Abstract:
Scientific and technological advances in monitoring systems allow for the automation of the measurement, transmission and data processing, as happens in the Portuguese dams with automated monitoring systems. However, as measuring has become increasingly easy, the increase in the amount of data available is significant, as well as the number of potential errors associated with the measuring process. The assessment of stored measurements is important because they are one of the main elements used in the activities related to the interpretation and safety control of concrete dams. A detailed assessment of measurements should be performed, for example, through the comparison of the Automated Data Acquisition System (ADAS) measurements, x_ADAS, and other values that may be used as a reference, such as Manual Data Acquisition System (MDAS) measurements, x_MDAS. In most situations, it is possible to perform both ADAS and MDAS measurements. In the case of MDAS measurements, the instruments used in concrete dam monitoring follow relatively simple physical principles and there is a lot of experience acquired over the years. This allows us to consider that the MDAS measurements are of good quality, which makes them a good reference element for the analysis of ADAS measurements. The main idea of the methodology proposed is to assess if ADAS and MDAS measurements from paired samples (x_ADAS,x_MDAS) represent the same population. For each pair of measurements (x_ADAS,x_MDAS), very close values are expected. Probability density function (PDF) may be used to characterise the measurement distribution of each sample. In the case of paired ADAS and MDAS measurements, two similar probability density functions (not necessarily identical due to random effects) are expected, that is PDF(x_ADAS)≈PDF(x_MDAS). If there are differences between ADAS and MDAS samples, these differences will be reflected in each PDF, and the causes may be identified.
