Qualification Considerations of Machine Learning Based Tools for Avionics System Development

Machine learning (ML) technology has advanced significantly in recent years, enabling its practical application in various domains, including avionics. However, before ML can be integrated into avionics systems, it must comply with certification or qualification standards. While much attention has been given to the certification of ML applications, the qualification of ML-based tools has been less extensively studied. This paper explores the qualification of ML-based tools in avionics system development, examining the unique characteristics of ML and their impact on tool qualification. We propose a methodology for qualifying low-criticality ML-based tools, aligned with the DO-330 software tool qualification standard, treating the ML tool as a black box. To demonstrate this approach, we present a case study of a ML-based avionics display testing tool.