A machine learning approach to damage detection is presented for a bridge structural health monitoring system, validated on the renowned Z-24 bridge benchmark dataset where a sensor instrumented, threespan bridge was realistically damaged in stages. A Bayesian autoencoder neural network is trained to reconstruct raw sensor data sequences, with uncertainty bounds in prediction. The reconstruction error is then compared with a healthy-state error distribution and the sequence determined to come from a healthy state or not. Several realistic damage stages were successfully detected, making this a viable approach in a data-based monitoring system of an operational bridge. This is a fully operational, machine learning based bridge damage detection system, that is learned directly from raw sensor data. / En maskininlärningsmetod för strukturell skadedetektering av broar presenteras. Metoden valideras på det kända referensdataset Z-24, där en sensor-instrumenterad trespannsbro stegvist skadats. Ett Bayesianskt neuralt nätverk med autoenkoders tränas till att rekonstruera råa sensordatasekvenser, med osäkerhetsgränser i förutsägningen. Rekonstrueringsavvikelsen jämförs med avvikelsesfördelningen i oskadat tillstånd och sekvensen bedöms att komma från ett skadad eller icke skadat tillstånd. Flera realistiska stegvisa skadetillstånd upptäcktes, vilket gör metoden användbar i ett databaserat skadedetektionssystem för en bro i full storlek. Detta är ett lovande steg mot ett helt operativt databaserat skadedetektionssystem.

[Truncated abstract] As a consequence of the increase in terrorist incidents, many comprehensive researches, both experimental and numerical modelling of structure and blast interaction, have been conducted to examine the behaviour of civilian structures under dynamic explosion and its impact. Nevertheless most of the works in literature are limited to response of simple structures such as masonry walls, reinforced concrete beams, columns and slabs. Although these studies can provide researchers and structural engineers a good fundamental knowledge regarding blast load effect, it is more likely for blast load to act upon entire structures in actual explosion events. The interaction between blast load and structures, as well as the interaction among structural members may well affect the structural response and damage. Therefore it is necessary to analyse more realistic reinforced concrete structures in order to gain an extensive knowledge on the possible structural response under blast load effect. Among all the civilian structures, bridges are considered to be the most vulnerable to terrorist threat and hence detailed investigation in the dynamic response of these structures is essential. This thesis focuses on the study of the response of a modern cable-stayed bridge under blast loadings. ... Firstly, analysis is conducted to examine the failure of four main components namely pier, tower, concrete back span and steel composite main span under close proximity dynamic impact of a 1000 kg TNT equivalent blast load. Secondly, based on such results, the remainder of the bridge structure is then tested by utilizing the loading condition specified in the US Department of Defence (DoD) guideline with the aim to investigate the possibility of bridge collapse after the damage of these components. It is found that failure of the vertical load bearing elements (i.e. pier and tower) will lead to catastrophic collapse of the bridge. Assuming that terrorist threat cannot be avoided, hence protective measures must be implemented into the bridge structure to reduce the damage induced by explosive blast impact and to prevent bridge from collapse. As such, a safe standoff distance is determined for both the pier and tower under the blast impact of 10000 kg TNT equivalent. This information would allow the bridge designer to identify the critical location for placing blast barriers for protection purpose. For the case of bridge deck explosion, carbon fibre reinforced polymer (CFRP) is employed to examine in respect of its effectiveness in strengthening the concrete structure against blast load. In this research, appropriate contact is employed for the numerical model to account for the epoxy resin layer between the CFRP and concrete. In addition, to ensure that the CFRP can perform to its full capacity, anchors are also considered in the numerical study to minimize the chance of debonding due to the weakening of the epoxy. The results reveal that although severe damage can still be seen for locations in close proximity to the explosive charge, the use of CFRP did reduce the dynamic response of the bridge deck as compared to the unprotected case scenario. Further investigation is also carried out to examine the change in damaged zone and global response through variation in CFRP thickness.