Smart sheet metal procurement with AI

Ensuring material availability at the lowest possible cost is a challenge for manufacturing companies in an increasingly complex and volatile world. Traditional methods are reaching their limits faster and faster. The solution: the AI software “AI-BOSS” developed at Fraunhofer IML.

The desire of manufacturing companies is clear: Raw materials should always be available in sufficient quantities and, if possible, on time. However, increasingly specific customer requirements mean that the number of material variants to be procured has considerably increased. On the other hand, delivery times and minimum order quantities of suppliers have to be observed. In practice, companies have used traditional methods to plan the procurement of materials: in particular, so-called ABC and XYZ analyses. With these methods, goods are usually classified according to past consumption patterns.

Ideally, the most common materials are stored with uniform requirements. Nowadays, however, material requirements and availability fluctuate strongly at many companies. Although the “ifo shortage indicator for manufacturing” of 31% in June 2023 had decreased since the peak of the Corona lockdown measures, it was still significantly higher than in 2016 (2.1%). The traditional methods of material planning are not well suited to the new so-called VUCA (volatility, uncertainty, complexity and ambiguity) world in which we live. Artificial intelligence (AI) methods can help here – Fraunhofer IML has therefore developed an AI solution for assembling sheet metal ranges in collaboration with Ferro Umformtechnik GmbH & Co. KG. 

Ferro Umformtechnik manufactures telescopic systems for cranes, lifting platforms and telescopic loaders, dumper bodies for vehicle construction as well as components for applications such as conveyor belts, bridge/wagon construction or renewable energy systems for its customers. For this purpose, the company procures raw sheets of different dimensions, steel grades and surfaces from its suppliers. This custom manufacturing is done in small batches, sometimes even with a batch size of one. However, the suppliers demand minimum order quantities that are often larger than required for the respective customer orders. The result: The company has a considerable stock of raw sheets.

For this reason, Ferro Umformtechnik and Fraunhofer IML have developed an AI-based solution to reduce the range. The basic idea: clustering similar metal sheets for different customer orders.

A practical example: The starting point is two customer orders with similar requirements for raw sheet metal. Here, similar means that the sheets have the same steel grade (e.g., S355J2+N), surface finish (hot-rolled and pickled) and thickness (e.g., 7 mm), and thus only differ in length and width. Let us assume that four sheets with a length of 10 meters and a width of 1.8 meters are used for one customer order and six sheets with dimensions of 10.5 x 1.6 meters are used for another order. Typical order quantities of these two raw sheets are 10 pieces. With order-specific raw sheet procurement, six sheets would therefore have to be stored for the first order and four for the second. However, it would be possible to procure an article of raw sheet metal with dimensions of 10.5 x 1.8 meters and to cut off the edge that is not required for the corresponding customer order. The resulting waste would be scrapped. However, this scrapping loss can become a profit if the warehousing costs (storage, handling and capital tie-up) are greater than the waste costs, for example.

If all possible combinations of sheet metal requirements were analyzed, it would be possible to optimize the sheet metal stocks. Other variables such as ordering costs or storage space requirements are also taken into account for the evaluation. However, the number of clusters to be evaluated increases very quickly with the number of similar sheets. While there are only five cluster possibilities for three similar sheets, the number of clusters to be evaluated for ten sheets increases to more than 100,000 combinations. It is not practical to “try out” all possible combinations. This is where the “AI-BOSS” solution (Artificial Intelligence Based Optimization of Sheet Sourcing) of Fraunhofer IML comes in. When AI processes are used, “smart” sheet clusters are created in just a few seconds. The researchers developed and applied the solution as part of a project with Ferro Umfomtechnik and a research project of the Dortmund High-Performance Center Logistics and IT.

Fraunhofer IML continues to work on AI-BOSS in order to use the solution for other companies and in related areas. For example, the range for steel bars can be created in a similar way. The same applies to procurement in wood and paper processing. The potential for smart procurement solutions compared to traditional methods is considerable and worthwhile, especially given the challenges in the “VUCA world.”