Interspectral coordinates SEK 7.4M Vinnova TRUSTAM project for federated learning in AM quality control

Interspectral AB is coordinating TRUSTAM (Trusted Federated Intelligence for Additive Manufacturing), a two-year project backed by Sweden's innovation agency Vinnova with a grant of SEK 7,413,300. Running from April 2026 to April 2028 under the Applied AI for industry call, the effort pairs Interspectral's AM data-fusion and visualization software with Scaleout Systems' federated-learning stack, AMEXCI's AM service production, and Saab as the industrial end user. The consortium will build a demonstrator for local model adaptation and federated model improvement aimed at calibrated, real-time, and traceable quality assurance in security-critical additive manufacturing environments.

Quality-control AI for industrial AM has long been constrained by data sovereignty: defect and process datasets sit behind OEM, bureau, and defense firewalls that block conventional central training. Federated learning addresses that bottleneck by improving models without moving raw production data-an architecture that fits Saab-class programs and multi-site service networks better than open data lakes. For Interspectral, TRUSTAM is a product-path trial: embedding federated workflows into its analysis layer could turn AM Explorer-style tools into shared quality infrastructure rather than single-factory dashboards.

The practical test is whether the 2028 demonstrator can survive real production governance-model provenance, audit trails, and acceptance by qualified end users-not only offline accuracy metrics.