By: Mahyar Asadi, VP Innovation, Novarc Technologies
Published By: World Pipelines – August, Pg. 37-40
Excerpt taken from Pg. 38
Using artificial intelligence to fully automate the pipe welding process
NovEye™ is a vision-based weld monitoring and control software that can analyze the weld pool and make adjustments in real-time, just as a human operator would use their eyes and biological neural nets to move their hands. NovEye increases the autonomy of the cobot, and allows the operator to work on more sophisticated tasks. NovEye monitors and makes real-time corrections to the welding operation, based on years of data collection and machine learning.
Being able to produce an AI system that can control a weld pool with the required latency and weld quality performance that’s required is a significant task and a challenge that is not immediately realized. AI systems are no longer explicitly programmed to deliver a perfect task. Instead, these smart systems are architectured to learn and improve over time. In other words, they can start with the current state of your welding process and self-direct through the machine-learning to a better state.
For example, initially NovEye had to be implemented in a low-level code so it could run efficiently on the robot. It took a number of years to develop as it required a considerable amount of effort to examine the weld data, and create high-quality and reliable code at low levels.
Today, the technology can record certain features in the weld dimension in exactly the same way welders would be looking at the weld pool (how it’s wetting on the sides, the position of the torch, etc.) and decide the next move, whether it’s increasing or decreasing weld parameters, wire feed speed or travel speed. Neural networks and artificial intelligence allow the cobot to follow the same procedures as the human operator, assessing those features as it’s been trained to command the control system to make the necessary corrections – in real-time – during the weld.
NovEye is now operating as a sophisticated piece of software that ties into the control system of the robot, and in that way, the robots – the actuators – can control the weld puddle in the same way a human welder would.
This is revolutionary for many industries, as they need to evolve their capabilities to remain competitive. Currently, there are many market demands for smart welding systems, including:
- Adaptive seam tracking and control of misalignments in the welding preparations;
- Maintaining the stability of the welding process through noises, fluctuations, and external instablers to retain a uniform quality along the weld;
- Auto-learning from good and bad weld practice over time to avoid repetitive mistakes, constantly exploring opportunities for a better weld.
Existing Industry 4.0 platforms spread out the learning domain to incorporate information from other welding machines. These platforms enrich the intellectual level of our local system to dive deep into massive data collected from around the globe and develop actionable insights…