Contribution of PCA and ANN in the Structural Diagnosis of a Masonry Lighthouse under Temperature and Wind Actions

摘要

Continuous monitoring was carried out on the Ile Vierge lighthouse, the tallest stone lighthouse in Europe (82 m), built in 1902, where the upper part of the tower presents a crack pattern. The monitoring was implemented over 2.5 years to characterize the actions of temperature and wind and to determine their links with crack opening. The variation of each factor measured was analysed and possible correlations between the variables were sought. In addition, the temperatures measured on site were compared with the actions required by Eurocode 1. The quantity of data extracted made it necessary to develop a data analysis method combining Principal Component Analysis, Multiple Regression Analysis, and Artificial Neural Network Analysis. Thus, the long- and short-term behaviour of the instrumented cracks could be identified. The relative influences of the wind and the temperature on crack opening were given by a percentage assessment.