Machine learning techniques for structural health monitoring

Structural Health Monitoring has become a hot topic in recent decades as it provides engineers with sufficient information regarding the damages on civil infrastructure by analysing data obtained from the monitoring sensors installed in the structures. Commonly, the process of implementing a damage identification strategy for aerospace, civil and mechanical engineering infrastructure is referred to as Structural Health Monitoring (SHM). The development of smart sensors and real-time communication technologies via Wireless Sensor Networks (WSN) has empowered the advancement in SHM. Recently, statistical time series models have been widely used for structural damage detection due to the sensitivity of the model coefficients and residual errors to the damages in the structure. Increasingly Machine Learning (ML) algorithms are employed for damage detection tasks. This research sheds light on the methodologies to predict the structural damage on concrete structures with the help of sensor technology by effectively combining data science and ML strategies. Experimental test results publicly available are used, where the tests have been performed with varying stiffness and mass conditions with the assumption that these sources of variability are representative of changing operational and environmental conditions in addition to changes caused by damage. To enhance the accuracy of damage detection, instead of the traditional time series analysis, ML is used for learning from prior experience. To detect the existence and location of the damage in the structure, we use supervised learning, and for measuring the severity of the damage, unsupervised learning is used. Accuracy results are obtained with three well-known ML algorithms (KNN- k Nearest Neighbour, SVM-support vector machine and RFC random forest classifier). In this study, the Random Forest Classifier algorithm generated good predictions on damaged and undamaged conditions with good accuracy, when compared to the KNN algorithm and Support Vector Machine algorithm under the supervised mode of machine learning. The utilisation of sensor technology effectively combined with aspects of Artificial Intelligence (AI) such as Machine Learning has the potential to implement a more efficient SHM system.