Full Professor, Bernd Rogge Professor of Global Production Logistics and President of Jacobs University, Bremen
Breaking the Wall of Unpunctuality. How Overcoming Equilibrium Thinking Helps Logistics in a Fluctuating World to be on Time
Globalization leads to networks in logistics
Logistics is about planning and control of all material and information flows in global networks. Actors in a production network are manufacturers, their suppliers, distributors and customers worldwide. Logistics involves all main processes as procurement, production and distribution as well as redistribution and recycling and after-sales services.
In our research we concentrate on multi-variant series production of technologically demanding products of high complexity. One can say that logistics is the control instrument of globalization, but obviously it is more than just transportation of goods.
Production logistics target
Logistics is also about achieving targets. Besides due date reliability, which means punctuality, we aim to achieve a low inventory level, short delivery times and a adapted capacity utilization of machines in order to produce with less costs.
The challenge is that these targets are in a trade off position and therefore it is not possible to achieve all targets in an optimal manner.
Linear process chain in logistics
But as pointed out in the title of the presentation: Logistics has to deal with fluctuations – and that is kind of terrible for achieving high punctuality.
Why do we have fluctuations? Why is it a challenge to achieve high punctuality in order to deliver products on time to customers?
To get answers to these questions an experiment with the audience is introduced. It is about passing balls through 12 rows. The only process is to hand the ball to the next person in front, sitting in the next row. The balls have to reach row one. The balls have been arrived but with different arriving times. Obviously, unplanned deviations occurred. One could see delivering on time is difficult – even in such a simple scenario.
Linear process chain in logistics
In order to elaborate this experiment we also simulated this experimental setting. We have a linear process chain represented by the 12 rows. For us every human represent a machine. We have a continuous steady inflow and a certain probability p to proceed to the next machine and a probability (1-p) to remain on the current machine – which is comparable that e.g. someone just dropped the ball.
We did this in a MATLAB computer simulation with 100.000 runs and p=0.9 and we calculated a frequency distribution with a mean value of 13.33 and a standard deviation σ = 1.48. The consequence out of this is that even simple processes produce fluctuating output. Equilibrium thinking means that one tries to keep the logistics system or machines in a balanced position based on mean values not taken into account fluctuations. These strategies use average values in order to simplify the reality.
Complexity drivers in an exemplary material flow network
The reality looks different as you can see on this example of a steel mill. The complexity drivers are the number of different production sites, a high number of steel types (approx. 650) and even more specific customer orders. The different technologically high demanding process starts with continuous casting, adjusting, hot roller mill, dimensioning, cold roller mill, annealing, hot coating, and again adjusting. But the flow restrictions show the real complexity of the material flow network. As an example, on the hot roller mill about 200 sequencing rules have to take into considerations.
Conventional control strategies
Conventional strategies try to control such a network in a centralized way, meaning that every machine, every order is directly connected with a central planning and control system. Obviously, it does not look like that this is a simple task, indeed to plan and to control such a network is very complex. It seems to be that centralized production control strategies are not able to deal with a high number of influencing factors and frequent changes.
In order to reduce the complexity we enable logistics objects e.g. nodes in a network to become intelligent. That way, these objects only decide upon their actual situation not considering the overall network. It seems to be that under certain conditions this approach is beneficial in comparison to conventional strategies. Per definition autonomous control means to process information, to render decisions and to execute these decisions by logistics objects [Hülsmann, M.; Windt, K. (eds.): Understanding Autonomous Cooperation & Control in Logistics – The Impact on Management, Information and Communication and Material Flow. Springer-Verlag, Berlin, 2007].
In general, alternatives are needed in order to run effectively autonomous control. In case alternatives exist the order is able to decide between these alternatives machines. For example, the order would prefer to get transported to the machine with the shortest queue length.
We concentrate our research on developing effective decision methods in a logistics sense as well. In experiments on graphs we figured out that specific decision methods performed better with increasing network complexity. We also found counter intuitive effects, e.g. to wait and allow other objects to act first gave very good results under specific constraint [Windt, K., Hütt, M.-T.: Graph coloring dynamics: A simple model scenario for distributed decisions in production logistics.
CIRP Annals - Manufacturing Technology 59/1 (2010), pp. 461–464].
Discovering general decision strategies on graphs
We actually started a new research approach. We try to learn from biology and therefore my workgroup co-operates with the Systems Biology Workgroup of Marc-Thorsten Hütt. We are motivated from the fact that metabolic systems in cells need to perform well for a wide range of environmental inputs. Obviously a cell is able to produce biomass under various conditions. As we could identify processes in cells as production and distributions in a cell the aim is to understand these dynamic processes and to transfer the knowledge to logistics. The first step is the development of a common model.
Production logistics of the future is not the logistics of averages. This means, that logistics systems should be prepared to act with fluctuations in unbalanced environments. Logistics as a science discipline tries to understand the complexity and its implications. Based on this understanding we develop new or adapted methods also inspired by other disciplines.
Thank you very much.