Ms Xiaofang Wen

Biography

Xiaofang Wen joined Digital Roads as a Research Assistant in November 2023. Her work focuses on Digital Twin (DT) process automation.

Starting her PhD in November 2024, her study focused on evaluating Digital Twins (DTs) for the Strategic Road Network (SRN). 

Research Interests

My research asks how observations of road infrastructure can become useful forecasts. I start with what can be measured reliably from inspection images, sensors, and maintenance records. I use this evidence to build a clear picture of an asset's current condition, then study how that condition changes under traffic loading, weather, and maintenance. I combine engineering mechanics with machine learning to make forecasts more physically sensible and transparent about uncertainty. The aim is not just to detect damage, but to help decide where and when inspection or maintenance is most needed.

Strategic Themes

• Observing infrastructure condition: Uses inspection images, sensor data, and maintenance records to identify what can be measured reliably about defects and asset condition. 
• Understanding the current state: Combines observations with engineering knowledge to build a picture of current condition and make uncertainty clear. 
• Forecasting change for maintenance: Studies how condition changes under loading, weather, and maintenance to support timely inspection, repair, and long-term planning. 

Research Project

National Highways iCASE PhD: From road observations to maintenance forecasts 

This PhD develops a practical digital-twin approach for road infrastructure maintenance. The work follows four linked steps: measure road condition from inspections and other available data; estimate the current state of an asset; model how its condition changes with loading and weather; and test whether predictions can help prioritise inspection or repair. Engineering mechanics are used to check whether the models remain physically sensible, while machine learning helps learn from complex and incomplete data.  

Research Opportunities

I welcome conversations and collaboration with infrastructure owners, engineers, researchers, and practitioners working across data, mechanics, AI, and computing. Useful evidence may include road-condition records, wider monitoring data/images, and laboratory observations of crack growth or material behaviour. I am also keen to exchange ideas with people with expertise in infrastructure engineering, AI, machine learning, fracture mechanics or related fields. Access to suitable computing resources for simulation, model development, and evaluation would also be valuable. More broadly, I welcome opportunities to discuss real maintenance questions and develop practical, well-tested approaches together.

Biography

Xiaofang Wen is a full-time PhD researcher in the Department of Engineering at the University of Cambridge, supervised by Professor Brian Sheil and supported through a National Highways iCASE studentship. Her research asks a practical question: how can observations of road infrastructure be turned into useful and trustworthy forecasts for maintenance? 

She begins with the evidence that is available in practice, such as inspection images, sensor data, and maintenance records. She then uses these sources to build a clear picture of an asset's current condition and studies how that condition may change under traffic loading and weather. Her work combines engineering mechanics with machine learning so that forecasts are not only accurate where possible, but also clear about their assumptions and uncertainty. The goal is to help engineers decide where inspection or maintenance is most needed, rather than simply produce a prediction. 

Xiaofang holds BEng and MSc degrees in Civil Engineering from Tongji University. Her current interests include reliable measurement of visible defects, state estimation from incomplete evidence, fatigue and fracture in infrastructure, and careful evaluation of predictions over time. She is keen to collaborate with infrastructure owners, engineers, and researchers who can bring together well-documented data, sound engineering understanding, and suitable computing resources to develop methods that are useful beyond a single demonstration. She aims to make the resulting forecasts directly useful for real decisions over time.