OCR Text |
Show - 5 - Some strong two parameter dependencies were suggested by the correlation matrices. Figure 5 is a surface map and contour plot of six-hour heat flux reduction as a function of initial heat flux and flue gas oxygen concentration. The irregularly spaced data were mapped using a nodal weighted averaging technique with equi-spaced search areas. Strong dependencies are self evident above oxygen levels of approximately 3.0% with increasing oxygen concentration and decreasing initial heat flux values resulting in lower heat flux reductions. Analyses of Variance and Covariance Generally fuel evaluation programs were structured as 2x2xn factorial test matrices, where n specified the number of coal types. Two levels of excess air and two levels of heat release rate were typical. This type of test design is quite conducive to traditional analyses of variance and each program was checked for block effect significance in three categories (heat release rate, oxygen concentration, and coal type). Heat release rate and coal type effects were consistently found to be statistically significant for the fuel evaluation programs. Slagging-induced heat flux reductions increased with heat release rate (as suggested by the scatter plots of initial heat flux versus heat flux reduction). Least significant difference (LSD) tests revealed pairwise coal differences with the most consistent potential contributor being Fe 203 concentration (larger concentrations of Fe 203 produced lower heat flux reductions). While the data base was still reduced into individual fuel evaluation programs, covariate operating condition and coal characteristic impacts on heat flux reduction were evaluated with simple regressive techniques. The following terms were found to be statistically significant (in order of decreasing frequency). • Initial Heat Flux • Si02 and Fe 203 • K20 and Na 20 • Ash Loading (lb/hr) • Sulfur • Median Fly Ash Size It should be noted that reducing the entire data base into individual f u e 1 e val u a t ion p r og ram s y i e 1 de d a los s 0 fin forma t ion • An un d e r 1 yin g oxygen level dependence, for instance, might be masked because each test program was performed under tighter ranges of this variable (i.e., 2-3% and 6-7% 02' instead of 2-7% 02). Stepwise Regression Comparisons of different coal performances within the same test program were easily made through the use of uniform operating conditions. The utility of the presented analyses was reduced as stated previously, however, due to the limited parameter ranges tested in individual programs. Although the unified data base displayed minor imbalance or bias in some of the independent variables, a great deal may be gained by analyzing the heat flux responses of the combined tests. Due to the |