Abstract

Subway networks play a key role in the smart mobility of millions of commuters in major metropolises. The facilities of these networks constantly deteriorate, which may compromise the integrity and durability of concrete structures. The ASCE 2017 Report Card revealed that the condition of public transit infrastructure in the U.S. is rated D-; hence a rehabilitation backlog of $90 billion is estimated to improve transit status to good conditions. Moreover, the Canadian Urban Transit Association (CUTA) reported 56.6 billion CAD in infrastructure needs for the period 2014-2018. The inspection and assessment of metro structures are predominantly conducted on the basis of Visual Inspection (VI) techniques, which are known to be time-consuming, costly, and qualitative in nature. The ultimate goal of this research is to develop an integrated condition assessment model for subway networks based on image processing, Artificial Intelligence (AI), and Non-Destructive Evaluation (NDE) techniques. Multiple image processing algorithms are created to enhance the crucial clues associated with RGB images and detect surface distresses. A complementary scheme is structured to channel the resulted information to Artificial Neural Networks (ANNs) and Regression Analysis (RA) techniques. The ANN model comprises sequential processors that automatically detect and quantify moisture marks (MM) defects. The RA model predicts spalling/scaling depth and simulates the de-facto scene by developing a hybrid algorithm and interactive 3D presentation. In addition, a comparative analysis is performed to select the most appropriate NDE technique for subway inspection. This technique is applied to probe the structure and measure the subsurface defects. Also, a novel model for the detection of air voids and water voids is proposed. The Fuzzy Inference System (FIS), Adaptive Neuro-Fuzzy Inference System (ANFIS), and Monte Carlo Simulation (MCS) are streamlined through successive operations to create the integrated condition assessment model. To exemplify and validate the proposed methodology, a myriad of images and profiles are collected from Montréal Metro systems. The results ascertain the efficacy of the developed detection algorithms. The attained recall, precision, and accuracy for MM detection algorithm are 93.2%, 96.1%, and 91.5% respectively. Whereas for spalling detection algorithm, are 91.7%, 94.8%, and 89.3% respectively. The mean and standard deviation of error percentage in MM region extraction are 12.2% and 7.9% respectively. While for spalling region extraction, they account for 11% and 7.1% respectively. Subsequent to selecting the Ground Penetrating Radar (GPR) for subway inspection, attenuation maps are generated by both the amplitude analysis and image-based analysis. Thus, the deteriorated zones and corrosiveness indices for subway elements are automatically computed. The ANN and RA models are validated versus statistical tests and key performance metrics that indicated the average validity of 96% and 93% respectively. The air/water voids model is validated through coring samples, camera images, infrared thermography and 3D laser scanning techniques. The validation outcomes reflected a strong correlation between the different results. A sensitivity analysis is conducted showing the influence of the studied subway elements on the overall subway condition. The element condition index using neuro-fuzzy technique indicated different conditions in Montréal subway systems, ranging from sound concrete to very poor, represented by 74.8 and 35.1 respectively. The fuzzy consolidator extrapolated the subway condition index of 61.6, which reveals a fair condition for Montréal Metro network. This research developed an automated tool, expected to improve the quality of decision making, as it can assist transportation agencies in identifying critical deficiencies, and by focusing constrained funding on most deserving assets.