Using Statistical Models and Machine Learning Techniques to Process Big Data from the Forth Road Bridge

Inspired by the development of the IT industry and the need for a continuous monitoring method, Structural Health Monitoring (SHM) systems have been installed on critical infrastructure like bridges in recent years. However, data from major infrastructure is often untouched apart from monitoring purposes – partly because most data owners lack the computational power for analysis. The high volume of data collected from the Forth Road Bridge’s SHM system has provided a unique opportunity to develop damage detection and response
prediction models for structure integrity assessment purposes. The goal of this paper is to apply statistic models and some machine learning methods to this SHM data, aiming to develop prediction models of bridge structural responses. Principal component analysis (PCA) reduced the dimensions of the dataset to save computational power for further analysis. Autoregressive Integrated Moving Average (ARIMA) models are used to predict traffic volumes. Comparisons between two different machine learning methods, Random Forest and traditional Artificial Neural Networks for building prediction models of strain data have been made. It is discovered that the Random Forest technique has higher
accuracy in this scenario. Based on the current research progress, future work is also proposed.