Solving the Data Issue for AI-based Inspection Tasks in Industry

The implementation of AI-based visual failure detection in industrial applications faces a significant challenge: the availability of high-quality training data. In collaboration with Siemens Energy, Gestalt Robotics presents an innovative approach to turbine blade repair inspection that minimizes reliance on real-world training images while maintaining industrial-grade robustness and performance.

Our solution leverages fully automated synthetic data pipelines and generative AI methods, integrated within an extensive MLOps framework. A digital twin ensures continuous data availability, enabling scalable AI training and validation in a highly controlled virtual environment. By incorporating edge, fog, and cloud computing, along with advanced networking approaches and digital twins, we establish a seamless interaction between AI-driven defect detection and real-world industrial processes.

This presentation will showcase how AI, synthetic data, and automation revolutionize quality assurance in production and maintenance, setting new standards for efficiency and reliability in industrial AI deployment.