FLARE - Fast and reliable human-centered-AI for high-rate non-destructive evaluation

In the transportation sector, new trends such as urban air mobility, where forecasts predict the use of 19,000 units by 2030, require extremely high production rates. To achieve this, production processes must be digitalized and automated while maintaining high safety standards. For the first time, non-destructive testing (NDT) methods should be used, which can inspect even the most complex components quickly and using non-contact imaging. As the wall thicknesses of UAMs are thinner than in civil aviation (less than 10 mm), near-surface methods such as active thermography (IRT) can be used extensively for the first time. Due to the higher part throughput (up to 1 part/min), the evaluation of the measurement results becomes a costly bottleneck.

The project aims to develop advanced technologies for automated defect detection based on artificial intelligence (AI) to reduce the evaluation time required per component to less than 50% of the measurement time. Depending on the legal requirements, the algorithms can be implemented as an assistance system for the operator or fully automatically. The objective is to develop human-in-the-loop (HiL) interaction concepts for IRT for the first time. By forming human-machine teams, humans are better integrated into the decision-making and learning process of AI. This enables an effective and legally compliant use of AI and can compensate for individual weaknesses of inspectors. The models and interaction concepts will be integrated into the smart and fully integrated inspection system of the start-up VOIDSY and made directly available to the industrial partners FACC, ENGEL and RORA. Resource- and energy-efficient decentralized computing is intended to solve the latency and security problems associated with cloud computing. With the support of AI systems, inspectors are expected to reach a decision up to 20% faster and detect errors 20% more reliably.
As a result of the project, inspection costs can be minimized and the competitiveness of Austrian aviation suppliers strengthened. By improving inspection technology, component quality in production can be increased and the probability of failure minimized. The project partners are focusing on sustainable manufacturing processes with minimal waste, such as organosheets with overmoulding and the optional use of recyclate (ENGEL), automated fiber placement (RORA) and pick & place for automated production (FACC). As there is still little experience with these innovative processes to achieve high cycle rates, 100% accompanying quality control is a top priority to achieve market entry. IRT testing enables more efficient production processes (through scrap reduction, process optimization and statements about the repairability of components) and thus the closing of material cycles.