Banca de QUALIFICAÇÃO: JORDANA TEIXEIRA DA SILVA

METHODOLOGY FOR EVALUATING SOUNDLANDSCAPE: MODELS PREDICTIVES AND APPLICATION IN URBAN SPACES IN THE CITY OF MACEIÓ- AL

 soundscape; ISO 12913, predictive model; linear multiple regression; artificial neural network; seafront.

The field of study in soundscape has been gaining strength in recent years, due to the need to overcome the approach to managing urban environments based solely
in noise control policies. Through a multidisciplinary approach, the assessment of the urban sound environment takes into account the numerous aspects that involve variables
subjective, related to users' perception, physical variables, specific to the location and context. It is essential to develop analysis tools capable of understanding and predicting the sound perceptual results related to certain decision-making within the scope of urban planning, based on predictive models. This thesis aims to develop a methodology for evaluating soundscapes based on the development of predictive models and application in urban public spaces in the city of Maceió, Alagoas, Brazil, with an urban area corresponding to the seafront of the Ponta Verde and Pajuçara neighborhoods.
The work corresponds to applied research, with data collection in the field, following the ISO 12913 normative series, application of sound walks, questionnaires and measurements. The proposal includes development of a new methodology for data collection, with
proposition of a mixed questionnaire, based on Method A and B of ISO 12913-2, plus contextual and visual information. A methodology is proposed for characterizing acoustic environments, which takes into account ISO 12913-3, with increased analysis of the interrelationships between aspects, followed by the study of correlations and assessment of relevance for the selection of indicators (collected information, input data ) and descriptors (measures of how
people perceive the acoustic environment, output data) in the model. Modeling follows the development of an analysis structure, through the construction of linear (linear multiple regression) and non-linear (neural networks) predictive models. Preliminary results indicate that through the predictive models under development, it is possible to characterize the urban sound environment and identify the main elements that interfere in the configuration
of the soundscape, which can serve as a tool for implementing public policies. It is expected that the developed methodology will be easy to implement, in order to contribute to the assessment, modeling, quality and management of the urban soundscape.