Data fusion of multiple machine intelligent systems for the condition assessment of subway structures

Abstract

Water intrusion through soil is considered the most significant structural issue and the major cause of concrete degradation in subway networks. An enormous amount of water infiltration may expedite the deterioration mechanisms, such as moisture marks, spalling, scaling, and cracks. These mechanisms can compromise the structural durability and jeopardize people’s safety. The condition assessment of concrete infrastructure is predominantly conducted based of visual inspection techniques, which are costly, time-consuming, and error prone. In this research, two main models for the condition assessment of subway networks are proposed. First, image processing techniques and machine intelligent systems are streamlined through successive operations to detect and quantify multiple surface defects automatically. Spatial and frequency domain filters are used to enhance the image clues, in tandem with artificial neural networks (ANNs) and regression analysis (RA) for defect recognition. The Monte Carlo simulation (MCS) is then leveraged to deliver advanced optimization and accurate estimation for each defect’s condition index in the subway element. The developed method was implemented on four stations in Montréal subway systems, whereby the performance of ANNs and RA was validated through R2 as 0.928 and 0.957, respectively. Moreover, the MCS forecast precision was recorded as 95% percentile, which proves the efficacy of the developed models. This research provides insights for infrastructure managers about maintenance and intervention plans in order to prioritize their spending policies.