Advances in modern mass-manufacturing using NDE 4.0 process based on artificial intelligence
R Gr Maev
Industries can save tons of time and money by making non-destructive inspection less error-prone. For that reason, research and development (R&D) in this direction began to be one of the top priorities and recent developments in this field demonstrate really promising results.
In this presentation the author will discuss a few of the most promising concepts that are currently under strong development by a few international research groups. Some successful results in this field will be presented, one of them being real-time automated spot weld quality analysis from ultrasonic B-scans using deep learning. This is the first work in this direction where a deep learning-based framework has been applied for in-line detection of objects of weld shape from ultrasonic B-scans. Applying a deep learning interpret detected components to numeric features and classify welds as good, acceptable or bad in real-time during production. This solution allows NDE specialists to improve existing automated system to produce high weld quality classification accuracy that matches with production-level satisfaction.
Another important another aspect of this problem is demonstrating how the results obtained from the deep learning model can be interpreted to understand the physical phenomena taking place inside materials during welding. It is crucial for manufacturing cycle time that this understanding needs to be achieved in real time. High-speed original algorithms allow immediate decisions to be made and in-line requests to be sent to the welder to modify welding parameters to be sure that at the end of each spot welding cycle the quality of each weld will be good, which is a great example of an advanced NDE 4.0 solution.
The author is confident that when the whole complex of various aspects of this problem are resolved and persistent results are demonstrated, it will allow for a completely revolutionised mass-manufacturing process using the artificial intelligence NDE 4.0 process at car assembly plants globally. This technology brings big savings in production cycle time, cuts on labour costs and eliminates unnecessary destructive tests, which are today still a part of quality inspection process.
In this presentation the author will discuss a few of the most promising concepts that are currently under strong development by a few international research groups. Some successful results in this field will be presented, one of them being real-time automated spot weld quality analysis from ultrasonic B-scans using deep learning. This is the first work in this direction where a deep learning-based framework has been applied for in-line detection of objects of weld shape from ultrasonic B-scans. Applying a deep learning interpret detected components to numeric features and classify welds as good, acceptable or bad in real-time during production. This solution allows NDE specialists to improve existing automated system to produce high weld quality classification accuracy that matches with production-level satisfaction.
Another important another aspect of this problem is demonstrating how the results obtained from the deep learning model can be interpreted to understand the physical phenomena taking place inside materials during welding. It is crucial for manufacturing cycle time that this understanding needs to be achieved in real time. High-speed original algorithms allow immediate decisions to be made and in-line requests to be sent to the welder to modify welding parameters to be sure that at the end of each spot welding cycle the quality of each weld will be good, which is a great example of an advanced NDE 4.0 solution.
The author is confident that when the whole complex of various aspects of this problem are resolved and persistent results are demonstrated, it will allow for a completely revolutionised mass-manufacturing process using the artificial intelligence NDE 4.0 process at car assembly plants globally. This technology brings big savings in production cycle time, cuts on labour costs and eliminates unnecessary destructive tests, which are today still a part of quality inspection process.
