Data-Driven Simulation for Production Balancing and Optimization: a Case Study in the Fashion Luxury Industry

Introduction

The objective of this paper was to propose a data-driven simulation model for production balancing and optimization in the leather luxury accessories industry. As widely reported in the literature, this sector is characterized by a highly fragmented supply chain (SC), with orders frequently rescheduled and high-quality standards to be met. In this context, frequent changes in the production mix had to be managed, often requiring the re-optimization or even re-design of production flows.

Despite this, standard production layouts and SC configurations could be found among different suppliers and brands. As a consequence, a common framework could be defined that would be able to represent the majority of the industrial case studies belonging to this sector.

In particular, the present work reports the preliminary results of a complex study aiming to propose a unique framework for modeling different actors of the luxury accessories SC (i.e., leather goods shops, maneuvers, and shoe joiners) by using a data-driven object-oriented simulation approach.

The main barriers to industrial applications of traditional discrete-event simulation models, widely studied in the literature to solve production-related issues, refer to the fact that they do not allow real-time support for business decisions in dynamic contexts due to the time-consuming activities needed to re-align parameters to changing environments.

A data-driven approach overcomes these limitations, giving the possibility to easily update input and quickly rebuild the model itself without any changes in the modeling. As a first real-case scenario, the proposed model has been validated within a shoe joiner.

In order to realize the proposed data-driven simulation model, AnyLogic has been identified as the tool for modeling and analyzing the entire SC due to the flexibility of the simulation approach and the Object Oriented (OO) architecture, which allows it to replicate the different configurations of the suppliers belonging to the fashion SC.

Simulation model

Production areas, each representing a sub-set of production processes (e.g., cutting, assembly, and packing), were connected by loading and unloading points, typically called "bays". Production could be managed as job shops or lines, according to the specific processes included in each production area.

Production phases usually alternated with transport phases: at the exit of a production area, articles were placed in the loading bay to be picked up and taken to the unloading bay of the next production area, as well as internal material flows that had to be managed by picking up material from one workstation and taking it to the next one within the same production area.

By analyzing the common characteristics of the different contexts, a generic framework for simulation of the leather luxury accessories industries has been defined, as reported in Figure 1.

The boundaries of the proposed framework could be summarized as follows:

  • The production layout was a job shop or a line with a small batch flow.
  • Items were processed at workstations or production areas.
  • The logic of the queues between different workstations, or production areas, was First In, First Out (FIFO).
  • A transport system moved items between production areas, placing them inside specific containers before being handled by workers, conveyor belts, or AGVs.

Example of objects configuration in the proposed model
Example of objects configuration in the proposed model

After theorizing the conceptual framework, it was possible to detail the elements of the model. First, every object had to be able to interface with the others. According to this, the enter and exit functions would be modeled in a parametric way: this had been done using "wireless" connections in the palettes of AnyLogic simulation software. Secondly, every process chart had to include incoming queues and processing times.

Furthermore, the objects "production area" and "bay" would be modeled in a different way. The “production area” object would include the drop of items from the container received as input, the movements of items between different nodes, the jobs directly processed by workers, and the pick of items into a different container to be sent to the next step.

On the other hand, the modeling of “bay” would include only the transport phase since the content and its container would never be separated. For this reason, the "transporter" object would also be modeled as an AGV, a worker, or a conveyor belt. As an example, a bay object with AGV is shown in Figure 2.

Example of a bay object with AGV
Example of a bay object with AGV

Results

The results of this work demonstrated that the proposed framework could be easily adapted to a shoe joiner using the pre-configured objects. Scenarios had been carried out based on what-if analyses regarding productivity, resources saturation, and bottlenecks.

This tool could therefore be used as a valuable support for production planning, scheduling optimization, workload management, and production cycle balancing.

Related posts