Artificial Intelligence (AI) is overrated. Yes, we could teach a Google AI engine to avoid defects due to blowholes, low mold temperature, or low pouring pressure, but it would take massive numbers of sensors and months if not years of data to analyze. Even then, this AI would need to start making mistakes and continue the second part of its learning. It is going to be a very long time before general-purpose AI is a credible tool in die-cast production.
However, a human can only digest and act upon so much data before even the best become overloaded. I believe that we are at a unique moment in time for production work. Humans are asked more and more to behave like machines and management looks for operational technology to have the insight of experienced people.
I have seen several attempts at deploying Artificial Intelligence in die-cast production. Most have been expensive wastes of time. If you want to succeed with AI deployment in die-casting, here are a few rules we follow:
Learn the process. An amazon smart speaker is already trained to understand a spoken language, so all it needs to learn is how you speak.
Focus on defect prediction and getting this information to a skilled operator. Leaving the man-in-the-middle makes the problem simpler and reduces risk of catastrophic failure.
Never use AI where classical controls work. This goes back to my first point of learning the process first.
Ensure there is a quick win as a proof of concept. Without it, internal support falls away even if the potential is enormous.
In a hyper-competitive manufacturing world, you need to streamline your production lines everywhere you can. You can gain real-time visibility into your factory floor processes and operations by collecting and analyzing the data produced throughout the die-casting process.