The Delta assembly line at Cummins depends on human visual inspection to identify external defects on turbocharger assemblies. Manual inspection introduces inconsistency, slows throughput, and degrades accuracy as operators experience fatigue over long production shifts. As a result, small defects such as surface scratches, corrosion, and geometric misalignments may pass undetected, leading to quality escapes, customer complaints, and increased warranty costs.
Cummins lacks an automated inspection solution capable of meeting its production and quality requirements. Existing commercial systems struggle with reflective metal surfaces, complex turbocharger geometry, short inspection cycle times, and integration within an active assembly line. Current tools also fail to provide adaptable detection methods or traceable inspection records required for modern, high-volume manufacturing environments.
This project investigates automated, non-contact inspection technologies capable of improving defect detection reliability on turbocharger assemblies. The goal focuses on identifying and evaluating sensing and analysis approaches that enable consistent, high-precision inspection, reduce dependence on manual judgment, and strengthen overall manufacturing quality control.
Cummins requires a non-contact, automated inspection system capable of detecting multiple physical defects on turbocharger assemblies; including surface damage, dimensional inaccuracies, misalignments, labeling errors, and missing or improperly installed components. The inspection must occur at a fixed station while the turbocharger is rotated for full 360° analysis, and the total cycle time must remain under 80 seconds to avoid interrupting production throughput. These parameters form the critical performance constraints of the problem.
Key factors contributing to the inspection challenge include:
Highly reflective and irregular metallic surfaces
Complex geometries with deep pockets and occlusions
Small but critical defects that are hard to detect visually
The need for repeatable, objective inspection data
Integration within an existing, fast-paced manufacturing workflow
Current commercial inspection systems cannot meet these requirements. Off-the-shelf cameras and rule-based vision tools struggle with reflective surfaces, cannot reliably capture all viewing angles, and lack the flexibility to detect new defect types without expensive re-engineering. They also cannot guarantee sub-80-second cycle times while maintaining accuracy across diverse turbocharger variants.
Because of these limitations, a tailored, AI-driven inspection system is required, designed specifically for Cummins’ Sabre turbocharger. Our team is responsible for developing this solution because it demands a combination of advanced computer vision, AI defect classification, synchronized rotational scanning, and custom mechanical integration—capabilities not provided by standard industrial systems.
The final system will ensure consistent defect detection, reduce rework, strengthen quality assurance, and uphold customer satisfaction for Cummins and its associate suppliers.