AirFusion - Advanced visual inspection for wind turbine blades

Client:

Industry:

Renewable Energy

WHAT WE DID:

Solution Delivery, IoT

RESULTS:

Improved accuracy of AI-based visual inspection for wind-farm operators

OVERVIEW

The project focused on developing an advanced visual inspection platform for AirFusion to automate the assessment of wind turbine blades using high-resolution imagery captured by drones.

The solution applies deep neural networks based on CNN architectures to perform object detection, semantic segmentation, and defect classification.

It identifies cracks, erosion, delamination, and other structural anomalies with high accuracy. The platform was delivered as a multi-tenant SaaS application, enabling scalable onboarding of multiple wind-farm operators while ensuring secure and isolated data handling for each client. AI-driven analysis, combined with automated reporting, significantly reduced reliance on manual inspections, improved the consistency of defect detection, and shortened inspection cycles. Following the acquisition of AirFusion, the technology and its further development continued in cooperation with mCloud.

See our approach

Results

High detection quality

The delivered system achieved high detection quality, with precision and recall rates exceeding 90% across key defect categories, including cracks, delamination, erosion, lightning strikes, and other surface anomalies.

An optimized inference pipeline and scalable compute architecture enabled near real-time processing, handling up to 500 images in under two minutes without affecting performance for other tenants operating within the multi-tenant SaaS environment.

High detection quality

90%

cross key defect categories

500

images in under two minutes

20+

tenants, each benefiting from secure data isolation

Why 7bulls

Engineering expertise behind the solution

7bulls provided the technical expertise and delivery capabilities required to design, build, and operate a system of this complexity.

The team brought hands-on experience in computer vision and deep learning, including training and optimization of CNN models for detection, segmentation, and classification on large and heterogeneous datasets. A strong cloud-native engineering background enabled the design of a multi-tenant SaaS architecture that scales horizontally, handles burst workloads, and maintains predictable operational costs.

Established MLOps practices covered dataset management, model versioning, automated validation and monitoring, ensuring reproducible results and controlled model updates. The project was delivered as a production-ready system with a strong focus on reliability, fault tolerance, maintainability, and security, supported by a cross-disciplinary team spanning data engineering, AI development, backend and cloud architecture, and deployment.

Looking to apply AI and scalable cloud architecture to complex inspection or monitoring processes? Let’s discuss how similar approaches can be adapted to your operational environment.