1. BaSaMa & HiWi
    2.  

Optimization-based Assembly and Production Line Planning: Extending the Open-Source Framework PyCAALP

Institute
Lehrstuhl für Umformtechnik und Gießereiwesen (TUM-ED)
Type
Bachelor's Thesis / Semester Thesis / Master's Thesis /
Content
theoretical /  
Description

Before a welded assembly such as an automotive exhaust or body component goes into series production, two coupled decisions must be made: the order in which the parts are joined (assembly sequence planning) and how these joining operations are distributed across the stations of a production line (production line planning). In industrial practice, both decisions are still largely made manually and based on experience, and they are usually treated sequentially. As a result, sequences that look good in isolation can lead to unbalanced lines, unnecessary stations, or long cycle times.

At the Chair of Metal Forming and Casting, we developed PyCAALP, an open-source Python framework that jointly addresses both problems. Assemblies are modeled as graphs; feasible joining sequences are found as paths through this graph; and the line layout is optimized using mixed-integer programming. A deterministic graph-reduction method keeps the problem tractable while guaranteeing the existence of feasible solutions. The approach has been published and validated on industrial welded assemblies [1, 2]. However, the current framework still relies on several simplifying assumptions: stations are assigned strictly sequentially, all parts are joined in a single linear sequence, and weld seam length is used as a proxy for actual operation time.

The objective of this thesis is to remove one of these limitations and extend PyCAALP towards realistic industrial production lines. The exact topic is chosen together with the student and scaled to the type of thesis. Possible directions include:

  • Parallel stations: duplicating bottleneck stations in the line, as is common in real production
  • Parallel sub-assemblies: pre-assembling independent groups in parallel and joining them later, instead of one linear sequence
  • Joining-process time model: replacing weld seam length with process-based times per joining technology (MAG, TIG, resistance spot welding, mechanical joining)
  • From CAD to graph: deriving part relations, tolerances and handling properties automatically from CAD data
  • Multi-objective planning: making trade-offs between investment, line balance and cycle time visible as Pareto fronts
  • Discrete-event simulation: testing planned lines under scattering process times and disturbances

In every direction, the extension is evaluated on industrial assemblies and on openly available assembly datasets, contributing to an open benchmark for assembly planning.

 

Work steps:

  • Literature review on assembly sequence and production line planning and familiarization with PyCAALP
  • Formulation of the extension (model, constraints, objective)
  • Implementation of the extension on the existing framework
  • Evaluation on industrial and open assembly data, including parametric and sensitivity studies (possibility of running on GPU workstation or CPU cluster)
  • Comparison with the current PyCAALP formulation and discussion of the results
  • Documentation of the work

 

Requirements:

  • High degree of independence, initiative, and a structured approach to scientific work
  • Interest in production engineering and manufacturing systems
  • Good programming skills in Python; experience with optimization, graphs, or simulation is an advantage but not required
  • Good CAD skills (SolidWorks, CATIA) depending on the selected topic

  • Ability to analyze results critically and document findings clearly

     

Literature:

[1] Hartmann, C., Demetriades, M., et al.: A Unified Framework for Automated Assembly Sequence and Production Line Planning using Graph-based Optimization. Journal of Manufacturing Systems (2026). doi.org/10.1016/j.jmsy.2026.09.019

[2] PyCAALP source code on GitHub: github.com/TUM-utg/PyCAALP

 

Application:

Please send your CV, transcript of records, and a short motivation letter by email.

Possible start
December 2026
Contact
Marios Demetriades
marios.demetriadestum.de