RESEARCH PROGRAM

Title: Machine Learning Based Factory Concept Layout Planning

 

Name:  Matthias Klar

E-Mail:  matthias.klar@mv.uni-kl.de

Phone:  +49 631 – 205 - 5629

 

Project description:

 

Starting situation

Concept layout planning is a central element of the factory planning process in which, in particular, the material flow is optimized by efficiently allocated functional units. Layout planning is characterized by its complexity, which is the result of considering multiple planning and optimization criteria. existing approaches use either methods that are based on the individual experience of the planner and therefore lead rarely to optimal solutions or methods from the field of classic mathematical optimization which only consider few optimization criteria and consequently lack in case of generality. Consequently, an approach is needed that enables a generation and optimization of concept layout in a sufficient time, concerning multiple optimization criteria.

 

Approach

To solve this problem, a reinforcement learning algorithm will be developed. The algorithm will generate valid concept layouts and optimize them concerning multiple optimization criteria. Furthermore, a material flow and energy simulation will be integrated into the algorithm, which will allow a layout evaluation. The starting point of the project is an analysis of relevant planning and optimization criteria. These enable a distinguishment between valid and invalid solutions. They also form the basis for evaluating different layout variants based on defined parameters. Subsequently, the reinforcement learning algorithm will be developed and evaluated based on different scenarios.

 

Expected Results

The expected result is a novel reinforcement learning algorithm that allows to generate valid concept layouts in a sufficient time and optimizes them concerning multiple criteria that can be individually prioritized.