Process Mining in Logistics

Process Mining in Logistics is a joint project of the Data Science Center Eindhoven and Vanderlande industries.

Project Description

Logistics systems, such as baggage handling or warehouse automation, handle large numbers of physical objects. Following pre-defined processes, the objects move over conveyor belts and routing elements to scanners, storage units, automated and manual processing, packaging steps, and may be grouped and re-grouped with other objects in smart automated batching and de-batching steps. Thereby, a single object participates in multiple different processes and cases of these processes. This renders the design, operation, and analysis of logistics systems extremely challenging:

  • Logistics systems form a network of numerous processes, cases, and objects that interact with each other in a multi-dimensional fashion: Multiple cases synchronize at batching and de-batching steps, multiple objects may queue in high-load situations over multiple processing steps, and the routing decisions for each object depend not only on its case, but also on the load in other system parts.
  • Logistics processes easily see a million process events per day as they handle 10,000s of cases involving 100's to 1000's of processing steps per case.
  • Such volumes require that the supporting logistics systems execute these processes with very high and reliable performance, ideally in steady flows, without deviations from process design, and with enough “free capacity” to provide operational resilience in case of unforeseen load.

These characteristics not only challenge logistics operations, but also render existing process mining techniques largely inapplicable. Learn more

Vanderlande is the global market leader in baggage handling systems for airports and sorting systems for parcel and postal services, and also a leading supplier of warehouse automation solutions. The company recognizes the emerging trend of more data driven business models and addressed ‘big data’ a key topic on the technology roadmap. Therefore, under the umbrella of the Data Science Impuls program, the DSC/e and Vanderlande joined forces in a research project.

Project Objectives

The objective of the “Process Mining in Logistics” project is to develop a data-driven feedback loop for improving the design, performance, and resilience of logistics processes and systems. The central research task is to lift process mining to the multi-dimensional dynamics of logistics processes. New process mining techniques analyze the underlying event data from all relevant angles and viewpoints to provide fact-based insights into logistics processes and systems. By having thorough and fast insight into logistics and business processes, improvements can be found, predicted, and implemented at Vanderlande delivered logistics solutions. Over the course of 6 years, we build this feedback loop in three stages:

  • Descriptive analytics for providing fine-grained, fact-based insights into how processes were actually executed, which deviations occurred, for root-cause analysis of performance problems, and for identifying potentials for improving process performance. This information helps process engineers to redesign and commission high-performing material handling systems.
  • Predictive analytics that uses machine-learning for predicting the near-term behavior of logistics systems from past data, allowing to signal performance bottlenecks or potentially problematic cases. This information helps process managers and service engineers to keep tabs on their processes and react to problems earlier than before.
  • Prescriptive analytics further combines predictions about future behavior with recommendations for mitigating upcoming process bottlenecks or delays. This information helps process managers to make the right decisions and to keep their operations running smoothly.

The project runs from September 2016 until August 2022.


Insights

  • 30-10-2018. One of the core challenges of process analytics from event data is to enable an analyst to get a comprehensive understanding of the process and where problems reside. As in logistics, insights into performance is key, we developed the The Performance Spectrum as a process map for performance insights for large-scale, high-volume processes.

Technology Developed

  • The Performance Spectrum Miner is a visual analytics tool to analyze very large amounts of event data, quickly identify temporary process deviations in very large processes, locate short- and long-term performance problems as well as gradual and abrupt changes in process performance, identify the root-cause of performance problems and deviations in logistics processes occurring only under certain conditions. The platform-independent (Java), open-source (LGPL) tool is available as ProM plugin and via https://github.com/processmining-in-logistics/psm

Publications


Staff Involved

Former Staff