Development of a Structural Health Monitoring Framework for Major Bridge Infrastructure in Nigeria Using Sensor Fusion and Machine Learning

Nigeria's major bridge infrastructure, including the Niger, Benue, and Third Mainland bridges, lacks systematic structural health monitoring systems capable of detecting progressive damage, overloading events, and foundation scour in real time, creating safety management gaps. This study develops an original structural health monitoring (SHM) framework for major Nigerian bridge infrastructure integrating multi-sensor fusion and machine learning-based damage detection algorithms. The framework architecture integrates accelerometers, strain gauges, corrosion potential sensors, and hydraulic scour monitors in a wireless sensor network, with data processed through a cloud-based analytics platform. Original machine learning models (convolutional neural network and recurrent neural network architectures) are developed and trained on numerically simulated structural response data incorporating 16 damage scenarios of varying severity and location. Transfer learning is applied to adapt simulation-trained models to measured data from two instrumented bridges. Finite element model updating is used to maintain accuracy as structural properties evolve. The framework is implemented and evaluated on a pilot installation on the 2nd Niger Bridge during its commissioning phase. Results indicate that the CNN-RNN combined architecture achieves damage detection sensitivity of 94 percent for severity class 2 damage (stiffness reduction above 10 percent) with a false positive rate of 6 percent. Scour detection from hydraulic sensor arrays achieves 89 percent accuracy. The framework constitutes an original contribution to SHM methodology for bridge infrastructure in sub-Saharan African operating environments.

Keywords: structural health monitoring, bridge infrastructure, machine learning, sensor fusion, damage detection