare, financial, and trade fronts. As the environment changed from stable to fluctuating, just-in-time gave way to just-in-case, prioritizing the resilience and viability of the supply chain at the expense of the bottom line. Given today's complexity of the supply chains, corporations, more frequently than you can imagine, rely on legacy technologies, providing a fragmented view, relying on delayed information, and lacking fast decision-making, which is so needed for adapting to the new normal, to be almost impossible.
We need a smart supply chain providing end-to-end real-time visibility to empower the intelligence, agility, and adaptability required to perform adequately in the always-changing condition of the new business environment. The digital twin (DT) concept satisfies these requirements. A Digital twin is a virtual representation of physical objects, processes, or combinations of both which is connected to the data layer enabling real-time monitoring and predictive analytics. We explore how this concept can be applied to the supply chain to the benefit of business incumbents, government agencies, and people consuming the goods and services of these entities.
Digital Twin
The digital twin concept was proposed 20 years ago by Dr. Michael Grieves. This concept has its origins in aerospace engineering and was designed to enhance product lifecycle management. [1] The concept is composed of two key elements: a real object and a virtual representation connected through the information exchange. With the updated data from the physical object, the virtual object, in addition to showing the actual state of the physical object, can predict its performance in the future, simulating its behavior over time and under various conditions. Such concepts took off in the realm of physical assets, significantly improving the efficiency of their operation. Examples are the aircraft engine industry, which, through digital twin simulation, was able to increase service time without incidents due to proactive maintenance, smart buildings that regulate and adapt energy consumption, or even smart stadiums pioneered in the Qatar soccer world cup which manage their operations through digital twins, reducing the cost of operations and reaching sustainability objectives. [2]
With time, the concepts broadened their applications and went beyond aerospace engineering and physical assets and became multidisciplinary. According to Gartner, digital twin is comprised of 4 essential components[3]:
- Entity metadata refers to information that describes an object in detail, such as its physical components, assembly, behavior, and specifications.
- Generated data is produced by IoT sensors and includes time-series data as well as other contextual data used in analytical models.
- Analytical models are software algorithms that process generated data to increase situational awareness and generate events.
- Software components consist of application logic, visualization tools, and other functionalities that act based on the events produced by analytical models.
- Grieves, Michael. "Origins of the digital twin concept." Florida Institute of Technology 8 (2016).
How World Cup Qatar Venues Are Utilizing a Digital Twin Solution to Centralize Operations Like Climate Control and Security. , 13 Dec. 2022,
“Use 4 Building Blocks for Successful Digital Twin Design.” Gartner, http://www.gartner.com/en/documents/3986128.
Use-cases
Even though the digital supply chain twin is a novel concept, there are already several successful use cases. Lu Wand described the implementation case at JD.Com.[4]
In today's retail landscape, several emerging trends, including retail decentralization, community group buying, social shopping, and live commerce, drive increased interaction and integration between retail and manufacturing. This makes the supply chain structure complex. JD.COM, one of the leading players in this space, operates an extensive network of 41 "Asia No. 1" logistics parks in China, boasting over 1300 warehouses and more than 9 million self-operated stock-keeping units (SKUs). The company's product offerings span various categories, including consumer packaged goods, information appliances, home appliances, clothing, fresh food, books, and even automobiles. For seamless transportation, JD.COM leverages multiple transportation modes, including land transportation and shipping. By harnessing digital and intelligent technologies and adopting multichannel models, JD.COM extends its supply chain planning and operations to upstream and downstream activities. However, the increasingly complex nature of supply chain management means that more than traditional planning methods and algorithms is required. The challenge was exacerbated by the COVID-19 outbreak, causing demand spikes for various goods categories such as masks, alcohol, household cleaning products, and others. This demand was hard to manage due to transportation network interruptions, shortages of labor and raw materials, and logistical disruptions, resulting in products being out of stock.
JD.com leadership built an end-to-end digital supply chain twin to handle such a challenge. The data synchronized between the physical supply chain and virtual replica included network configuration, procurement system, transfer system, and fulfillment system. Optimization and simulation algorithms were orchestrated together, gaining intelligent insights. These insights supported various decisions, from long-term strategies to mid-term plans and short-term operations.
To handle sporadic distribution centers and warehouses' unavailability, the supply chain should be reconfigured so that orders can be replenished throughout alternative distribution centers. Such a decision often implies adverse effects on the alternative distribution center performance indicators, such as order fulfillment and transportation cost. To mitigate the negative impact, JD.com team validated alternative solutions through digital twin simulation to find the optimal one that has minimum negative impact due to reconfiguration and was able to handle affected orders. The team reduced the response time for the disruptions twice. Before digital twin, the team spent several days reconfiguring the supply chain for 200,000 SKUs, which now takes less than an hour.[5]
4. Wang, Lu, et al. "Digital twin-driven smart supply chain." Frontiers of Engineering Management 9.1 (2022): 56-70.
5. Wang, Lu, et al. "Digital twin-driven smart supply chain." Frontiers of Engineering Management 9.1 (2022): 56-70.
Another use case of digital supply chain twin came from Michelin [6], the second largest tire manufacturer in the world. They have a complex global supply chain structure with eight levels, including three manufacturing and supplier levels, distribution centers, warehouses with 3,5 million square meters, and two types of customers. Logistics costs are close to 2 Billion euros, and 4 billion in inventory. Michelin has several levels of digital twins depending on the horizon and usage frequency; they discern the following:
- Strategic/tactical horizon, providing insights on the strategy questions.
- Operational/Tactical for flow optimization between plants, and dynamic flow analysis mitigates various adverse effects such as bullwhip or ripple effects.
- Operation/execution - dynamic simulation of plant or warehouse flow.
For example, strategic digital twin helps leadership to update twice a year the sourcing strategy for China factories, maximizing profit over 5 years. China is the biggest market in terms of original equipment and replacement. Sales volume is 20 million per year; the number of tire types is 1700 and it is growing to have solid differentiation. Part of these products is imported from Canada, North America, Europe, or Thailand; the rest is produced in China. The issue is how to balance the sourcing and production between various locations to maximize profit.
The built digital twin model had various parameters such as capacities, the demand curve, demand uncertainties (variability), operations costs, cost of moving tire production from one plant to another, and product life cycle time. The output takes into account not only profit but also service level, total cost, factory utilization, and inventory level. To find the optimal strategy, digital twin executed 80 000 scenarios! The identified optimal scenario helped Michelin gain tremendous results, optimizing global profit by more than 5%, customs costs by 60%; transportation costs by 60%, and in-transit stock costs by 60%.
