Wood has been awarded the second place prize at The Newcrest Crowd’s Hydro Saver Data Science Challenge, sponsored by Unearthed and Newcrest Mining Limited.
The Newcrest Crowd is a new and ongoing programme which aims to work with the best innovators worldwide and identify, develop and implement new solutions for real-world, multi-million dollar applications.
The Hydro Saver Data Science Challenge saw hundreds of data scientists and machine learning experts from a wide range of industries come together to compete. Each participant was tasked with building an algorithm that predicts tailings underflow density (and therefore process water content) three hours ahead of time.
Wood’s automation & control team, which included Sahil Masand, George Pickering, Danyal Rasheed and Troy Holdsworth, utilised a dense neural network combined with a recurrent neural network to predict the gold processing tailings underflow density.
The approach combined experience designing and modelling process systems with advanced machine learning techniques to give the Wood team an advantage over competitors with backgrounds solely in data science or machine learning. Examples of the process knowledge applied to problem include:
- Identification and filtering of miscalibration or instrumentation error data points;
- Use of an unsupervised learning algorithm to identify specific process states; and
- Consideration of additional process parameters to further guide the machine learning.
The team impressed judges finishing with a score of 1.48% (Root Mean Square Error), finishing just 0.06% behind the winning team from Three Springs Technology, a specialist technology and artificial intelligence consulting company.
Congratulations to the automation & control team in Perth for this fantastic result. Not only does this have direct application to our mining business but it also demonstrates our expertise in process analytics which can be applied across a whole range of industries.