Artificial Intelligence (AI) - Projects 2022

Improving the efficiency of empty container handling in maritime logistics

Schiffscontainer sind entscheidend für das Funktionieren internationaler Lieferketten und den Welthandel.
© Studio Bogumil /Fraunhofer
Shipping containers are a vital component in the running of global supply chains and international trade.

Global trade chains are based on standard container units, particularly in the case of maritime transport. Globally, there are almost 30 million such shipping containers in use, with a total capacity of 47 million TEU (one TEU = 20-foot standard container). It is therefore essential to the successful operation of the supply chain that containers are turned around quickly and that empty containers are swiftly returned. 

And yet there is a bottleneck in this process — caused by regular inspections of containers and, in some cases, the need for repairs. Every new container is inspected by specialists on a regular basis after five years — for mechanical or weather-related damage, for example. Improving the efficiency of container inspections was therefore the goal of the COOKIE (Container Services Optimized by Artificial Intelligence) project funded by the German Federal Ministry for Digital and Transport (BMDV). The project partners were HCCR Hamburger Container- und Chassis-Reparatur GmbH (coordinator) and the Fraunhofer Center for Maritime Logistics and Services CML, part of the Fraunhofer Institute for Material Flow and Logistics IML. Another objective of the project was to optimize tank container cleaning operations using an AI system. 

Port inspectors use handheld devices to document and assess damage through imagery. These “checkers”, as they are called in port jargon, are supported by 3D models developed in the course of the project, enabling highly reliable and detailed localization and identification of the damaged areas. This forms the basis for transparent repair and cost proposals, thereby boosting the confidence that shipping and leasing companies have in the repair service provider. Improving data quality is also the cornerstone for continued development and successful deployment of AI models that can help automate the inspection process in the future. 

Project description


Vehicle parts to be remanufactured instead of recycled

© Fraunhofer IPK/Larissa Klassen
AI-supported assistance system for semi-automated sorting of used components.

The circular economy is an important tool for climate protection. Remanufacturing, the process of restoring used parts to their original condition, is a key aspect of the circular economy. To make this process efficient and profitable, clear identification and evaluation is critical: Many products, especially in the automotive sector, are virtually indistinguishable from one another and are difficult to identify due to dirt and wear. Up to now, this task has been carried out manually by specialists. 

The EIBA project, funded by the German Federal Ministry of Education and Research (BMBF), addresses this issue by way of an assistance system based on artificial intelligence (AI): In future, the specialists will be supported by a reliable, AI-based suggestion system when identifying and assessing defective wear parts such as starters, air-conditioning compressors and alternators. The consortium partners are Circular Economy Solutions GmbH, the Fraunhofer Institute for Production Systems and Design Technology IPK, the Technical University of Berlin and acatech — the German National Academy of Science and Engineering.

The consortium’s first step was to combine image-based processing using on convolutional neural networks and analysis of specific business data — such as the origin, customer, date and location of parts. The information is processed and merged by two AI systems simultaneously. The outcome is presented to the specialist in the form of a suggestion list with a preview image and part number: This means that the expert will continue to make the final decision. 

Every year, about five to seven percent of one million used parts processed by Circular Economy Solutions GmbH — that is, up to 70,000 parts — are discarded because they cannot be identified. A study conducted as part of the project revealed a recognition accuracy of 98.9 percent. Seen in terms of the 70,000 used parts that are discarded, it is expected that AI-based identification will al-low 67,200 more used parts to be fed back into the cycle than before. This reduces waste, lowers the CO2 footprint and extends the service life of products.


Press release »Turning old into new: A second life for vehicle components«

Designing trustworthy AI — securing and assessing AI systems

Die Software-Toolbox ScrutinAI wurde für die Analyse von Bild- und Videodaten entwickelt. Sie kann das Vorgehen von KI-Modellen sichtbar machen.
© Fraunhofer IAIS
The ScrutinAI software toolkit was developed for analyzing image and video data. It has the power to make the internal actions of AI models visible.

Artificial intelligence (AI) could be used in all sorts of ways to simplify our daily lives and automate processes — in the automatic analysis of application documents, for example, or image recognition procedures that support quality assurance in manufacturing processes. However, in many cases, AI systems can only really be deemed ready for companies and users to deploy once the systems’ trustworthiness, reliability and decision-making processes have been secured and demonstrated. Standards and laws for the demonstrably reliable use of AI, such as the EU AI Act, are already in preparation. The KI.NRW (AI in North Rhine Westphalia) competence platform is also working on a certified AI project with the objective of widespread use of secure AI. In this project, researchers from the Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS have teamed up with partners from the German Institute for Standardization (DIN) and the German Federal Office for Information Security (BSI) to develop standardized testing methods with the “made in Germany” label. Fraunhofer IAIS released an AI assessment catalog back in 2021. This practical document provides companies with a tool that they can use to evaluate and improve their own systems right from the development stage, so as to be prepared for future regulatory requirements. Independent testing organizations can also use the assessment catalog as a basis for product testing.

Some industry partners have already launched concrete testing services based on Fraunhofer IAIS expertise. Since mid-2022, Fraunhofer IAIS have been supporting Munich Re’s CertAI service as a technical partner. This testing service for AI applications enables companies to evaluate AI systems in terms of trustworthiness, and thus develop, use or purchase them as needed for their business success. In addition to defining and operationalizing quality requirements, the AI experts at Fraunhofer IAIS are also ensuring that the AI system evaluations conducted through the CertAI service reflect the latest progress in research and development.

KI-Prüfkatalog - Fraunhofer IAIS

Forum Zertifizierte KI

Inventory planning helps local pharmacies

Mit KI-basierten Prognosen lässt sich die Bestandshaltung für Apotheken optimieren
© iStock
AI-driven predictions can help optimize inventory management

Even in 2020, the Federal Union of German Associations of Pharmacists (ABDA) was reporting a steady decline in the number of pharmacies in Germany. However, a direct supply of medications and the opportunity to consult with a specialist in person are still cornerstones of the healthcare infrastructure, and not just for older people. Inventory planning is the decisive factor in enabling local pharmacies to compete with their online equivalents. This is why the Supply Chain Services working group at the Fraunhofer Institute for Integrated Circuits IIS has been researching ways of optimizing this process since the beginning of 2022. The group is concentrating on providing AI-powered predictions of the demand for various medications, taking into account the impact of seasonal factors and patterns, along with the recurring needs of regular customers.

A mathematical optimization model combines the predictions with restrictions such as the storage space and current ordering conditions, so that it can provide the optimal decision for what to order. This makes it possible to serve customer needs directly while simultaneously keeping the amount of capital tied up in stored goods to a low level. The ordering procedure developed in the project is automated to a great extent, which leaves more time for specialist personnel to spend on consultations with customers.

When optimizing pharmacy inventory planning, the research team was able to refer to previous results from projects that focused on wholesale distribution in the sanitation, heating and air conditioning sectors. The pharmacy inventory planning project is being funded until the end of 2024 by the Bavarian Ministry of Economic Affairs, Regional Development and Energy and the Bavarian Collaborative Research Program’s (BayVFP) Digitization funding line.