Designing a digital twin for a mixed model stochastic assembly line for the reduction of cycle time

Abstract

The shift from traditional manufacturing techniques to the fourth industrial revolution era has brought about significant changes to the manufacturing landscape. One such noticeable change is the shift from mass production to customised production. This shift has emanated from the demand for product variety in today’s market, implying that assembly lines, which are the backbone of a production line, must now be re-designed to produce multiple variants of a product. However, this accentuates the system’s complexity and has therefore created a need for a system that can perform real-time monitoring and fault-finding in order to avoid the bottlenecks that are caused by an increased cycle time. Cycle time is a key factor in assembly line balancing, as it influences efficiency and cost of an assembly line and as such cycle time reduction is seen as a critical element in the optimised operation of an assembly line. Currently, most of the techniques used for cycle time reduction in an assembly line revolve around heuristics approaches and simulation-based techniques. The drawback of these approaches, as postulated in the literature, is that there is limited research on using real-time data as inputs, and they are mostly only suitable for a limited set of inputs. This is not suited for the stochastic nature of the modern-day assembly line. The industry 4.0 era has brought about many new strategies to ensure that production processes are customised to satisfy individual customer demands. Among the new strategies developed to improve production efficiency in the manufacturing arena is the digital transformation of the manufacturing line. Two of the popular tools that are used for implementing this digital transformation are Digital Twins (DTs) and Digital Shadows (DSs). Unlike heuristics and simulation-based techniques, DTs can utilise real-time data as inputs and therefore provide real-time visualisation and prediction of several what-if scenarios. This research looks at the design of a real-time data-driven Digital Twin (using MATLAB/SIMULINK) for a Mixed Model Assembly Line Balancing Problem (MALBP), with focus on cycle time reduction. A case study of a water bottling plant operating at the Central University of Technology (CUT) capable of producing multiple variants of water bottles will be used in this research. The Digital Twin will be synchronised with the plant for real-time monitoring of the process flow with the aim of reducing the cycle time and optimising production. The results of this research study highlight the impact of real-time data and using Digital Twins in a Mixed Model Stochastic Straight-type Assembly Line, with the cycle time reduced by 19% on average, compared to when using the Digital Shadow.