Research Projects Ongoing

Click to expand the following past research projects of the Production & Logistics Networks Workgroup.

Research Motivation

steel production is organized as hybrid flow shop with multiple stages
the steel production stage has highly differing requirements regarding the sequence of production
e.g. continuous casting requires to group orders according to the steel type, whereas hot strip milling is concerned with material width and coating with selection of the customer until today, it is not possible to generate a production plan that is able to consider all constraints to different production levels
during rough planning, when the order due dates are determined for the different production stages, only a very limited amount of constraints are considered (e.g. capacity)
during detailed production planning, the actual sequence is generated, which satisfies all local constraints

Figure 1: Production planning in steel production [1]

a conflict arises between competing targets (e.g. the production cost and due date performance)
whenever a cost optimal sequence is derived (lower left side of Fig. 1), shifting orders to other planning periods than originally set during rough scheduling leads to high lateness
a due date optimal sequence (lower right side of Fig. 1) increases the amount of setups and hence production cost

Research Approach

Problem Description

the problem of continuous casting consists of five highly connected sub decisions
ten different, partly conflicting objectives have been identified (e.g. costs, due date fulfillment)
18 different constraints have been described, located at different decision levels

Optimization Method

exact optimization procedures are developed to verify the solutions generated by heuristic procedures
heuristic procedures are developed to achieve the requirement of generating good (near optimal) solutions within reasonable computation time

First Results/Outlook

within the initial test model, the first three decisions (work system selection, slab design and charge batching) were carried out using simple pre-processing procedures, in order to derive a pool of charges as jobs for the scheduling procedure
for cast batching and sequence selection, a minimal model based on family scheduling was derived
no steel specific constraints but only those required to describe a sequence of jobs as the solutions were used
the selected objectives have been tardiness (to measure due date fulfillment) and makespan (to evaluate production costs)
total tardiness and makespan are conflicting targets and the trade-off can be displayed using a Pareto-curve
the minimal model is not sufficient to generate solutions that can be applied in practice
current research is concerened with developing exact and heuristic procedures for an extended model that includes the important constraint of hot metal consumption


Dipl.-Ing. oec. Oliver Herr
Research Associate
Production & Logistics Networks Workgroup
Department of Mathematics and Logistics
Phone: 0421 200 3030
Fax: 0421 200 3078


1. K. Windt, P. Nyhuis, O. Herr, Exploring Effects of Sequencing Modes towards Logistics Target Achievement on the Example of Steel Production, Procedia CIRP, Volume 3, 2012, Pages 620-625

Research Motivation

Manufacturing systems, transportation networks, and supply chains are all logistics systems, which are required to operate efficiently without waste of resources or time.
Sustainable innovations are inventions providing an essential progress concerning social, economic and ecological concerns.
Frugal innovation is an inclusive approach to innovation that maximizes value for customers, shareholders, and society – while significantly reducing the use of financial and natural resources in developing countries.
Reverse innovations are frugal products and services successful in developing markets that make their way back to industrialized countries by creating new market segments.
Roland Berger Strategy Consultants (2013) estimate that frugal products and services are to double their global market share within the next five years.
Research on frugal innovation, reverse innovation and sustainability is still in its infancy.

Research Objectives

The main research question is how frugal and reverse innovation can strengthen sustainable development, and how can business models can be systemized and described in this context.
The main research goal is to fill the existing gap in literature with empirical investigation of frugal products and services and their potential for sustainability.

First Results

Entrepreneurs and companies offering frugal and reverse products and services manage to combine the business model elements in an insightful manner and create economic, social and environmental value.
It is necessary to understand the roles of Multi-National Corporations as well as NGOs for the understanding of business models in the context of frugal innovation and sustainability.
A sustainability impact differs for different directions of innovation in different dimensions.


Eugenia Rosca, M.Sc.
Research Associate & PhD Student
Production & Logistics Networks Workgroup
Department of Mathematics and Logistics
Phone: 0421 200 3077
Fax: 0421 200 3078

Research Motivation

manufacturing systems and their environments are faced with increasing complexity, which consists of structural aspects (e.g., increasing numbers of resources, growing product variety) or dynamic aspects (e.g., supply disturbances, demand variability)

Figure 1: Structural and Dynamic Aspects of Manufacturing Systems. Modified version of this figure in [1].
in complex manufacturing systems, failures are more likely to propagate through the network and can thus negatively influence the performance, e.g., if raw material arrival from a supplier is delayed and no buffers are available, production orders might be delayed
performance (e.g., due date reliability, throughput times, inventory costs) of manufacturing systems needs to stay at steady levels even in volatile conditions

one means to achieve performance robustness towards fluctuations is redundancy; redundant elements are present in manufacturing systems in different forms, e.g., redundant resources, safety stocks, buffers or excess capacity.
yet the decision between robust (many redundancies) or efficient (few redundancies) design describes a trade-off between contradicting targets in manufacturing

Figure 2: Expected trade-off between performance robustness and efficiency [2]
recent approaches to design or adjust manufacturing systems resources (e.g., machines, workforce or buffers) have put a strong focus on increasing system efficiency by reducing costs, which does not necessarily have a positive influence on performance robustness

Research Objectives

classify different elements in manufacturing systems that can be redundant investigate the role of redundancies as a means to achieve robustness in manufacturing systems

Table 1: Redundant elements in manufacturing systems [2]

Figure 3: Sustainability as a function of efficiency and resilience (acc. to Goerner et al. 2009)
biological structures show a high performance under a broad range of perturbations, and their robustness has been studied intensively (e.g., finding an optimal balance where ecosystems are most robust has been suggested by Ulanowicz et al. 2009, see also Figure 3)
thus we use them as potential role models for manufacturing systems and analyze performance robustness using interdisciplinary methods
develop a model that allows evaluating the robustness and efficiency for different manufacturing aspects in manufacturing systems

the exact trade-off relation between robustness and efficiency in manufacturing is unclear (one possible relation is depicted in Figure 2, yet the curves in Figure 3 are also possible)

Figure 4: Possible robustness and efficiency relations in a manufacturing system [3]

First Results

a definition for redundancy in manufacturing systems was developed, which distinguishes between identical, functional and structural redundancy definition is preliminary and is currently being enhanced

Figure 5: Redundancy types in manufacturing [3]
a first analysis approach that uses pathway analysis (as applied in metabolic network analysis) to analyze the systems structural redundancy was applied in a manufacturing context (see Figure 5)

Figure 6: Path counting approach [4]
a simulation model was set up to depict and analyze the trade-off between performance robustness and cost efficiency

Figure 7: Evaluation model for different manufacturing systems and their configuration


Dipl.- Wi.-Ing. Mirja Meyer
Research Associate & PhD Student
Production & Logistics Networks Workgroup
Department of Mathematics and Logistics
Phone: 0421 200 3073
Fax: 0421 200 3078


1. Windt, K., Hütt, M.-T., Meyer, M., A Modeling Approach to Analyze Redundancy in Manufacturing Systems, in: H.A. ElMaraghy (Ed.), Enabling Manuf. Compet. Econ. Sustain., Springer, Berlin, Heidelberg, 2012: pp. 493–498.
2. Meyer, M., Apostu, M.-V., Windt, K., Analyzing the Influence of Capacity Adjustments on Performance Robustness in Dynamic Job-shop Environments, in: Procedia CIRP, Elsevier B.V., 2013: pp. 449–454.
3. Windt, K., Hütt, M.-T., Meyer, M., A Modeling Approach to Analyze Redundancy in Manufacturing Systems, in: H.A. ElMaraghy (Ed.), Enabling Manuf. Compet. Econ. Sustain., Springer, Berlin, Heidelberg, 2012: pp. 493–498.
4. Meyer, M., Apostu, M.-V., Windt, K., Analyzing the Influence of Capacity Adjustments on Performance Robustness in Dynamic Job-shop Environments, in: Procedia CIRP, Elsevier B.V., 2013: pp. 449–454.

Research Motivation

Manufacturing systems, transportation networks, and supply chains are all logistics systems, which are required to operate efficiently without waste of resources or time.
Synchronization is a widely used term in connection with logistics systems, and it promises to increase efficiency by coordinating supply and demand over time and space.
The potential of synchronized activity for increasing the performance of logistics systems has led to a wide range of research activities in production logistics and supply chain management (SCM) as part of the just-in-time philosophy and as a possible mitigation of the Bullwhip Effect.
However, there is neither a common understanding of synchronization in logistics systems, nor an accepted way of measuring and quantifying it.

Existing Definitions

Sync. is “the adjustment of rhythms due to interaction” (Pikovsky et al. 2003)
“systems are synchronized when rigid correlations between their internal dynamical states appear” (Manrubia et al. 2004).
logistic synchronization occurs when a flow-oriented coordination is present in the production-logistic process chain (Fastabend 1997)
logistic synchronization is the output-input coupling, i.e. a firm determination of the input by the output (Wiendahl 1998)

Research Objectives

The main research question is whether logistics systems should be synchronized and to what extent. We aim to investigate synchronization from an engineering perspective in order to gain a quantitative understanding of its emergence, its mechanisms, and its influence on logistics performance.
The main research goal is the comprehensive description of the synchronization black box (shown on Figure 1) as well as the future development of methods to utilize knowledge about synchronization in the design and operation of logistics systems with the effect of increased efficiency and robustness.

