17/12/2025
As we wrap up 2025, CGA is incredibly proud to share the milestones achieved through Civil Group’s ongoing collaboration with Flinders University. This year marked a pivotal shift from theoretical modelling to validated field application, proving that Machine Learning can revolutionise how we interpret Investigative Drilling (ID) data.
Here are the key takeaways from our 2025 R&D Annual Report:
🚀 Field Validation Success
We moved beyond the lab. At our Queensland yard and a large-scale transmission infrastructure project (gathering data from ~188 structures), we tried to validate our ID vs. Conventional (SPT) drilling models. Our simulated rock drilling on 35MPa and 70MPa concrete piles exceeded our expectations, providing a benchmark dataset that confirms the reliability of our rigs in real-world conditions.
📊 AI & Machine Learning Breakthroughs
Led by our PhD candidate Fei Huang, we’ve developed a Rock UCS Prediction Model – a game-changer for real-time site assessment. We’ve also pioneered new ML-based data cleaning methods that significantly outperform traditional statistical approaches, ensuring our data is cleaner and more actionable than ever.
📚 Published Research
Our work isn't just staying on site; it's contributing to the global body of knowledge. We had three major papers accepted and published this year:
- Predicting Rock Strength from Investigative Drilling Data (Presented at AGS Adelaide)
- Data-Driven Rock Strength Classification from Investigative Drilling Using Machine Learning (Presented at ARMA US Rock Mechanics Symposium)
- A preliminary study on the correlations between drilling parameters and pile capacities (Presented at the 1st International MWD Conference)
- Machine learning approaches for automatic cleaning of investigative drilling data (Published in Geomechanics and Geoengineering Journal)
- Preliminary study on the potential of the ID method for quantitative site characterisation (Published in Australian Geomechanics Journal)
- Comprehensive evaluation of tree‑based machine learning algorithms for soil classification from investigative drilling data (Published in Machine Learning for Computational Science and Engineering Journal)
🤝 Knowledge Sharing
In July, we hosted a successful ID/MWD Workshop at Flinders University, welcoming around 60 in-person and online attendees to discuss the future of Measurement While Drilling and its application in real-world projects.
🔮 Looking to 2026
Momentum continues next year as we target commercialisation, expanding rock strength models to 0.6-200 MPa and developing scalable software prototype for automated MWD processing.
A huge thank you to the R&D team – Ben Juett, Ben Evans, Michael Turner, Fei Huang, Dr Hongyu Qin, and Dr Masoud Manafi – for your dedication.
📩 Interested in the technical details? Reach out to discuss how ML-enhanced drilling can optimise your next geotechnical campaign.
