naar top
Logo Print
09/10/2018 - ASTRID BAXTER


Using agent-based modelling (ABM) to improve strategic decision-making in industrial processing

In the past century, manufacturers have dealt with the same suppliers, customers and competitors. They concentrated on economies of scale and on optimizing the efficiency and effectiveness of their processes. But we now live in a world in which many of our previous assumptions are no longer valid. Raw materials become depleted. Climate change drives large scale energy transition. Governments impose increasingly stringent regulations on what manufacturers can do. Demographic shifts disrupt societies around the world. All this means that manufacturers must learn how to manage price volatility in their raw materials, incorporate greater flexibility into their operations, and make sure their logistical operations are resilient to sudden changes.



“But it's not all doom and gloom", says Andreas ten Cate, Director International Business Development at the Institute for Sustainable Process Technology (ISPT) in the Netherlands. “On the contrary. The threats brought by constant change are also bringing opportunities. New technical capabilities are freeing up manufacturers in the design of their supply chains. For example, many companies are currently developing and applying mobile, flexible processes. Modular, 'container-based' processing solutions tremendously enhance the flexibility to rapidly change your location or feedstock choices and therefore your company's position in the supply chain." Working out how to best exploit these new capacities in complex, global value chains is one of the great challenges - and opportunities - for the designers of today's process industry supply chains.

One approach to dealing with this kind of systems is agent-based modelling (ABM). ABM's have shown their strength in areas as diverse as social behaviour, road traffic and portfolio management. “ABM makes it possible to analyse the behaviour of a large network of companies as the results of the actions and interactions of the individual companies - the independent 'agents' - that follow quite simple rules." An analogy is the simulation of traffic, which may be seen either as a single complex entity (a road with five kilometre of traffic jam as one big occurring event), or as the product of independent cars, cyclists, pedestrians etc. that respond to their environment, such as the weather, time of day, season, roadworks, occurrence of accidents, and so on. “Mimicking this system by simulating individual cars then can predict the appearance of e.g. traffic jams bottom-up by applying the simple rules of individual users. This computational approach is flexible and versatile, while the increasing availability of computing power means that nowadays complex and realistic supply chain problems can be analysed in reasonable time frames."


Quantification of the effects of operational choices
Quantification of the effects of operational choices in company performance flexibility in manufacturing. Impact of feedstock flexibility, mobile plants, modular plants in industry network.



“With the rising interest in flexible assets, the question appeared how industries can best take advantage of these new capabilities. Opposing economies of scale, the industry was trying to create commercial roll-out of small-scale flexible processing units. But against the logic of economies of scale, the costs would increase in these systems, unless benefits could be obtained in different ways, by operating differently in the supply chain. This realisation was the start of the Economy-of-Chain project. Four years ago, ISPT initiated this project, led by scientists from Royal DSM and TU Delft, to explore how ABM models can help to identify the competitive benefit of modular plants. “By exploring different scenarios with advanced ABM models we would be able to study the disruptive impact of modular plants on large-scale industrial value chains. This was very exciting," says Andreas, “as it was the first time the concept of agent-based modelling had been used to tackle challenges of this scope."



“We started by looking, as a proof of principle, at how ABM can simulate market dynamics in manufacturing supply chains," says Andreas. “The researchers built models for decentralized markets. This basically means that we simulate the evolvement of market prices between companies that trade goods amongst each other- feedstocks and products. ABM's had been used before to model centralized and fully transparent markets - the stock exchange - but not for distributed and less-transparent markets. With this model in place we could map existing supply chains and analyse the market dynamics and observe price dynamics and productivity performances that were very close to real market phenomena. From here on we could modify the markets by introducing new assets and observe how the supply chains develop under these changes. From these simulations we obtain answers to questions such as 'Do we need to relocate, and if so, when?', 'Should we start processing new raw materials?', 'What effects will regulatory changes of the government have?', 'How should we deal with changes in prices?', 'Will we need to distribute to different customers?', and so on. This led to very promising results. Agent-based modelling was clearly able to simulate the dynamics of the market and provide insight into a whole range of possibilities that could be explored by taking different pathways."


agent-based modelling (ABM)
Model applied to fuel transition in Port of Rotterdam vessels through fuel market modelling


In 2015, Gerben Bas, scientist involved in the Economy-of-Chain project, was asked to apply agent-based modelling for decentralized markets to an issue faced by the Port of Rotterdam: how to support the transition of international vessels from heavy fuel oil to LNG. “This is an urgent question," says Gerben, “given the need to eliminate high-sulfur maritime fuels for environmental and regulatory reasons. So we set out to explore the different ways in which the Port of Rotterdam could anticipate a change in sulphur emission regulations, and how creating different circumstances in Rotterdam affects the adoption of Liquid Natural Gas (LNG), which is widely considered to be the most acceptable alternative to heavy fuel oils."

As he worked on his agent-based models, Gerben and his team came up with a number of interesting findings. “First, we found that, contrary to expectations, the price of LNG was not an issue: vessels will switch to LNG at a wide range of prices. We also found that vessels are more likely to switch to LNG if a secure supply of LNG is available in enough ports around the world. And third, we found that vessels are more likely to switch to LNG if the authorities encourage the retrofitting of vessels." On the basis of this model-based analysis, Gerben's team was able to make three recommendations to the Port of Rotterdam:

(1) to maintain its leading role in the bunker fuel market, the Port of Rotterdam should invest in storage facilities for LNG;
(2) it should coordinate efforts with other large bunker ports to develop a global network; and
(3) it should start initiatives to stimulate the retrofitting of vessels.



Projects such as Economy of Chain require substantial resources, both in terms of expertise and funding. “That's why companies are increasingly exploring ways of joining forces," says Andreas ten Cate. “This is exactly what ISPT facilitates. In our projects, scientists from industry, universities and knowledge institutes work together in a trust-based network to speed up and improve innovation processes. The Institute not only gathers knowledge, but also develops, demonstrates and applies breakthrough technologies, especially in the field of process technology." It makes sense to work together in this way, says Andreas. “After all, the issues we typically work on benefit everyone. By improving the efficient use of resources, or enabling manufacturers to reduce their carbon footprint, for instance, we're working towards a more sustainable world." 



The possibility of successfully applying ABM to the complex world of industrial processing opens up exciting prospects, says Andreas ten Cate. “It enables companies to prepare for unexpected changes and respond quickly and flexibly. It also helps them to better understand cause and effect and incorporate this knowledge into their strategies, and ultimately take better decisions. But we've only just begun. Many other companies could benefit from applying ABM to their processes, and a number of project partners are already joining forces to refine ABM, study additional aspects of economies of chain and thus validate the method still further. We look forward to working with anyone interested in taking this step."For more information, contact Andreas ten Cate at


Simulation of the supply chain with an agent model for factories