Figure 1: Synchronization as a black box. One view on synchronization in logistics is the coordination of actions, followed by the observation of synchronized behavior, without a deeper understanding of the mechanisms in this “black box”. Another view is the observation of non-coordinated actions, leading to an emergent synchronization. The explanation of this phenomenon requires quantitative measures and their linkage to logistics performance.

Proposed Definition

(Chankov et al. 2013)
Various definitions of synchronization are established in the literature (mostly in the natural sciences)
They cover different views: temporal, causal, endogenous, exogenous, state, process and object
Based on these different views, the following definition is proposed:

“Synchronization in logistics systems is the observable and quantifiable phenomenon which represents the temporal coupling and performance-related coupling of different system elements or processes due to direct or indirect interaction. This phenomenon can occur within a single logistics system or between multiple logistics systems.”

As indicated in Figure 2, there are two dimensions of synchronization: temporal coupling and performance coupling. The performance of the two elements follows exactly the same pattern, but there is a time as well as a performance shift between them, as indicated by the intersections of the dotted lines.

Figure 2: Illustration of the two dimensions of a synchronization phenomenon in a logistics system. The two logistics elements (e.g., work stations on a shop floor) exhibit a correlated development of their state, represented by a performance figure (e.g., the current WIP). The coupling is two-dimensional: there is a constant temporal gap, the time lag, between the performance development of the two elements, and there is also a constant difference in performance.

First Results

Cross correlation

c.f.Becker et al. 2013:
measure of linear synchronization between two discrete univariate time series
it describes how strong the development of the two time series is coupled, depending on the time gap \(\tau\)
c ranges from -1 (opposed sync.) over 0 (no sync.) to 1 (perfect sync.)

$$c_{x,y}(\tau) = \frac{1}{N-\tau}\sum_{t=1}^{T-\tau}\left(\frac{x_t-\bar{x}}{\sigma_x}\right)\left(\frac{y_{t+\tau}-\bar{y}}{\sigma_y}\right)$$

Cross-correlation is a suitable measure to identify if two WSs are synchronized

Figure 3: Work in Process (WIP) development for 4 work systems (WS) from a job shop manufacturer

Figure 4: Example of presence of synchronization. WSs 22 and 28 are highly synchronized with WS 1 for TAU=-1

Figure 5: Example of absence of synchronization. In contrast to WS 1, a significant correlation cannot be identified for WS 2

Phase synchronization

describes the coupling of two or more devices that repeatedly perform an event
uses the Kuramoto model to quantify phase synchronization for a set of events
$\beta$ index is used to determine the degree to which a number of events are synchronized

$$\beta (\omega) = \left|\frac{1}{N}\sum_{j=1}^N e^{i\phi_j (\omega)}\right|$$ where $$\phi_t(\omega) = \frac{2\pi}{\omega}\left(t \bmod \omega\right)$$

the level of synchronization of a work system or a whole manufacturing environment can be determined (depending on a phase length)

Figure 6: Average synchronization index depending on phase length in days. Feedback data from six job shop manufacturing companies was used. The synchronization index for each machine was calculated based on the starting time of each operation. The average of the synchronization index for all machines is displayed on the graph. The highest peaks of synchronization can be observed for phase length of 0.5 (half a day), 1 (one day), and 7 (one week).


M.Phil. Stanislav Chankov
Research Associate & PhD Student
Production & Logistics Networks Workgroup
Department of Mathematics and Logistics
Phone: 0421 200 3076
Fax: 0421 200 3078


Becker, T., Chankov, S.M. & Windt, K., 2013. Synchronization Measures in Job Shop Manufacturing Environments. Procedia CIRP, 7, pp.157–162

Chankov, S.M., Becker, T. & Windt, K., 2014. Towards Definition of Synchronization in Logistics Systems. Procedia CIRP, In Print.
Fastabend, H., 1997. Kennliniengestützte Synchronisation von Fertigungs- und Montageprozessen. Leibniz Universität Hannover.

Manrubia, S.C., Mikhailov, A.S. & Zannette, D.H., 2004. Emergence of Dynamical Order: Synchronization Phenomena in Complex Systems, London: World Scientific.

Pikovsky, A., Rosenblum, M. & Kurths, J., 2003. Synchronization: A Universal Concept in Nonlinear Sciences, Cambridge: Cambridge University Press.

Wiendahl, H.-H., 1998. Zentralistische Planung in dezentralen Strukturen? – Orientierungshilfen für die Praxis. In E. Westkämper & R. D. Schraft, eds. Auftrags- und Informationsmanagement in Produktionsnetzwerken – Konzepte und Erfahrungsberichte. Stuttgart: Fraunhofer IPA, pp. 79–107.

