Automated Surface Defect Inspection – Drew Schiltz, Zeiss Industrial Quality Solutions

INTRODUCTION

Current universal methods for quality control of surface defects largely rely on the human visual inspection process which is notoriously subjective, unrepeatable and unreliable. The implementation may look slightly different across various industries, but the overarching process remains the same; the visual inspector visually identifies defects on the surface then judges if the defect is acceptable based on a previously established quality standard. Results from this process can vary significantly from inspector to inspector. Even for the same visual inspector, results will drift based on time of the day and day of the week. Although there are numerous reasons to move away from this outdated process there are also underlying complexities that have thwarted many attempts to date, which is why human visual inspection remains a pervasive component of most manufacturing processes.

As manufacturing continues to evolve into an era of digitalization, the visual inspection process must be automated with a suitable level of information acquired, analyzed and presented for immediate feedback. As such, advanced imaging techniques have been developed to acquire various layers of information necessary to observe numerous defect classes on varying surfaces and finishes. In conjunction with these innovative data acquisition methodologies, refined evaluations are rapidly being established, including intricate machine learning techniques. These two critical advancements are key enablers of inspection solutions that can not only replace but outperform the human visual inspection process.

TUTORIAL TOPICS

1. Overview of human visual inspection process.
2. Common defect and surface types
3. Fundamentals of optical imaging
4. Optical imaging for high gloss surfaces
5. Current technology: cameras and lenses
6. Industry case study: design of an automated imaging cell
7. Advanced imaging techniques for surface defect inspection
8. Classical Image analysis techniques for defect detection
9. Machine learning techniques for defect detection
10. Future of surface defect inspection
11. Practical experiments in surface defect image acquisition

TUTORIAL OUTCOME

After the tutorial, attendees will understand how visual inspectors acquire information and the underlying data that needs to be collected to automate this process. They will learn about the fundamentals of surface defect imaging and capabilities. Attendees will explore the basics of evaluation for defect detection and judgement, including differences between classical image analysis and machine learning techniques. Lastly, attendees will get to see real experiments in surface defect inspection.

Drew Schiltz, Zeiss Industrial Quality Solutions

In 2015, Drew joined Zeiss in an applications development role geared toward automating surface defect inspection. Drew has worked directly with customers, understanding their needs in the visual inspection process while developing in-line surface defect inspection solutions for the last 6 years. These applications have spanned numerous industries including aerospace, automotive, consumer electronics, semiconductor and nuclear energy. Drew is currently the technology and applications manager for SurfMax inspection systems at Carl Zeiss Industrial Quality Solutions in Maple Grove, MN.