How can we maintain an overview of supply chains in times of crisis and ensure planning security for companies? Initial situation As a globally active company, Phoenix Contact works with many thousands of suppliers every day. Therefore, a stable and transparent supply chain is essential to ensure a functioning business. The Corona pandemic and the current war in Ukraine show how fragile existing supply chains are. Raw materials are scarce and expensive. Important components are partially or completely unavailable. For companies, this results in extremely rising prices, severely restricted ability to deliver or even the risk of complete inability to do business. For some time now, it has no longer been possible to speak of planning security. Objective The objective of the Challenge is to develop a prototypical solution for the near-real-time visualization of expected material flows, incipient events or coordination deficits within supply chains. This should help to forecast at which location which goods are expected to arrive in which quantity and which possible alternatives are possible to minimize risks. As a result, Phoenix Contact GmbH & Co. KG hopes for increased transparency across the entire supply chain, increased planning accuracy and delivery capability, and more time to react to unplannable events in the value creation stages.

How can AI be used to detect error patterns from sensor data in small bakery production and use them to optimize process parameters? Initial situation: The company WP Kemper GmbH builds machines and systems for industrial cookie production and offers its customers the entire process chain from dough production to dough preparation and baking. Especially in food production, the aspect of sustainability is of crucial importance against the background of increasing population figures and raw material shortages. Changes in environmental conditions and fluctuations in flour properties lead to altered dough characteristics, which quickly result in reduced dough quality and thus in rejects. AI-based early detection of defect patterns with subsequent optimization of process parameters can significantly reduce scrap. Objective: WP Kemper provides production data of a dough sheet line for the production of dough pieces, which sporadically contain target-actual control differences of the dough thickness during the rolling out of the dough sheet. Using AI, error patterns within all available process parameters and sensor data of the upstream processes are to be identified in order to find the cause of these control differences. Based on the identified error patterns, suggestions for automated optimization of the production parameters will be developed in order to reduce scrap. What is the context of the data? OpcUA sensor data / From the production of dough products /. Time series data / Characteristics including but not limited to: Time stamp, dough inlet thickness, and dough passage thickness. What does the data indicate? Label available for part of the data / Distinction between normal and abnormal / Abnormal data indicate dough strands that are too thick or thin What should be the solution to the challenge? Step 1: AI-based pattern recognition; detect deviation in production. Using AI, error patterns are to be extracted and a conclusion drawn as to the cause. In the later course of the project, not part of the data challenge, a sensor concept is to be developed with which an automated evaluation of the product quality can be carried out. Which IT systems are involved? Sensor data available via a dashboard

Initial situation: The management consultancy Unity and the mechanical engineering company Kraft have joined forces to address unused potential in machine-based production. Existing machines are often not optimally utilized. Especially in interlinked production lines, there are waiting or idle times due to retooling, malfunctions and due to individual stations that represent bottlenecks. This is not only lost revenue, but more and more often a form of waste of resources, which in times of rising raw material and energy prices is more and more in focus. Objective: The objective of the Challenge is to continuously and automatically optimize machine-based production in a sustainable and resource-saving manner using AI. AI-controlled production planning, identification and optimization of bottle necks of individual machines and interlinked production lines, as well as simulation of throughput, process and setup times with different configurations of intermediate storage and sorting stations. No rigid control system is to be created, as these already exist in sufficient numbers. The solution sought should use AI based on data analytics and machine learning, with historical production and machine data and the respective current conditions to learn what an optimal solution looks like and must also be able to react to short-term disruptions. This ensures resource-efficient value creation and resilience through forward-looking production planning. What is the context of the data? Energy consumption data collected from various measuring points within the Hesse Mechatronics plant as well as reference and feed-in data from the PV plant. What does the data indicate? Energy consumption as well as energy feed-in and generation. What should the solution to the challenge look like? Assistance system to support energy management managers. What IT systems are involved? Data is collected and managed by a proprietary system.

Initial situation: Against the backdrop of the ongoing climate and energy crisis, ensuring a secure and cost-effective energy supply is an ever greater challenge for companies. They are therefore increasingly motivated to produce electricity themselves, to hold reserves and to make demand plannable. In order to be able to react to outages or extreme price conditions in the future, the companies Hesse Mechatronics and Weidmüller are looking for solutions for intelligent energy management. Objective: As part of the Makeathon, a prototype for an intelligent energy management system is to be developed. On the one hand, this should be able to forecast future electricity production, e.g. on the basis of weather data. On the other hand, the prototype should make it possible to predict future electricity demand, e.g. taking into account consumption data, vacation periods or weather conditions. Taking into account electricity supply and demand or in the event of a power outage, the system should thus provide recommendations for action, for example, to shut down consumers, shift production and attendance times, or dimension storage facilities. What is the context of the data? Energy consumption data recorded from various measuring points within the Hesse Mechatronics plant as well as reference and feed-in data from the PV plant. What does the data indicate? Energy consumption as well as energy feed-in and generation. What should the solution to the challenge look like? Assistance system to support energy management managers. What IT systems are involved? Data is collected and managed by a proprietary system.

What is the context of the data? Measurement data from machines and tools used in the milling process What does the data indicate? Indications during production that lead to wear and problems of the production tools What should be the solution to the challenge? Detection of anomalies in production that indicate wear. Also, a warning as to whether the tool should be replaced before the next process. Which IT systems are involved? Sensor data is stored in a database at inno focus.

How can AI accelerate the creation and use of IP rights? Initial situation: Intellectual property rights play a central role in securing competitive advantages. In this context, companies are faced with the challenge of examining in detail an almost unmanageable, existing patent landscape during new and further product development. At the same time, there is the challenge of gaining knowledge of the unauthorized use of existing proprietary rights or new proprietary rights within the company's own scope of protection. Details are often decisive, so that the process of patent identification and examination is always associated with high expenditure. Objective: The objective of the Challenge is to develop an AI solution for the automated identification and analysis of relevant patent protection rights. The special feature of the solution lies in particular in the recognition of the relevant context and the corresponding comparison with existing patents. As a result, Phoenix Contact GmbH & Co. KG hopes to achieve massive time savings - and thus efficiency in the processing of tasks and, in particular, speed and agility in the registration and use of patents. Inventions and their protection against imitation are extremely important for companies in order to secure jobs and economic growth. Otherwise, any company could copy an invention, which would effectively devalue it. However, effective protection by a patent is only possible if it is possible to show a difference from the already known prior art. For this purpose, it is necessary to analyze patents with regard to the respective semantic technology concepts and to identify the relevant prior art for the respective invention on a context basis. With this knowledge, a patent application can be focused specifically on the core of the invention. The solution of the task below is the beginning of a new era for us as a company, through which a fast and equally effective patent protection is achieved by AI-based context-sensitive technology. What is the context of the data? a. Abstracts of patents b. From patent database What should be the solution to the challenge? Assignment of patents to main class: Each patent office uses predefined patent classes (IPC classes) into which patent applications are sorted by an official technical expert depending on the subject matter. This classification allows a quick assignment to a respective responsible examining office. The IPC classes are arranged in a tree structure and subdivided into main classes and subclasses. These patent classes can also be used to locate nearby prior art. Using a limited selection of patents assigned by the patent office to the preselected subclasses of the IPC main classes H01, H02, H03 or H04, a machine learning method is to be trained and evaluated on test data. This will provide evidence for a context-sensitive mapping.