Appraisal and performance prediction of Senai-Desaru expressway cable-stayed bridge subject to earthquakes

• Main Author: Aizon, Nabila Huda
• Format: Thesis
• Language: English
• Published: 2021
• Subjects: TA Engineering (General). Civil engineering (General)
• Online Access: http://eprints.utm.my/id/eprint/101806/1/NabilaHudaAizonPSKA2021.pdf; http://eprints.utm.my/id/eprint/101806/; http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:146125
• Institution: Universiti Teknologi Malaysia

Application of cable-stayed bridge types in Malaysia is getting higher demands by the developer due to its cost-effectiveness. However, its high flexibility and low damping behaviour may exhibit a critical response due to earthquake loads. In addition, there are challenges for the bridge authorities to decide the severity of bridge condition in post-earthquake events. Therefore, the objectives of this research are, (i) to assess the seismic performance of a cable-stayed bridge under the earthquake time history loading in terms of dynamic behaviour and acceleration response, (ii) to establish fragility curves as guidelines of potential damage level of the cable-stayed bridge components under various earthquake loadings, and (iii) to investigate the applicability of Artificial Neural Network (ANN) as a prediction tool of damage level for cable-stayed bridge’s intelligent decision -making tool in Bridge Health Monitoring System (BHMS). The seismic performances of cable-stayed bridge were obtained by developing the 3D Finite Element Model (FEM). Two types of seismic analyses were implemented in this research, Free Vibration Analysis (FVA) and Non-Linear Time History Analysis (NTHA). Earthquake loads scaled to Peak Ground Acceleration (PGA) with low, moderate, and strong earthquake loads were utilised. The obtained results from the NTHA was then fed to the Feedforward Artificial Neural Network model to obtain a damage level prediction model for the cable-stayed bridge. The acceleration data of structure response were used as input in the ANN model. Meanwhile, the output of training network used the damage levels from the analysis which were Immediate Occupancy (IO), Life Safety (LS), and Collapse Prevention (CP). Data used for the ANN training was 70% of total data, while data used for testing and validation were 15% of the whole data, respectively. The results showed that the proposed artificial intelligent prediction model could provide prediction of up to 83.54% rate of accuracy and the least mean squared error of 0.1549. Next, the fragility curves of each component subjected to earthquake time history parameter were determined. Scaling factors for 14 time histories were calculated to generate IDA curves, which were needed to develop fragility curves. The pushover analysis and NTHA also needed to build the IDA curve. From the fragility curves, it was concluded that different bridge components have different probabilities of damage level occurrence due to specified earthquake time history parameters. This research was verified using on-site modal testing to compare the modal parameter with the FEM. The studied dynamic behaviour of cable-stayed bridges, as well as damage assessment through fragility curves and ANN approach, will greatly assist authorities in maintaining the structural integrity of their bridges by detecting and predicting the likelihood of damage under earthquake loads.