Louis Schäfer, M.Sc.

  • 76131 Karlsruhe
    Kaiserstraße 12

Louis Schäfer, M.Sc

Area of Research:

  • Product-Production-Codesign
  • Machine Learning in Production Planning and Control
  • Industrie 4.0

General Tasks:

Projects:

  • MoSyS - Human-oriented design of complex System of Systems
  • teamIn - Digital leadership and technologies for the team interaction of tomorrow

 

Curriculum Vitae:

since 10/2020

Research Associate at the Institute of Production Science (wbk) at Karlsruhe Institute of Technology (KIT)

10/2014-08/2020 

Study of Mechanical Engineering at Karlsruhe Institute of Technology (KIT)

Publications

[ 1 ] Kuhnle, A.; Schäfer, L.; Stricker, N. & Lanza, G. (2019), "Design, Implementation and Evaluation of Reinforcement Learning for an Adaptive Order Dispatching in Job Shop Manufacturing Systems". Procedia CIRP, eds. Elsevier, pp. 234-239. 10.1016/j.procir.2019.03.041
Abstract
Modern production systems tend to have smaller batch sizes, a larger product variety and more complex material flow systems. Since a human oftentimes can no longer act in a sufficient manner as a decision maker under these circumstances, the demand for efficient and adaptive control systems is rising. This paper introduces a methodical approach as well as guideline for the design, implementation and evaluation of Reinforcement Learning (RL) algorithms for an adaptive order dispatching. Thereby, it addresses production engineers willing to apply RL. Moreover, a real-world use case shows the successful application of the method and remarkable results supporting real-time decision-making. These findings comprehensively illustrate and extend the knowledge on RL.

[ 2 ] Schäfer, L.; Burkhardt, L.; Kuhnle, A. & Lanza, G. (2021), "Integriertes Produkt-Produktions-Codesign", wt Werkstattstechnik online, vol. 111, pp. 201-205. 10.37544/1436-4980-2021-04-23
Abstract
Eine steigende Variantenvielfalt, hohe Marktvolatilität und heterogene Prozesslandschaften betonen die Bedeutung einer simultanen Betrachtung von Produkt- und Produktionssystem für produzierende Unternehmen. Vor dem Hintergrund einer wandlungsfähigen Produktionsplanung und -steuerung stellt dieser Beitrag eine Methodik zur Implementierung eines integrierten Produkt-Produktions-Codesign vor. Bestandteile sind ein ganzheitliches Änderungsmanagement und die Identifikation von Lösungsmustern.

[ 3 ] May, M.; Albers, A.; Fischer, M. D.; Mayerhofer, F.; Schäfer, L. & Lanza, G. (2021), "Queue Length Forecasting in Complex Manufacturing Job Shops", forecasting, no. 3, pp. 322-338. 10.3390/forecast3020021
Abstract
Currently, manufacturing is characterized by increasing complexity both on the technical and organizational levels. Thus, more complex and intelligent production control methods are developed in order to remain competitive and achieve operational excellence. Operations management described early on the influence among target metrics, such as queuing times, queue length, and production speed. However, accurate predictions of queue lengths have long been overlooked as a means to better understanding manufacturing systems. In order to provide queue length forecasts, this paper introduced a methodology to identify queue lengths in retrospect based on transitional data, as well as a comparison of easy-to-deploy machine learning-based queue forecasting models. Forecasting, based on static data sets, as well as time series models can be shown to be successfully applied in an exemplary semiconductor case study. The main findings concluded that accurate queue length prediction, even with minimal available data, is feasible by applying a variety of techniques, which can enable further research and predictions.

[ 4 ] Kandler, M.; Schäfer, L.; Gorny, P. M.; Ströhlein, K.; Lanza, G. & Nieken, P. (2021), "Learning Factory Labs as Field-in-the-Lab Environments". http://ssrn.com/abstract=3862569
Abstract
A central challenge in the implementation of digitalisation and Industry 4.0 in companies is the human-centred development and design of the new technologies. These technologies have a major impact on the way people work and thus also on the motivation and satisfaction of employees. A thorough understanding of underlying drivers of employees' technology acceptance and possible resistance to change is crucial for a successful implementation of such technologies. Experimental economic research methods comprise a way to record the effects on human behaviour and working methods allowing for causal evidence. Controlled laboratory conditions offer the option to vary only individual variables and measure their influence on human behaviour, motivation, satisfaction, or working methods. In contrast to standard computer labs, learning factories offer the possibility to carry out experiments in a real production environment and thus observing behaviour in real work tasks in a realistic environment. This leads to increased external validity while at the same time the strict experimental protocol still allows making causal claims. Learning factories so far do not fulfil laboratory requirements. We first outline the prerequisites for the execution of empirical experiments. We then introduce our research concept, based on the example of the Learning Factory Global Production of wbk. The goal of this paper is to create an environment that allows collecting empirical and meaningful research data in a learning factory. Finally, we exemplify our concept with an experimental design on how digitalisation can foster decentralised decision-making. This research will inform a more human-centred design of digitalisation technologies and its effects on employee behaviour.

[ 5 ] Kuhnle, A.; May, M.; Schäfer, L. & Lanza, G. (2021), "Explainable reinforcement learning in production control of job shop manufacturing system", International Journal of Production Research, 10.1080/00207543.2021.1972179
Abstract
Manufacturing in the age of Industry 4.0 can be characterised by a high product variety and complex material flows. The increasing individualisation of products requires adaptive production planning and control systems. Research in the area of Machine Learning demonstrates the applicability and potential of Reinforcement Learning (RL) systems for the control of complex manufacturing. However, a major disadvantage of RL-methods is that they are usually considered as "black box" models. For this reason, this paper investigates methods of explainable reinforcement learning in production control. Based on a comprehensive literature review an approach to increase the plausibility of RL-based control strategies is presented. The approach combines the advantages of high prediction accuracy (e.g. neural networks) and high explainability (e.g. decision trees). In doing so, understandable control strategies such as heuristics can be generated, and an advanced RL-system can be designed including specific domain expertise. The results are demonstrated based on a real-world system, taken from semiconductor manufacturing, which is investigated in a simulated approach.

[ 6 ] Höger, K.; Schäfer, L.; Schild, L. & Lanza, G. (2021), "Towards a User Support System for Computed Tomography Measurements Using Machine Learning" in Production at the Leading Edge of Technology. WGP 2021. Lecture Notes in Production Engineering., eds. Behrens, B.; Brosius, A.; Drossel, W.; Hintze, W.; Ihlenfeldt, S. & Nyhuis, P., Springer, Cham., pp. 506-514. ISBN/ISSN: 978-3-030-78423-2
Abstract
Increasing importance of X-Ray Computed Tomography (CT) for industrial applications demands suitable approaches for user support. These are intended to guide the choice of setting parameters to improve measurement quality and enable CT-technology access also to non-experts. In the past, several user support systems (USS) have been developed. One promising approach uses knowledge gained from historical measurements. However, this initially requires time-consuming measurements and thus limits generalization, and the potential of machine-learning (ML) techniques. Further, no statement about the achieved quality is possible. In this paper, a concept and the required workflow to build a USS using ML is introduced. Therefore, a suitable ML approach is identified by examining patterns between ML algorithms and their characteristic applications. To provide the required database and analyze the measurement quality, the use of a virtual CT is suggested. Based on the proposed concept, future work will focus on the implementation of the USS.