| OCR Text |
Show - 4 - fusibilit. Additional relationships reflecting imbalance in the test matrix include a significant correlation between oxygen level and Fe 203 , sulfur, and pyritic sulfur, and between heat release rate and oxygen level, Ti0 2 , A1 203 , sulfur, and pyritic sulfur. Some correlation between coal elemental concentrations is fairly unavoidable, for example, CaO and S03 (r=0.753), Fe 203 and Si02 (r=-.834), and sulfur and pyritic sulfur (r=.785). Other significant correlations are not as easily explained and may again be due to imbalance in the test matrix. Suffice it to say that the correlation matrices (matrices by firing mode were also constructed) were a constant source of information during later regressive analyses where the effects of multicolinearity (strong intercorrelation) were to be avoided. SLAGGING-INDUCED HEAT FLUX REDUCTION TO A SIMULATED WATER-WALL Several techniques were used to elucidate the magnitudes of the dependencies of slagging-induced heat flux reductions on operating conditions and coal characteristics. In addition to the correlation matrices (which identified bias in the data and some significant linear explanatory variable dependencies) four other methods were employed: • Scatter Plots and Contour Maps • Analyses of Variance • Analyses of Covariance • Stepwise Regression Techniques The remainder of this section covers a discussion of the more pertinent results from these investigations as well as a comparison of the developed correlations with traditional slagging predictors. Scatter Plots Severa 1 data plots revea 1 ed a dependence of hea t fl ux degrada t i on on some of the independent variables. These relationships were statistically substantiated using 95% confidence limits. Figure 2 is a plot of two-hour heat flux reduction versus six-hour heat flux reduction. The wall-fired test data, which span the range of two-hour reductions from approximately 0-40%, do show some nonlinearity. Generally, over all modes the six-hour heat flux reductions were fairly linearly related to the two-hour values. On average 50% to 80% of the total heat flux reduction occurred in the first two hours. The implication of these observations is that the dynamics of the heat flux reductions show some similarities despite firing mode (flow pattern) differences. Figures 3 and 4 are plots of initial (or zero-hour) absolute heat flux versus t'ft'o-hour heat flux reduction and six-hour heat flux reduction, respectively. Also shown for comparative purposes are the regression lines for each firing mode. Although a statistically significant positive correlation was displayed for each firing configuration, the magnitude of the dependence was strongest for the wall-fired data. |
