Overview

With SonicTC 2026.2 we take the next step in evolving our software platform for acoustic quality testing. This release focuses on three core areas: AI-powered anomaly detection, inline performance and industrial connectivity.

ML Classifier: Higher Selectivity

The revised ML classifier uses a new ensemble method combining gradient-boosted trees with neural networks. Selectivity between OK and NOK improves by an average of 18 % compared to the previous version – using the identical training dataset.

The new algorithm excels particularly with acoustically "quiet" defects such as fine hairline cracks in ceramic parts or minimal imbalances in rotors. The false-positive rate drops by up to 40 %, directly reducing pseudo-scrap and increasing overall equipment efficiency.

Faster Inline Classification

An optimized signal pipeline and GPU-accelerated FFT computation reduce classification time per part by 35 %. For typical automotive applications this means: from 1.2 s down to under 0.8 s per part – fully compatible with series production cycle times.

PROFINET S2 Support

SonicTC 2026.2 now natively supports PROFINET IRT with S2 device class. This enables the test station to participate as a full member in highly synchronous automation networks – without additional gateway hardware. Integration into existing Siemens and Beckhoff environments has been successfully validated in three pilot projects.

Additional Improvements

  • New dashboard widget for trend monitoring across shifts
  • Export of test protocols in VDA-compliant XML format
  • Improved REST API for CAQ integration (SAP QM, Babtec)
  • Bug fixes and performance optimizations in the signal recorder

Upgrade Path

Existing SonicTC customers receive the update free of charge through their standard maintenance contract. Migration of existing classifiers is automatic – your training data and models are fully preserved. Contact us if you require an assisted migration.