Defect detection in mass production using images

Traditionally, monitoring the quality of manufactured products has been a time-consuming process with inconsistent accuracy. With recent advances in deep learning, this process can be easily automated. At the end of this article you’ll find a link to how to build your own defect detection model using the Peltarion Platform.

... how to find a needle in a haystack?

02/ What can be achieved?

Using AI for automated detection of production defects allows for reduced production cost, speed and accuracy.

03/ The problem

Traditional product quality testing is slow and inefficient which can lead to production bottlenecks. With human inspection, the accuracy largely depends on each particular inspector and traditional automated systems are both expensive and difficult to implement.

04/ Opportunity for deep learning

A deep learning approach will increase efficiency by allowing quality control to be integrated into a fully automated production line and enable more accurate analysis of the quality of each individual part. The fact that these deep learning models are automated and quick allows them to be implemented in more stages of the production to avoid defective parts from going through additional processes in the production process.

05/ How is the AI model implemented?

Cameras mounted on the production line feed images to the model. Modern production lines can then act upon the prediction made by the model to either remove the part from production or allow it to continue to the next step.

06/ Data requirements

The model is trained on images of manufactured parts, classified either as defective or non-defective.

07/ Test it yourself

Use the app below to test the functionality of a model we built using the Peltarion platform to detect defects in metal castings. We have removed 10 images from the main dataset in the data library. This means that these images have not been seen by the model at all so far. Click here to download the demo images.

Click the icon below to upload an image of the part. Once you have uploaded an image, you can click it again to try another one.

08/ Where to learn more?

If you're interested in learning more, have a look at the in-depth article about defect detection. You can also read about Peltarion has been working with the Industrial 3D printing company Amexci to create a proof of concept around quality control in Industrial 3D printing.

Detecting defects is an example of binary Image classification. To learn how to do this for this use case, follow the Defect detection tutorial.

With good input data, deep learning can do a whole lot more than simply identifying defective parts. The technique used in the Car damage assessment tutorial could be used to classify different types of defects and the Skin cancer detection tutorial can be used to identify the locations of a particular condition.