PhD Defence Johannes Unglert

designing future factories - a novel approach to configuring production systems by combining set-based and automated design

Johannes Unglert is a PhD student in the research group Design Production & Management. His supervisor is Professor Fred van Houten from the faculty Engineering Technology (ET) 

Production technology innovations introduced in the last decades have helped manufacturing companies to create the capabilities to produce new and complex products more efficiently. One type of system concepts that made this possible are Reconfigurable Manufacturing Systems (RMS). An important characteristic of RMS is that they enable the production system engineers to change the configuration of the production system. Hence, the capabilities and capacities of RMS can be adjusted in response to changing customer demand, which allows re-use of the system components, an extended lifecycle of these assets and a maximum value-added of the system components. Nevertheless, the design of system configurations bears a high complexity for the production engineers, since numerous configuration opportunities exist, each with its own implications. This so-called design space of RMS configurations can be heterogenic, large and difficult to grasp by the decision-makers. Every time the system is reconfigured, the system designers have to come up with suitable configuration designs, determine and evaluate their implications and eventually choose one of these solutions. In this challenging process, computational tools can relieve the system designers from tedious tasks by automating the design, analysis and evaluation of system configurations and thereby support the design process. In this thesis I determine the basic objectives and required features of computational tools for automated design of reconfigurable production systems and propose a novel approach to automating RMS design. Subsequently, I evaluate the approach based on experimentation to determine its implications and conduct an empirical study to indicate its usefulness and applicability in industrial practice.

In addition to the number of configuration opportunities, the complexity and extent of the design task depends significantly on the possible number and types of the involved subsystems. This implies that a large number of design variables can exist, whose impact on the performance of the resulting systems cannot be stated generally and has to be determined for each specific case. Furthermore, it can be required to consider interdependencies between the variables of the design problem, as well as extrinsic pre-conditions for system configurations, such as space limitations associated with the existing factory infrastructure. An important implication with regard to these factors is that boundary conditions for planning the future system - such as customer demand - are often subject to high uncertainty. Hence, the assumptions used as a basis for developing the design are often difficult to forecast and justify. Individually and also collectively these circumstances can make the configuration design a complex endeavor. Design automation software can be used to support the system designers in mastering these difficulties. By automating tedious tasks associated to the synthesis, analysis and evaluation of system configurations, the main benefits of such tools can be an increased effectiveness and efficiency of the configuration design process. As a result, decision-makers can get an understanding of possible alternative solutions and thus obtain well-performing systems. Moreover, the tools can reduce the time required for assessing a broad spectrum of solutions. Yet, the development of design automation tools commonly requires significant efforts that have to be justified considering their prospective capabilities. In this context, the re-usability and versatility of the support tools play important roles. This research discusses the link between the characteristics of RMS configuration design and design automation approaches to establish a framework of the relevant features and objectives that tools for automated design of RMS configurations should expose. The main objectives of such tools are efficiency and effectiveness of the automated generation, analysis and evaluation of system designs. The key features to realize these objectives are algorithms and user interfaces that make it possible to generate solutions for various formulations of the system design problem, flexibly integrating preference information for the solutions and integrally comparing the resulting solutions. The innovative design automation tool developed here realizes the functionality and objectives of the framework by combining automated design, customizable visualization of the resulting design spaces and the rationale of set-based design, i.e. decision-making based on multiple alternatives. The system model and algorithms are tailored to the concrete problem case of a manufacturing company, which is presented in detail. The anticipated functionality of the developed software is explained subsequently, which enables the user to iteratively test and explore the implications of various ways of formulating the design problem and associated assumptions.                                                                                             Following the presentation of the developed tool, its evaluation is described. The first part of the evaluation is focused on the tool’s intrinsic characteristics with regard to the synthesized system designs. It examines the de-facto coverage of the potential design space by the algorithms, the implications of changed user input and the potential to suggest solutions for various problems. The results of this evaluation confirm the aspired effects of the approach, namely its capability to efficiently generate a broad spectrum of different solutions for different problem formulations that can be assessed and compared by the users. In addition, the evaluation results make it possible to identify improvement potentials with regard to the effectiveness of the design synthesis algorithms. The second part of the evaluation focuses on the extrinsic effects. An empirical study was designed to let industrial system designers determine the practical applicability and usefulness of the tool. For this purpose, the designers used the tool to solve two experimental cases, which represent realistic system design problems. After the experimental cases, qualitative data was collected by leading semi-structured interviews with the objective to document the system designers’ opinion with regard to the applicability and usefulness of the tool. The designers stated that the tool would allow them to efficiently increase the number of configurations considered and to strategically approach the design of configurations. The results of these interviews indicate a high use value and applicability of the tool in industrial practice.

This research provides three major contributions. Firstly, the proposed reference framework makes it possible to evaluate existing and future support tools and detect their respective implications. Secondly, the support tool and its detailed description represent an innovative approach to support production engineers by means of design automation, which supports the users in determining well-performing system configurations and reducing the duration of the design process. Thirdly, the evaluation and validation of the tool with regard the aspired improvements of efficiency and effectiveness of the design process allow a differentiated impression of the tool’s strengths and opportunities for improvement. In this context, the innovative design of the empirical study for assessing the tool in industrial practice has to be highlighted, as such an approach to evaluation has not been found in existing literature on support tools for production system design.