Living in a world of disruptions, supply chain management professionals know that many processes have become more fragile over the last few years. Now scientists from Germany are trying to learn from nature itself to give those processes the “anti-fragility” they need in the current times. Artificial intelligence plays a key role in this endeavor.
Fragility is a current problem in supply chain management. Lean processes, automation, just-in-time logistics and resource savings make industrial production more prone to disruption. Simultaneously, the high transport intensity of goods due to a global mesh of networks of transport links, warehouses and transshipment centers increases the probability of errors.
Supply chain managers have to react to unpredictable and strong market fluctuations, shorter product life cycles, a greater number of variants and shorter delivery times. In terms of logistics, the decisive performance of a company is based on delivery capability, delivery time and reliability, as well as delivery quality and flexibility.
Furthermore, the ability to deliver is determined by the ability to predict the mostly unknown and fluctuating incoming customer orders, and by the security of supply from suppliers. These suppliers generally also face similar challenges as the company itself.
The question now arises of how to buffer demand fluctuations on the one hand, and to cope with supply gaps caused by disruptions further down in the supply chain on the other. Of course, there is a very simple and practical solution – increase inventory! However, this is associated with enormous capital commitment, higher processing costs, and of course lower profits.
Antifragility as a concept against stressors and volatility
Scientists of RWTH Aachen University, Germany, are looking for a solution to this problem. Marco Becker of the Laboratory for Machine Tools and Production Engineering (WZL)explains: “Nature has developed mechanisms to deal with disruptions. Our immune system, for example, is activated and strengthened by external attacks.” The German researchers want to find and understand these natural mechanisms.
As part of a project launched in June 2020, the researchers are now carrying out preparatory work for future research projects that will transfer these mechanisms from nature to logistics and production. Regardless of the type of production or product, the researchers want to investigate in general whether antifragility makes a relevant difference in logistics and production.
A good example of anti-fragility are foxes as they are exceptionally adaptable. They eat mice, earthworms or fruit, depending on the food supply and the season. In the countryside, they steal from farmers, while in the city they plunder garbage cans. These are concepts against volatility, i.e. fluctuations in food availability. If an unusually large number of foxes die due to hunting or disease, they then have considerably more offspring. This is a concept against stressors, i.e. stimuli that cause stress.
“Stressors and volatility can strengthen rather than weaken systems in nature,” explains Daniel Trauth, project manager at WZL. “Traffic jams on the highway or extremely hot summers, an epidemic or a violent storm, many factors can paralyze a production system. Now we want to draw inspiration from nature and learn how to better deal with disruptive variables.”
Biotechnology and philosophy should help
“Stressors are quite normal in biology,” says philosopher Dawid Kasprowicz from the Chair of Philosophy of Science and Technology. In mechanical engineering it is entirely different. “Willfully damaging a system in order to improve it is initially counter-intuitive for mechanical engineers.” His chair is more of a mediator between the two worlds of science.
The expertise of Professor Ulrich Schwaneberg, Head of the Department of Biotechnology and member of the scientific management of the DWI-Leibniz Institute for Interactive Materials, lies in the field of directed evolution. Directed evolution makes it possible to reprogram enzymes from nature and thus make them usable for industrial processes.
Frances Arnold – whose team Schwaneberg was part of from 1999 to 2001 – was awarded the Nobel Prize in Chemistry for the development of the method in 2018. “It would be conceivable that biodegradable building blocks that are tailor-made for applications, by means of directed evolution, would make it possible to use computer-aided methods to develop optimal coating systems that would then optimize themselves in the event of external stress influences during production,” he says. “The composition then changes according to temperature and humidity, so that, for example, antimicrobial or water-repellent finishes remain free of defects.”
Artificial intelligence plays a key role
The scientists plan to use artificial intelligence (AI) to find ways to transfer the mechanisms from nature to technical systems. “I have a digital software management system in mind,” reveals Trauth, “that supports or warns people working at machines in making decisions.”
AI should provide the information basis for improving the volatility, randomness and stressors of antifragile systems in different areas. The WZL researchers are currently testing the implementation concept on a CNC grinding machine. “We are getting an enormous amount of information that we cannot use properly yet,” explains Becker. “The systems are complex and people cannot understand the huge amount of data.”
The factors that influence the process in this particular example range from the grain distribution of the grinding wheel to the humidity during processing and the availability of the machines. This again depends on the delivery and arrival time of feedstock which is influenced, for example, by weather conditions, the traffic situation, route profile, data on the transported goods, day of the week and time of day, etc. The more information available, the more accurate the data processing and correspondingly high benefits. But not only “technical” factors are important, a holistic perspective also includes the people integrated in the whole process, for example the production employees and the drivers’ constitution: Will the worker at the machine become ill or injured? Is he or she perhaps less focused on Mondays or tired during the night shift? The amount of data can be expanded almost indefinitely.
During operation of the CNC machine, more than 1,000 signals arrive simultaneously, which cannot be processed without AI. Moreover, CNC processing is only one small processing step. In the production of a car, the many individual processes, work steps and parts deliveries add up.
First results are very promising. For example, the researchers used a second laser as a stressor in a laser welding process. The result was surprising: the additional laser improved the result on several levels, especially the quality benefited. It remains to be seen how far the anti-fragile concept can be transferred to supply chain management. The researchers are focusing primarily on evolutionary algorithms. These AI algorithms are already being used for other problems or for optimization, such as the development of sensor networks, stock market analysis or RNA structure prediction.