A Theoretical Structural Impairment Detection System for Timber Railway Bridges

The objective of this research is to develop a theoretical Structural Impairment Detection System (SIDS) for timber railway bridges. Due to fatigue, the timber stringers in timber railway bridges develop shear cracks. These shear cracks lead to higher bridge deflections, higher stresses in the stringers and rail, and shorter fatigue life of the system. A SIDS is proposed which links wheel path accelerations obtained from traversing freight cars to the condition of the bridge.

In order to develop the SIDS, two models of timber railway bridges with various levels of structural impairment were developed. The first model was a quasi-static model developed from classical beam theory and implemented in MATLAB. The second model was a dynamic, finite element model created in LS-DYNA. Traversing axle loads were imposed on the models. The results obtained from the model were the wheel paths the axles take as they traverse the bridge. The paths were expressed as vertical displacements as a function of position on the bridge. Wheel path accelerations were obtained by numerically differentiating the vertical displacements.

The accelerations were then used to train neural networks to have an input of an acceleration vector and an output of a bridge condition vector. The neural networks were trained on results from both models under three train speeds: 40 mph, 30 mph, and 20 mph. The networks were able to determine the correct bridge condition 90% of the time when the train speed was 40 mph and 70% of the time when the train speed was 30 mph. The networks were not successful in determining bridge condition when the train speed was 20 mph.