Principal component analysis to compress acquired data offshore

Abstract

Telecommunications offshore have connectivity in virtually all parts of the globe via satellite, with increasing bandwidth and lower cost, but still far from levels that are onshore. The principal component analysis (PCA) is a statistical technique that has found application in fields such as biometrics or compression of images, being a common tool for finding patterns in multidimensional data sets. The hypothesis for this work was that it was possible to use the theory of PCA to compress, with sufficient accuracy, the large amount of data that are collected on board to a vessel and then sent by satellite in a more economical or rapid way than the traditional one. The material used were 44 samples of 182 different signals, collected from 19 different equipment on board to “Castillo de Villalba” Liquid Natural Gas carrier vessel. With these data, the PCA algorithm was applied using a computer program developed by the authors, generating new data packets to send by satellite. Different strategies were used in order to ensure that the coefficient of correlation r between original and reconstructed data onshore were equal or greater than 0.95. The results showed that it was possible to save 46.9% in the number of data sent via satellite, in the case of grouping all the 182 signs, with a mean r = 0.95 ± 0.08. This strategy is appropriate for onshore vessel equipment telediagnostic and maintenance decision making, with telecommunication cost or time savings.