The research project aims to develop innovative tools for the damage assessment of critical structural systems such as buildings and bridges exposed to external shocks (e.g. earthquakes, floods) and/or degradation (e.g. corrosion, fatigue), exploiting the information from non-destructive assessments and sensors.
An essential requirement for civil and mechanical structures is to ensure a high level of operational safety, in order to protect human life and avoid financial losses. The performance and reliability of structures inevitably deteriorates with time as a result of the effects of various shocks and the deterioration induced by the environment, unless retrofit and maintenance measures are implemented.
The accurate evaluation of the damage sustained by critical infrastructural assets due to hazardous events such as floods and earthquakes is of paramount importance for optimal emergency management decisions and to ensure their safe use under the threat of future shocks. It is also essential for a correct allocation and prioritization of resources to minimise further casualties and business interruptions and to speed up recovery from disruption. Conventional practices for damage assessment of structures are based on visual inspections, which are labor-intensive, time-consuming, and potentially inaccurate. Sensor data can provide very useful information on the size of the event and on how structural components behaved. However, this information is often limited and needs to be combined with hazard and vulnerability models in order to evaluate the actual damage sustained by the structure.
Structural degradation is also a problem of concern to asset managers, which is likely to be exacerbated by the effects of climate change. The assessment of structures exposed to deterioration also relies on visual inspections and non-invasive and non-destructive tests that do not provide direct estimates of the structural state and are affected by many uncertainties. These need to be taken into account in estimating the reliability of the structure.
This project aims to develop a digital-twin based platform for damage diagnosis and prognosis of structural assets that incorporates physical modelling and data from sensors, non-destructive assessments and visual inspections. The platform will exploit the recent advancement in machine learning algorithms for supervised learning (e.g., Artificial Neural Networks) and in Dynamic Bayesian Networks. The machine learning algorithms will be used to generate surrogate models describing the structural response to the stressors leading to deterioration or damage, whereas the Dynamic Bayesian Networks will enable the integration of modelling and heterogeneous sensing data for achieving accurate estimates of the structural state.
The output of the proposed platform, enabling more accurate evaluations of the current and future state of structures, will help asset managers to make better-informed decisions concerning asset management, by overcoming the limitation of current structural assessment approaches. It can ultimately contribute to enhanced resilience of structures exposed to natural hazards and deterioration.