Bridge Health Monitoring Using MATLAB, Signal Processing and Deep Learning

A student-laptop structural health monitoring (SHM) system for BR-07, a two-span continuous steel-concrete composite girder bridge (2 x 30 m), instrumented with 8 vertical accelerometers and an air-temperature sensor that record 60 s at 100 Hz. The whole chain is built and measured: a data-quality gate, filtering and spectral analysis (Welch, spectrogram, Morlet wavelet), automated operational modal analysis (Frequency Domain Decomposition and covariance-driven SSI) and a 1-D convolutional network (DamageNet) that locates damage in one of 8 girder zones or at the pier and estimates its severity, next to an autoencoder novelty index and classical baselines. On 622 held-out records DamageNet reaches a location accuracy of 0.846 and a detection AUC of 0.966 - but a random forest on engineered OMA features locates better (0.873), and the fusion of both gives the best detection AUC (0.980). In a simulated monitoring year with a frost spell, a sensor fault and a growing crack, the temperature-normalised methods raise no false alarm, the CNN detects the crack at 9.7 % stiffness loss after 52 days and the Mahalanobis index at 13.2 % after 71 days, while a naive frequency band gives 7 false-alarm days in the cold spell. Honest scope: all data are simulated by the project's own finite-element model (no measured bridge data), and MATLAB is NOT installed on the build machine - the 10 MATLAB scripts were written but never run; every number comes from the Python implementation.

Category: Civil Engineering / Structural Health Monitoring / Bridge Engineering / Structural Dynamics / Signal Processing / Operational Modal Analysis / Deep Learning / MATLAB (scripts) with Python

Civil Engineering Project C03 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.