Challenge
Atlas Copco is a global industrial leader specialising in manufacturing advanced air compressors and industrial machinery components, including high-precision rotors and casings essential for compressor operations. In their relentless pursuit of operational excellence, Atlas Copco sought to significantly reduce costly defect-related returns by accurately identifying product defects before shipment. Traditional manual inspection methods were insufficient, resulting in occasional missed defects which resulted in high return costs and potential reputation damage.

Key underlying factors causing this critical quality control challenge included:
- Limited defect visibility: Subtle defects were difficult to consistently detect through manual inspection due to the highly reflective surfaces of rotor components.
- Scarcity of defect examples: The low defect rate (<1%) resulted in insufficient data for reliable training of computer vision models.
- Operational inconsistency: Manual inspection introduced variability and subjective judgments, affecting defect definition and annotation clarity.
Atlas Copco’s ambition was clear: ensure consistent defect-free deliveries to maintain customer trust. Achieving this required an automated visual inspection capability leveraging AI, something manual inspection alone simply could not deliver.
Jeroen Boeye – Head of AI, Faktion
Solution
Faktion designed and implemented an advanced, AI-powered computer vision solution to automate the detection of subtle defects, significantly enhancing quality control reliability. Central to this was the establishment of a comprehensive MLOps framework based on Azure Machine Learning, empowering Atlas Copco to consolidate and standardise all machine learning initiatives company-wide.
The innovative aspects of the solution included:
- Deploying advanced computer vision models (segmentation and object detection) capable of accurately detecting microscopic defects on reflective rotor and casing surfaces.
- Building a robust and reusable MLOps environment, enabling efficient and standardised training, fine-tuning, deployment, and maintenance of ML models across diverse projects.
- Enabling seamless edge deployment via NVIDIA Jetson devices for real-time defect detection integrated directly into Atlas Copco’s operational machinery.
- Integrating AI-generated quality insights into existing industrial software (Kepware & Ignition), ensuring streamlined operator visibility and immediate actionability.
The combined solution empowered Atlas Copco’s internal teams by reducing dependency on external experts. Teams could now independently train, iterate, and fine-tune their ML models rapidly and consistently, dramatically accelerating AI adoption.
Approach
Faktion’s approach to delivering the solution included:
- Conducting a thorough initial process and technical assessment to define clear defect criteria, data collection strategies, and infrastructural requirements.
- Developing an end-to-end standardised ML pipeline within Azure ML, providing a structured method for annotation, model training, evaluation, and deployment.
- Implementing robust feedback loops with Atlas Copco’s quality experts, iteratively refining model accuracy, ensuring clear defect annotation guidelines, and enhancing overall AI performance.
- Executing meticulous validation and operational integration, including edge computing infrastructure, ensuring seamless interoperability with industrial control systems and visualisation platforms.
This structured, collaborative process provided Atlas Copco a sustainable, scalable foundation for ongoing AI-driven quality improvements across the organisation.
Outcome
Since implementing Faktion’s computer vision solution, Atlas Copco achieved critical improvements in the quality inspection workflow and operational consistency:
- Automated defect inspection integrated within the production line, eliminating manual inspection variability.
- Established an operational AI-driven inspection pipeline for future scalability, easily adaptable to new product lines or changing requirements.
- Reduced inspection latency, enabling faster decision-making and shorter production cycles.
The broader impact on Atlas Copco’s business included strengthened customer confidence, reduced risk of costly product returns, and enhanced competitive differentiation through increased reliability and product quality consistency.