Machine learning application in lifetime assessment of RC structures

Ageing infrastructures have been a subject of growing concern in the past few years, and structural assessment has become a considerable challenge in the structural engineering field. Infrastructure, especially bridge structure, is a complicated structural system essential to the transportation network. Structural service life is influenced by uncontrollable factors, i.e., the growth of traffic and heavy vehicles shortening fatigue life, changing climate exacerbating structural deterioration, and a lack of proper maintenance due to costly, risky and laborious processes. The chloride ingress and crack growth analysis model adopted a simplified solution of Fick’s second law of diffusion and Paris’s law of crack propagation. Given the complexity of the nonlinear relationship variables, the ML approach has potential since artificial neural networks can learn data patterns and yield a prediction result. This study will demonstrate the integration of ML approach in reliability analysis and utilise the time-variant fatigue-corrosion of RC structures analysis results to predict the reliability of RC structures. The FNN model is shown to perform well in predicting the variable target, i.e., time to corrosion initiation, time to ultimate crack under the given set of material properties and environmental factors.