PREDICTING QUALITY AND PERFORMANCE OF OIL-FIELD CEMENTS WITH ARTIFICIAL NEURAL NETWORKS AND FTIR SPECTROSCOPY

Portland cement is used almost exclusively for primary and secondary c ementing despite the fact that its performance is variable and not und erstood completely. Cement variability makes slurry performance testin g difficult and is often a major factor in operational failure. This p aper shows that the Fourier transform infrared (FTIR) spectrum of ceme nts yield information on the factors that control hydration and cement variability, which establishes the infrared spectrum as a ''signature '' of cement composition and performance. Statistical models, based on linear statistics and artificial neural networks, that allow predicti on of cement composition, extent of aging, particle-size distribution, and limited slurry performance from cement spectra have been construc ted. In addition, the technique can be used to detect and to quantify the presence of contaminants, such as barite, clay or silica. Case stu dies demonstrate that spectra can be used to predict the nature and co ndition of the cement in a way not given by API measurements.