Featured Product
This Week in Quality Digest Live
Metrology Features
Dirk Dusharme @ Quality Digest
The DSX1000 digital microscope provides 20X to 7000X magnification in multiple observation modes
Ryan E. Day
Olympus’ LEXT OLS5100 microscopes can spearhead the move to increased productivity and profitability
Bryan Christiansen
And what are its applications?
Joe Chew
New system holds the key to unprecedented accuracy in beam control

More Features

Metrology News
Update offers primary support of QIF 3.0 with bi-directional translation and validation of the MBD portion
The reshoring move will simplify operations
Alerts scheduled by actual use of the machine rather than number of days. See at PackExpo, Sept. 27–29, 2021.
Additional option for interfacing and controlling innovative mobile surface measurement system cuts implementation costs in half
Booth No. 1120, Donald E. Stephens Convention Center, Rosemont, Illinois, Oct. 26–28, 2021
CNC lens turret allows nine automatic, programmable magnification changes within a CNC program
Features high-resolution, high-color contrast display to mimic smooth motion of an analog meter
20 point clouds per second, maximum deviation of 0.6 mm at object distances of 50 cm

More News

Raghava Kashyapa

Metrology

Automated Surface-Defect Detection of Bearing Rollers

Vision and AI quickly help detect and quantify bearing defects during manufacture

Published: Tuesday, August 3, 2021 - 12:03

Bearings are important components of mechanical equipment. They are specifically designed to convert the direct friction from parts in relative rotation into rolling friction or sliding friction of the bearing. As a result, bearings are extremely important in reducing the friction coefficient and ensuring the long-term stable operation of a machine.

The bearing surface and bearing rollers both have an important impact on the installation performance, use, quality, and life of the bearing. Common defects on a bearing roller’s surface, such as wear, cracks, bruises, pitting, scratches, or deformation, can lead to machine vibration and noise, accelerate oxidation and wear, and even damage the machine. It is thus paramount to inspect the surface of the bearing as well as the bearing rollers to prevent defective products from entering the market.

The assembly of bearings globally has been fully automated for the most part. However, the bearing roller inspection and surface inspection of the bearing before and after assembly is still largely based on manual inspection. The method is labor-intensive, inefficient, costly, and easily affected by such factors as inspector qualification and experience, visual resolution of the naked eye, and fatigue.

Common types of defects found on bearing rollers

Pairs of tapered roller bearings are used in numerous vehicle wheel bearings, where they must cope simultaneously with large radial and horizontal axial forces. Tapered roller bearings are commonly used for moderate speed, heavy-duty applications where durability is required. Other real-world applications include agriculture, construction and mining equipment, sports robot combat, axle systems, gearboxes, engine motors and reducers, propeller shafts, railroad axle-boxes, differentials, and wind turbines.

As with any typical mechanical production process, defects are natural in the production of bearing rollers. Below are some of the most common bearing defects.

Face end defects

Not cleaned up: Contamination can be caused by foreign substances getting into bearing lubricants or cleaning solutions. These substances include dirt, abrasive grit, dust, steel chips from contaminated work areas, and dirty hands or tools.

Shallow recess: A lack of appropriate recess may have consequences such as entire ball bearings not fitting as intended.

Grinding damage: Grinding damages shorten the fatigue life and can cause critical failures in dynamically loaded, critical components.

Outer diameter defects

Spiral effect: This is an undesirable effect that is hard to avoid in the manufacturing process if the roller bearing possesses any kinds of defects.

Defects in a production unit are a natural and common byproduct of the manufacturing process. However, the challenge lies in separating defective pieces from the perfect pieces. Let’s look at one of Qualitas Technology’s clients and how we helped them overcome their issues.

Our client’s background

Our client is the market leader in their industry and engineers, manufactures, and markets bearings, gear drives, belts, chains, couplings, lubrication systems, and related products. The company offers a spectrum of powertrain rebuild and repair services. As the leading authority on tapered roller bearings, they have been successful in applying their expertise in metallurgy, tribology, and mechanical systems to improve the reliability and efficiency of equipment, machinery, and vehicles worldwide.

Main issues

Rollers are the main pressure-bearing part of rolling bearings and are easily damaged due to defects and other factors. If there are defects on the roller surface, the stability of the bearing will be heavily reduced during use. Among the rolling bearings, deep groove ball bearings are mainly used in small and medium-sized equipment, while roller bearings are widely used in medium-sized and large machines. They are widely used in passenger transportation, aerospace, and other transportation fields, as well as agricultural machinery, industrial machinery, medical equipment, and other related machinery industries.

Defects naturally occur on the surface of bearing rollers during the production pipeline. The defects are mainly observed on the cylindrical surface, chamfers, and end surfaces. Common defect categories include damage and scratches caused by mechanical collision; corrosion caused by mechanical ageing; material lacking at the chamfer; and defects caused by insufficient grinding during the production process.

Although our client was aware of these problems in their production process, they had no way of ensuring that each and every defective piece was successfully separated and discarded. It was clear that the accuracy of human inspectors varies a lot and is subject to factors such as fatigue, turnover, and inconsistent classification of defects. These inconsistencies led to quality variations across operator shifts. With a production speed of almost three rollers per second, it was almost impossible for proper manual inspection to catch all defects.

These problems had some serious implications. The low-quality output impaired the company’s brand image and created an overall distrust amongst customers and clients. The defective rollers could result in leaks and cost their customers efficiency. Surface defects also reduced the lifespan of the bearings. The recall rates were higher than desirable.

Bearing inspection and bearing roller inspection

We suggested that the company use AI and machine vision-based technologies to automate the bearing and bearing-roller inspection process.

We applied our paradigm of the 4I methodology—an abbreviation for install, instruct, inspect, and improve—to help the company bring about the desired outputs.

The core of our solution worked in the following way.

Install: A set of three cameras at different angles with red-panel lighting was set up to capture the defects on the rollers. Red lights are used instead of white to minimize reflection on the bearing surface, which could lead to overlooking some surface defects.

Instruct: A solution was developed using the acquired images of the roller defects. The machine vision system was trained to recognize each type of defect with the help of a different set of images.

Inspect: The AI-based anomaly-detection technique was used to correctly identify the defects on the roller body and face.

Improve: Deep learning (DL) programs were created to train the machine vision system to understand the various surface anomalies (defects) on the bearing rollers. The results are displayed on the user interface in real-time.

We face two key challenges during this project. We had to design a system that was fast enough to accommodate the high production speeds as well as train the model to inspect various types of defects while maintaining an accuracy of more than 85 percent. Using state-of-the-art technology and research, we achieved an accurate and fast bearing-inspection system, successfully satisfying all the stakeholders involved.

Conclusion

With the machine vision system, all the defects were identified correctly and defective products were rejected. This success was strengthened by the following special features:
• The accuracy of identifying NCU, shallow recess, grinding damage, and spiral defects was 98 percent
• The inspection cycle time was a quick, 150 to 200 milliseconds
• No manpower was required for the inspection

Discuss

About The Author

Raghava Kashyapa’s picture

Raghava Kashyapa

Founder and Managing Director of Qualitas Technologies Pvt Ltd, a successful Machine Vision Solutions company in India. A passionate technologist, Raghava completed his Engineering from Bangalore University and a Master's from Washington State University where he was a research fellow. After several years in leadership roles at Epic Systems and Microsoft, Raghava returned from the US in 2011 and started operations in Bangalore. Raghava is passionate about reading and playing the guitar.