18th Edition of the Journal of Innovation Rapid Advancements in Digital Transformation | Page 87

Digital Transformation in Steel Inspection
As a grade determination service , that AI inference model develops its own analysis skills , automatically , with each steel sample it analyzes . It also adapts to the analysis from veteran human inspectors , who have the final judgement decision . As the human inspector performs their normal measurement analysis , they are automatically training the AI model , with no additional burden on the inspector .
As a training service , the AI model can provide real-world image samples to junior inspectors and teach them to measure for non-metallic inclusions . Since the veteran inspectors continually train the AI model , they are also providing new training material for the junior inspectors . Thus , the burden on veteran inspectors is reduced , and junior inspectors gain experience from real-world data samples .
The value of this new service offering was immediately recognized by the customer , as it reduced the number of veteran inspectors needed for training purposes , while simultaneously increasing the skill levels of junior inspectors .

5.2 TECHNOLOGY ENABLERS The key technology enablers for the metal inspection services were as follows :

Company Perspective
Artificial Intelligence
Cloud Computing
Image Recognition
Data Management
The Steel AI Image Processing Service can analyze steel samples , automatically determine non-metallic inclusions , and estimate grades of steel grain sizes . The AI Model can also be trained by inspectors , continually improving the model .
The Steel AI Image Processing Service runs in the cloud , so there is no need to add additional hardware on-site . The service is therefore not bound by local processing power , and updates are automatically applied with no downtime . In the case of an Internet outage , the overall system can still function without the AI Service , providing fault tolerance .
The image recognition processing algorithms are applied at the imaging station ( i . e ., at the Edge ) and are compared for speed against standard steel inspection methods ( ASTM and JIS ), and for accuracy against visual inspection from an electron microscope .
All image data for the steel grading is securely stored and managed to ensure the data ’ s integrity and consistency . The data management system allows any inspectors to read the data , but not update the data .
80 November 2021