HYDROSAVER22 Jan - 20 Feb
The Newcrest Crowd
This competition has finished.
Come rock with us, join the community
program you join.
resources industry. All are welcome.
Reduce water usage in gold processing through tailings density prediction
Can you build an algorithm that predicts tailings underflow density (and therefore process water content) three hours ahead of time?
In metals processing, large tanks (thickeners) are used to separate the mineral solids from the process water. Thickeners separate the mineral solids and water by discharging a stream of the settled mineral solids in the form of a sludge (underflow) and a stream of clarified water from the top (overflow). The separation of water from tailings (mineral processing waste product) is a critical process as it enables the recycled water to be reutilised within the processing plant. This reduces the operating cost for ongoing processing of the ore and improves the environmental sustainability of the operation.
Sustaining a high tailings underflow density is a challenge Newcrest routinely endures. By only achieving a low tailings density (<54%), this results in several unfavourable consequences. A low tailings density reduces the recoverable water content that can be reused within the processing plant. It also results in larger volumes of slurry being pumped to the tailings storage facility to store the necessary mass of waste product. The culmination of these issues is a higher cost of operating due to increased energy consumption and water consumption, and it also impacts the social reputation of Newcrest during periods of sustained drought within the Orange region. Increasing the annual average tailings density by 1% saves approximately 500ML of water per annum.
- Based on provided processing data, can you predict the tailings density % 3 hours from now?
- A separate prize will be awarded to best additional insights for process improvement.
Data links will be available at the bottom of this page from event launch.
The datasets cover over approximately 24 months worth of tailings underflow density, underflow discharge flowrate, underflow discharge pressure, thickener feed rate, thickener bed pressure, thickener rake torque, flocculent addition rates, slurry pH values, process water addition rates and more.
Operational constraints relating to underflow pipeline pressure and flow, thickener rake torque, bed pressure and overflow water clarity are also provided.
The data is split into 2 sets:
A training set, and a test set containing test data for public and private scoring.
The training set contains the complete data as well as target values. The aim of the competition is to predict the value of DIC88023.PV (Underflow Solids Concentration by Mass %) three hours ahead of time. In the training set, we have copied and moved the actual readings of DIC88023.PV three hours ahead into the target column.
The test set contains the data for the public and private component but the target values are kept secret. The private and public test sets are roughly equal in size. During the competition, the competitors can evaluate their scores against the public test set, while the private test set scores are only evaluated once at the end to identify a winner.