Smart Blade Inspection
& Robotic Repair

Clobotics runs a five-stage AI pipeline on every inspection image: blade classification, blade segmentation, then defect recognition using sliding-window parallel analysis. The blade classification model achieves 99% accuracy, segmentation 98% accuracy, and defect recall exceeds 95% overall — with major-defect recall above 98% and major-defect precision above 80%. In practice, on a 400-image turbine inspection the AI automatically clears roughly 240 images as defect-free, reducing human review time from 1.5 hours per turbine to 25 minutes.

Accuracy in defect detection is a data problem as much as a technology problem. Clobotics has conducted blade inspections across 40+ countries — building one of the largest real-world blade defect datasets in the industry. Every completed inspection is archived into a data lake that feeds continuous model retraining on a PDCA loop. The result is a model that improves with every campaign, not one trained on a static dataset and left unchanged.

Most inspection vendors cover external blade surfaces only. Clobotics inspects the full blade: IBIS drones capture the exterior at sub-millimeter resolution (avg 0.66mm/pixel, best case 0.25mm/pixel); the Kiwi robot crawls inside the blade for internal structural defect detection; and Hummingbird performs lightning protection resistance testing — all three data streams are cross-referenced on the IRIS platform to give a complete internal-external blade health picture. No other vendor offers all three in a single integrated workflow.