Matthew Perras, assistant professor, Department of Civil Engineering at the Lassonde School of Engineering is the principal investigator on the project “Using machine learning to understand ancient climatic influences on the stability of cliffs and tombs in the Theban Necropolis of Egypt.” Working with an international research team that includes his colleague Usman Khan, also an assistant professor in the Department of Civil Engineering at Lassonde, who is a co-principal investigator on the project, the research focuses on the Theban Necropolis, a UNESCO World Heritage site comprised of tombs and temples near Luxor, Egypt. The project received $250,000 in funding.
The tombs in the Theban Necropolis are often shallow excavations with entrances at the base of cliffs. The tombs hold evidence of rock mass collapses during construction through to recent deterioration leading to potential instabilities. Climatic variations are known to cause rock to deteriorate, however, there is debate about the exact influence on crack growth rates. Due to lack of detailed observations and experiments on long-term crack growth in rock, since such experiments span many months or even years, current numerical tools are not capable of capturing the influences of climate change on crack growth. This leads to challenges in determining when instabilities will develop and problems designing preservation strategies. To address these challenges, Perras and the research team propose to utilize machine learning (ML) to aid in analyzing existing climate data and crack growth indicators to predict instability. A ML algorithm will be trained on current measurements (weather & crack movement), then on historic climate & photographs of crack growth.
Ancient climate records and models (Nile sedimentation, tomb flooding & collapses) could be used to back analyze the influence on crack growth with time. With the expertise of geotechnical engineering, geology, archaeology, data and climate science, the researchers will seek to understand the prevailing conditions that led to the current state of stability and develop guidelines for preserving the stability into the future. The novelty of this research is in the combination of machine learning with archaeology and geological engineering. Machine learning in both fields is in its infancy, however, such techniques allow for nuanced behaviors to be extracted from large and complex data sets as in this project. Understanding the current measurements, past influences and applying it to predict future instabilities will help to identify key areas for protection and aid in preserving this UNESCO site for generations to come.
Link to the full article on YFile: