Estatística espacial aplicada às falhas e confiabilidade de sistemas de abastecimento de água

Abstract

Since the end of the 1980s, such characteristics, problems and challenges in the operation have been evaluated in an approach of analysis of failures and risks of the components or systems as a whole. The reliability of these systems has been approached, for the most part, in a probabilistic or stochastic way. Through the use of probability distributions, the factors that stress the system, as well as those that define its capacity, are considered as random variables, or, the system is approached as a black box in the failure time analysis. In this work we demonstrate the applicability and efficiency of using spatial statistics techniques of dot pattern analysis, clustering spatial correlation. The geospatial analysis tools applied to faults in SAA are recent and this work is unique in its application to a large data set, integrating the temporal and spatial approach, through descriptive and inferential statistics. The use of machine learning techniques (K-means and DBSCAN) to identify clusters made it possible to define the boundaries of suitable areas for network maintenance services, and could also help define measurement and control districts. In addition, such techniques helped in choosing an optimal influence radius, serving as a more technically robust definition for the definition of the input parameter in the kernel density analyses. For future work, I recommend crossing hydraulic and water quality information with information from service orders for maintenance and repair, in such a way that it allows the assessment of cause and effect between variables such as pressure and supply failures