Application of Statistical Machine Learning Algorithms for Classification of Bridge Deformation Data Sets

This paper presents the design and development of a structural health monitoring (SHM) system specifically tailored for transportation infrastructure components such as bridges. If focuses mainly the application of statistical machine learning (ML) algorithms to classify deformation datasets of a bridge. A model of a steel bridge was constructed and contactless sensors were placed to collect deformation data. Four loads were applied at each of the pre-defined four locations identified to represent heavy loads across the real bridge. Computer simulation in ANSYS and application of gradient boosting neural networks were performed to produce a comparative and predictive analysis of the behavior of transportation infrastructures, which can be used to understand the health of the structure and make informed decisions. Deformation levels at 100 critical locations on the bridge model were collected in each experiment by using sensors. The experiments were repeated to get average data for processing. Python programming language was used for coding and the analysis was performed in a Google Collaboratory Notebook. Development and training of the models were done using the Pycaret, which is a Python based framework that supports a variety of ML tools. Performance of each ML technique was evaluated by means of the accuracy. The final is capable of simulating multiple load conditions on structures, identifying possible failure points, and detecting and predicting failure scenarios. Both hardware and software implementations of a model of a bridge were performed as a pilot project to validate the proposed system.

Automated summary

This paper presents the design and development of a structural health monitoring system tailored for transportation infrastructure components, focusing on the application of statistical machine learning algorithms to classify deformation datasets of a bridge. The study utilized computer simulation and gradient boosting neural networks to predict the behavior of transportation infrastructures and evaluate the performance of ML techniques.