‘Digital Twins’ drawing attention in multiple sectors

Long before today’s onset of Digital Twins (a virtual representation that serves as the real-time digital counterpart of a physical object or process), supporting simulation technologies made the adoption possible. These technologies are used in manufacturing, marketing and strategy, and the generated science will facilitate the widespread usage.

Three main types of simulation methodologies aide the development of Digital Twins:

1. Discrete-event simulation and manufacturing. Discrete manufacturing (as opposed to process manufacturing) can be modeled as queueing systems with each workstation considered a subsystem that processes work that is either raw material or work in progress. The methodology comprises modeling a discrete event and an activity. An inherent variance in the processing at each workstation and a probability distribution is assumed and simulated for the arrival of material at each of these workstations. The number of finished products produced per hour is the throughput level of the plant. Multiple products can be accommodated.

Metrics for work-in-process (the number of semi-finished products waiting in the queue at each of the workstations) and cycle time (time measured from when the raw material enters the production process until the time when the finished product leaves the process) is computed at the plant level. Cycle time is an indicator of the innovation time necessary to get the product to market. Work in progress is indicative of the working capital required. Little’s Law [1] describes the relationship between cycle time, work-in-process and throughput.

2. System dynamics and marketing. The framework involves a system that dynamically evolves over time. Here, the marketing function is modeled as a set of differential equations. On a computer, however, they are represented as difference equations that are solved using standard Runge-Kutta methods. These difference equations take parameters as values to simulate the marketing process. The variable of interest is usually sales or profits.

The model is represented as stocks and flows with external parameters along with other model variables and constants. Units flow into each stock, and the stock either gets depleted or replenished. The system accounts for and incorporates positive and negative feedback loops and time delays. Multiple products can be accommodated. The methodology began and took root at MIT with Jay Forrester and his colleagues in the 1950s.

3. Agent-based modeling and strategy. Agent-based modeling can be used to represent companies as agents. Each of these agents interacts with other agents and the environment comprising of customers and competitors. These agents are represented as objects having an organizational form. During each time step, the company changes its form to either adapt, imitate or innovate depending on the feedback received from the environment. A fitness landscape provides economic rent. Thus, the firms traverse this landscape that is characterized by two parameters: N, the size of the organizational form, and K, the dependency within the organizational form that determines the ruggedness as modeled in Levinthal [2]. The model illuminates and helps us understand the predictors of firm performance. 

Adoption and Future Use

Digital Twins have recently gained wide publicity with the acquisition of Gamma by the Boston Consulting Group. McKinsey & Company has announced the formation of a Strategy Analytics Center (STAC), which has the potential to accelerate the adoption of the technology within client organizations. The biggest advantage possibly arises from building a virtual counterpart to the physical entity for experimentation as a lab where the impact of different strategies can be tried out while incorporating advanced analytics.

References

1. Little, J.D., 1961, “A Proof for the Queuing Formula: L=λW,” Operations Research, Vol. 9, No. 3, pp. 383-387.
2. Levinthal, D.A., 1997, “Adaptation on Rugged Landscapes,” Management Science, Vol. 43, No. 7, pp. 934-950.