Stop Collecting Data. Start Finding the Loss.
How plant teams can turn existing operating data into better tonnes, recovery and decisions.
By Krutik Parikh, Director & Metallurgist, Liberate Metallurgy
Every processing plant has days when the numbers do not quite make sense. Throughput is down, recovery has drifted, reagent consumption is up, or the grinding circuit is making tonnes but not the grind the downstream plant needs.
The usual response is to open more trends, compare shift reports and start looking for the variable that changed. The problem is that most plants do not suffer from a lack of data. They suffer from too many disconnected signals.
Start with the operating question
A useful plant review should begin with a question that matters at the morning meeting:
- Where are we losing tonnes or recovery?
- What is driving the loss?
- What is it worth?
- What can the plant safely change today or this week?
That sounds simple, but it changes the way data is used. Instead of building another dashboard, operating data is connected to the process mechanism and then to business value.
Grinding is a good example
A mill can deliver high throughput while producing a coarser P80 that gives away recovery downstream. Alternatively, the circuit can protect grind but sacrifice tonnes, power efficiency or media life.
Looking at feed rate, mill power, density, cyclone pressure, circulating load, P80, media consumption and downstream recovery together can reveal the operating window that has historically produced the best overall result.
For the operator, that means clearer operating boundaries. For the Metallurgy Manager, it creates evidence for a plant trial. For the General Manager, it translates the discussion into tonnes, ounces and cost.
The same applies to CIL
Overall recovery is important, but it does not explain where the gold is being lost. Grind, dissolved oxygen, cyanide, residence time, carbon activity, carbon loading and tails grade all interact.
A plant averaging 94% recovery may be performing well — or it may have a repeatable 0.3–0.5 percentage-point opportunity hidden inside normal variability. The important question is whether that additional recovery can be achieved economically and sustained.
Process Intelligence should finish with an action
This is how Liberate Metallurgy approaches Process Intelligence through GrindScope™ and CILScope™: use existing plant information to find the loss, understand the likely cause, estimate the value and identify a practical test.
AI and advanced analytics can help identify relationships across large datasets, but they should remain behind the operating decision. The plant team needs an explanation it can understand and an action it can verify.
The best output is not a sophisticated model. It is something much simpler:
Problem ? Evidence ? Value ? Action ? Result.
If the analysis cannot help the plant make a better decision, it is just more data.