Multi-sensor Data Fusion System for Enhanced Analysis of Deterioration in Concrete Structures

Multisensor data fusion provides significant advantages over single source data. The use of multiple types of sensors plays an important role in achieving reasonable accuracy and precision. A novel integrated heterogeneous multi sensor data fusion approach in structural health monitoring is proposed. The study concerns to find a simple and affordable monitoring strategy for Alkali-aggregate reaction (AAR), which is one of the root causes for structural deterioration in concrete. Many researchers have stated the process of gradual structural deterioration to be very complicated because of the random distribution of aggregate, the arising and growth of concrete crack shows remarkable abnormity and discontinuity. Although researchers have developed several test methods to identify potential reactivity of aggregate. There is no universally valid standard testing method for all cases of AAR. The conventional methods namely petrographic examinations, expansion tests, and chemical analyses are cost expensive and require high skilled person to carry out tests. More over they qualitatively determine the possibility of AAR presence but least quantitatively predict whether the AAR will be deleterious. In order to develop a monitoring strategy for AAR in small size concrete structures, it is necessary to simulate the AAR expansion and cracking within reasonable laboratory timescale. A standard method to accelerate AAR expansion is employed on four samples which are prepared with different level of alkali concentrations. Different sensor systems are used at surface and internal level. Acoustic sensor system, electro-mechanical system, optical systems are employed to obtain surface level damage. The internal level of damage is obtained by embedded sensors within the structure. Features extracted from heterogeneous sensors are fed to Decentralized Kalman filter. The fused global estimates and individual source estimates are fed to artificial neural network (ANN), which characterize and quantify the level of damage. The research is focused on establishing correlation among surface damage level, internal damage level and the amount of gel concentration in the structure. To emphasize the expected improved accuracy using data fusion, evaluations are done on efficiency and accuracy of single source data system comparing with the fused heterogeneous data.