Unsupervised and supervised pattern recognition of acoustic emission signals during early hydration of Portland cement paste

摘要

AbstractSeveral studies have been conducted to investigate early age Portland cement hydration using acoustic emission technique, with different mechanisms attributed by different authors. In the proof-of-concept research presented in this paper, acoustic emission (AE) was employed to explore relationships between recorded signals associated with elastic stress waves and potential mechanisms associated with cement hydration. Ordinary Portland cement paste samples having water/cement ratio of 0.3 and 0.5 were monitored during the first 72h of curing using broadband AE sensors. The acoustic emission signals were analyzed using unsupervised and supervised pattern recognition algorithms to address limitations of acoustic emission parameter analysis. Wavelet analysis was utilized as a complementary method, which can be considered as a map for identification of patterns in the signal set. Unsupervised methods are useful when there is no history or background data concerning the pattern of a phenomenon such as the hydration process.]]>