How to investigate the defects of the steel bars

Challenge

The automatic inspection system promptly alerts the customer to the presence and type of defects in the product.

The design of the inspection system has been conceived so that it can be easily integrated into existing production lines, immediately upstream of the packaging process. This way, it is possible to equip the plant with this new function without the need for any plant alterations.

Providing intelligent process control systems puts the customer in a position to supply high-quality products, thereby improving their competitive positioning in the industry.

Approach

We have developed an automated product inspection system that integrates hardware engineering solutions to obtain reliable product images under harsh environmental conditions, coupled with AI software solutions that ensure fast and accurate classification of defects in these images.

The system's hardware is optimized for product speed and resolution, while the lighting and imaging conditions are carefully designed to ensure proper visibility of defects in the captured images. Moreover, the positioning of the imaging system along the rolling line is strategically chosen to maximize defect visibility.

Thanks to these design criteria, we have achieved a highly sensitive device capable of capturing indisputable images indicating the presence or absence of defects. The system operates in real-time, capturing a large number of images per second (hundreds of megapixels) and swiftly processing them. This ensures that all defects on the material are detected and reported.

By utilizing AI techniques in image processing, we obtain reliable results even in classifying the type of defect detected. This enables the generation of selective reports, indicating the nature and cause of each defect. As a result, end-users have a basis for taking corrective action on the processes to reduce the occurrence of defects.

Our defect classification has been optimized for identifying typical defects specific to this process. This has been made possible through our in-depth knowledge of this process and access to hundreds of thousands of examples of defect images. These examples have allowed us to effectively train and instruct the system to perform accurate defect identification.

Conclusion

The automatic inspection system promptly alerts the customer about the presence and type of defects in the product. The implementation of this device has resulted in two significant effects. The first effect is immediate, as it allows for the diversion of material with critical defects and facilitates a more objective management of customer complaints. The second effect, although slower, is equally important. It involves the optimization of processes upstream of rolling (such as steelmaking and continuous casting). By combining defect information with material tracking, it becomes possible to analyze the process conditions that caused the defects in the initial production steps.