The “Mobile Repair Factory” project, carried out under the WIR! program and the WIR! alliance, brought together B.I.G. Technology Services, Keyence, Scansonic, Oscar PLT and the Technical Information Library (TIB) to develop and validate advanced defect‑detection technologies for turbine blades and welded components. The initiative was funded through the WIR! program, which supports regional innovation and technology transfer. Over the course of the project, the partners coordinated the acquisition of test parts, the deployment of sensors, the collection of data, and the integration of the systems into a mobile processing container.

B.I.G. Technology Services led the experimental work, providing a turbine‑blade specimen that had developed a fatigue defect of only about 0.2 mm. The defect was first captured with a Scansonic SCeye camera, which delivers high‑resolution video of the surface and the weld pool. The camera system, equipped with a Photonfocus EM‑D1312IE‑160‑LC12 sensor and a Computar MLH‑10X zoom lens, was illuminated with a dedicated light source to optimise contrast for the subsequent AI‑based image analysis. The captured weld‑process footage served as training data for a machine‑learning model that could recognise typical process faults in real time.

For geometric measurement, a Keyence LJ‑X8080 laser sensor was mounted on a robotic arm and scanned the blade surface. The sensor operates at 405 nm, delivers a 10 mW output, and offers a repeatability of 0.5 µm in the Z‑direction and 1 µm in X. ScanView software from Dr. Thomas von Chossy visualised the data in 3‑D and enabled the extraction of the defect’s dimensions. By varying the incidence angle and the scan direction, the team demonstrated that an angled approach improved defect visibility, although it slightly reduced overall measurement quality.

The project then expanded to weld inspection. A Scansonic QH6D sensor, essentially a Keyence LJ‑X8080 with enhanced protection, was used to scan the geometry of seven adjacent welds simultaneously. The QH6D offers a measurement range of ±34 mm (Z) and 72 mm (X), a repeatability of 1 µm (Z) and 3 µm (X), a frequency of up to 16 kHz, and a resolution of 25 µm. Using QH6D‑View software, the team measured weld width, height, spacing and profile, and identified edge‑crack artefacts that were later eliminated by fine‑tuning exposure time and laser power. The final scans accurately reflected the true geometry without scan‑induced artefacts.

After validating the sensors in the laboratory, the SCeye camera and QH6D sensor were installed in a mobile processing container. Control boxes were mounted, power supplies were secured, and the systems were aligned to the processing head. The SCeye camera was positioned above the weld, with additional illumination precisely aimed at the weld seam. Operation was managed through Jockey software on a dedicated PC, enabling real‑time data acquisition and monitoring.

Throughout the project, the partners maintained close collaboration: Keyence supplied the laser sensors and technical support; Scansonic provided the SCeye camera, QH6D sensor, and software; Oscar PLT supplied the weld samples and process knowledge; B.I.G. Technology Services coordinated the experiments and data analysis; and TIB granted rights for metadata creation and public dissemination. The project achieved its milestones—sensor integration, defect detection, AI training data generation, and mobile deployment—demonstrating a robust workflow for rapid, on‑site inspection of critical components in the manufacturing sector.