How exactly does an artificial intelligence reach a certain decision or arrive at a specific prediction? Trust in AI technology depends largely on whether users can understand the decision-making processes of an AI. This also applies to the application of artificial intelligence in industrial production. After all, AI technologies harbour immense potential to organise industrial production processes more efficiently, safely and sustainably - however, they are currently being used only hesitantly. In the EU project XMANAI, 15 project partners are working on improving the comprehensibility of AI decision-making processes for industry and production. To this end, a platform is being created under the heading of “Explainable AI” which provides services for the creation and execution of AI models for industrial production.

XMANAI demonstrators: Four companies test AI use cases in pilot operations

The XMANAI research results are now being validated under real-life conditions through demonstrators: Four companies involved in the project test XMANAI technologies for various use cases under real conditions - the car manufacturer FORD, the manufacturer of household appliances WHIRLPOOL, the agricultural machinery manufacturer CNH Industrial and UNIMETRIK, a company for solutions in the field of measurement technology. The diversity of the demonstrators enables the project participants to take into account heterogeneous industrial plants as well as different data sources and requirements for AI models. The specific use cases range from demand forecasts and the optimisation of plant performance to predictive maintenance and the reduction of downtimes.

AI decisions simple to understand via a connected platform

With the integration of all demonstrators into the XMANAI platform, one of the most important steps of the project has now been completed: All project partners can now use and test the XMANAI services in their final version - via a single platform that enables access and management of the necessary data. At the same time, each demonstrator is connected to the platform via its own app with a user interface developed specifically for the explainability of AI. This enables functionalities, visualisations and metrics that are specially tailored to the respective use case. Knowledge graphs and visual representations of the effects of various parameters (e.g. heat maps) aid in understanding the relationship between the data supplied and the decisions taken by the AI.

The Digital Public Services (DPS) business unit at Fraunhofer FOKUS contributes as a scientific partner for real-time data management. XMANAI will continue until April 2024 and will be continuously developed over the coming year.