Recently, I was asked where the next big improvement in mineral processing would emerge. In my opinion, it is likely to come from the rapid estimation of ore properties. And when I say ‘rapid’ I mean 30 seconds and not 12 hours (like shift composites).
Imagine a scenario when the mill feed includes BOTH bulk sensor analysis (targeting % copper or % chalcopyrite) combined with particle-by-particle imaging using something like an XRT sensor. Can anyone see the potential for pattern recognition? This is where the strengths of AI should be applied.
- Low-grade material with coarse particles showing a poor XRT pattern?
Reject the screen oversize before it reaches the grinding circuit (where power and water are consumed). - High-grade material with a consistently good XRT pattern?
Send it to stockpile when it can be processed under the best plant conditions.
However, as an industry we can’t just sit back and wait for AI processing to tell us when ore properties change. We need to be prepared to do something with this information upon identifying bad or waste material in the ore stream:
- What can be done about it? Redirect it?
- Where would it go?
This is what’s missing in our current process plant design – the ability to reject material that’s been labelled as ‘ore’. Conversely, the ability to segregate excellent material for optimal plant conditions.
In SRK’s work looking at coarse pre-concentration with sensor sorting, we’re investigating where the value is, but is waste hiding among the good ore?
Have questions? Reach out to Adrian directly, here.
