Data-driven discovery of manufacturing processes and performance from worker localisation

Tools

Aslan, Ayse and El-Raoui, Hanane and Hanson, Jack and Vasantha, Gokula and Quigley, John and Corney, Jonathan; (2023) Data-driven discovery of manufacturing processes and performance from worker localisation. In: The 32nd Flexible Automation and Intelligent Manufacturing International Conference (FAIM2023) : Lecture Notes in Mechanical Engineering (LNME). Lecture Notes in Mechanical Engineering . Springer, [Cham].

[thumbnail of Aslan-etal-FAIM-2023-Data-driven-discovery-of-manufacturing-processes-and-performance-from-worker-localisation]
Preview
 Text. Filename: Aslan_etal_FAIM_2023_Data_driven_discovery_of_manufacturing_processes_and_performance_from_worker_localisation.pdf
Accepted Author Manuscript
License: Strathprints license 1.0
Download (971kB)| Preview

Abstract

In complex manufacturing industries that involve the coordination of people, equipment and materials in the completion of tasks with non-canonical sequences (such as in shipbuilding) it is important to identify deviations from the planned production schedule and locate bottlenecks in order to improve efficiency. This is not an easy task as it requires data on how workers are actually performing the manufacturing activities. Ultra-wideband (UWB) tags, which are sensors that track worker movement, can be used to collect this data. Previous research has mostly focused on using this data to detect faults and anomalies and to ensure worker safety. However, this paper presents a method for using UWB data to discover process models of manufacturing activities using process mining techniques. These models allow us to identify deviations from prescribed process steps and find bottlenecks by predicting performance metrics such as the mean duration of activities. The method involves extracting event logs based on workers’ positions, defining activities based on their time spent in proximity to specific zones, and using the "Inductive Miner" algorithm to generate process models. We applied our method to a real assembly line with UWB data and were able to identify deviations from the prescribed process steps and bottlenecks in the assembly line, which indicated that the first assembly step can take twice as much time compared to other steps.

ORCID iDs

Aslan, Ayse, El-Raoui, Hanane ORCID logoORCID: https://orcid.org/0000-0002-9079-3248, Hanson, Jack, Vasantha, Gokula, Quigley, John ORCID logoORCID: https://orcid.org/0000-0002-7253-8470 and Corney, Jonathan;
CORE (COnnecting REpositories)