Milad Mousavi, a RIIS Hub student and researcher from the School of Civil and Environmental Engineering at UNSW Sydney, was honored with the Best Presentation Award at the 10th International Conference on Machine Learning Technologies (ICMLT 2025), held in Helsinki, Finland from May 23–25, 2025. His paper, “Online Deep Transfer Learning and Multi-Sensor Analysis for Enhanced Underground Monitoring,” introduces a novel approach to improving safety and efficiency in underground environments.

Tackling the Challenges of Underground Monitoring

Underground operations, such as coal mining, face significant challenges due to complex spatial layouts, unpredictable geological conditions, and hazardous gas emissions—particularly methane. Traditional monitoring systems often struggle with data scarcity and adaptability, especially during the early stages of operation.
Milad’s research addresses these limitations by integrating multi-sensor analysis with online transfer learning, enabling real-time prediction and monitoring of methane concentrations. His model leverages a Long Short-Term Memory (LSTM) neural network trained on data from a Polish coal mine and fine-tuned using online learning with data from a Chinese coal mine.

Key Innovations and Results

The study demonstrates that combining historical data with real-time updates significantly enhances prediction accuracy. Notably:

– The online transfer learning model achieved an R² score of 0.93, indicating high reliability in methane concentration predictions.
– The model trained with transfer learning showed faster performance improvements compared to models trained from scratch.
– This approach helps mitigate the “cold start” problem in new underground projects, where data is initially limited.
By continuously adapting to new data, the model supports timely interventions, reduces false alarms, and improves operational efficiency—ultimately contributing to safer underground environments.

A Vision for the Future

Milad’s work, as a RIIS Hub student supported by the ARC Research Hub for Resilient and Intelligent Infrastructure Systems (RIIS), sets a new benchmark for intelligent underground monitoring. His presentation at ICMLT 2025 not only showcased technical excellence but also highlighted the practical impact of machine learning in infrastructure resilience and safety.
As underground operations become increasingly digitized, innovations like Milad’s will play a crucial role in shaping the future of smart, adaptive monitoring systems.