Research Motivation

Self-organization as a system of thought to explain the observed behavior of natural and social systems has gained increasing attention over the last four decades [1, 2]. In Production Planning and Control, the company function in entrusted with the efficient allocation of company resources to production jobs, the distribution of decision capacity and authority is believed to increase system performance and robustness in the wake of increasing production complexity, at the expense of non-deterministic, non-optimal performance [2–4]. While the idea is getting more and more track in both academia and practice, the understanding of the necessary conditions (success factors) for the application of self-organization in production planning and control are still largely unknown and relationships between system parameters, degree of autonomy and logistic performance achievement can only be hypothesized (c.f. Fig. 1).

Research Approach & Previous Work

We investigate Graph Coloring Dynamics (GCD) as a minimal model of both scheduling operations and agent interaction. GCD is the agent-based modification of the well-known Graph Coloring problem, defined on graphs. While Graph Coloring is a minimal model for scheduling problems [6], the agent-based variation places GCD firmly in the context of Cellular Automata on general graphs and hence also serves as minimal model for distributed decision making and system evolution over time. Kearns et al. [7] investigated the performance of human subject networks to solve a simple game of social differentiation in Networks (Graph Coloring) as a function of the network structure. They found a distinct increase in performance when adding ’central nodes’ to the graph architecture that are systematically connected to large portions on the graph. Hadzhiev et al. [8] transferred the experiment and applied a first systematization to the local decision rule space. They found dramatic impact on the performance evolution with increasing graph complexity (chords in a ring graph; see Fig. 2).

Figure 2: Impact additional chords in a ring graph on coloring performance (in terms of required color changes). Hadzhiev et al. [8]

Graph & Experiment Setup

We represent operations that have to interact to form a feasible schedule as a ring-graph of \(n\) nodes. With \(X\) colors available to solve the graph, each node is connected to its \(2 * ( − 1)\) neighbors. The ring-structure avoids artefacts at the ends of the observed operation pattern (periodic boundary condition). Central coordination is represented by additional nodes (’Master-Nodes’), connected by \(m\) edges each to the underlying ring. Both random shortcuts and Master-Node links are added such that \(X\)-colorability is maintained. The network is fully described by the 5-tuple \((n, X, s, m, c)\), were \(c = 1\), if and only if the Master-Nodes are connected among each other (see right column in Table 1), and 0 otherwise. Results shown here were obtained, using the AW local decision heuristic (c.f. [8])
To investigate the distinct impact of intentionally emphasized Master-Nodes, we compare the number of required color changes \(x\) of a test-network, defined by 5-tuple , with that of a reference network \((n_{ref}, X, s_ref,0,0)\), where

$$n_{ref} = n + X$$
$$s_{ref} = s + X * m + \begin{cases} \begin{matrix} {X(X-1) \over 2} & c= 1, \\ 0 & c = 0 \\ \end{matrix}\end{cases}.$$

First Results


The experiments show a strong impact of degree of central coordination (parameter \(m\)) on the relative system performance. For non-connected Master-Nodes, we found evidence for the existence of an ’optimal degree’ of central coordination, backing the hypothesized relationship by Philipp et al.[5]. For connected Master-Nodes however, no significant decrease in relative performance was observed for high values of \(m\). This can be seen as evidence that improved coordination within a central coordination can significantly improve its performance and effectively outperform distributed decision making approaches.

Next Steps

In following steps, we will try to get an understanding for the mechanistics behind the observed phenomenon. We will then shift our focus towards more sophisticated models of distributed decision making in production logistics to try to verify the result.


Henning Blunck
Production & Logistics Networks Workgroup
Research Associate / PhD Student
+49 421 200 3075


1. T. De Wolf, T. Holvoet, in Engineering Self-Organising Systems, ed. by S. A. Brueckner et al. (Springer, 2005), vol. 3464, pp. 1–15.
2. K. Windt, M. Hülsmann, English, in Understanding Autonomous Cooperation and Control in Logistics, ed. by M. Hülsmann, K. Windt (Springer Berlin Heidelberg, 2007), pp. 1–16.
3. D. Dilts et al., Journal of Manufacturing Systems 10, 79 –93 (1991).
4. D. Trentesaux, Engineering Applications of Artificial Intelligence 22, Distributed Control of Production Systems, 971 –978 (2009).
5. T. Philipp et al., English, in Understanding Autonomous Cooperation and Control in Logistics, ed. by M. Hülsmann, K. Windt (Springer Berlin Heidelberg, 2007), pp. 303–324.
6. M. M. Halldórsson et al., English, Algorithmica 37, 187–209 (3 2003).
7. M. Kearns et al., Science 313, 824–827 (2006).
8. B. Hadzhiev et al., Advances in Complex Systems 12, 549–564 (2009).