| OCR Text |
Show - 3 - TEST PROCEDURES The tests used in this statistical evaluation were completed in the fo 11 owi ng rna nner. The combus tor wa s confi gured to the proper fi ri ng mode and preheated using natural gas for about 12-16 hours. After initiating coal feed and achieving stable coal flames, the natural gas burners were removed and the unit was allowed to equilibrate for approximately 24 hours. Just prior to each test, the slag panel and fouling probes were cleaned with an air lance. Heat flux degradation data and convection section pressure drop changes were then monitored for the test duration utilizing data acquisition software running on a VAX 11/750. Following the conclusion of each six-hour test, solid samples were acquired and qualitative determinations of deposition cleanability were made. STATISTICAL TECHNIQUES A variety of statistical techniques were used to evaluate the slagging and f ou 1 i n 9 data. Th e s e inc 1 u de d the com pi 1 at ion and man i p u 1 a t ion 0 f the d a tab a s e us i n g c h e c k s for norm ali t y , va ria b 1 e t ran s form at ion s , analyses of variance and covariance, least significant difference tests, correlation matrices, explanatory variable scatter plots, multivariate stepwise regressive techniques, residual influence assessments, and linear discriminant analyses. The goals of the statistical evaluation were to identify any imbalance in the data base and to determine which explanatory variables (operating conditions and coal characteristics) influenced slagging and fouling in the pilot-scale combustor. No detailed explanations of the mechanics of the statistical methods are covered here (for the sake of brevity), but the analyses follow along the same lines of numerous related correlative works in the ash deposition field.(l, 10, 14) THE DATA BASE Table 1 is a compilation of the means, mlnlma, maxima, and standard deviations for most of the parameters used in this study. All ash chemical properties are for the as-fired coal; the locational deposit compositions are not presented here. Notation is also provided where the base da ta were trans formed into norma 1 dis tri bu t ions. Tab 1 e 2 is the correlation matrix for the unified data base (all firing modes). In this instance a value of .202, for the correlation coefficient (r), is statistically significant at the 95% confidence level. The fouling qualifiers, deposit tenacity, and convection section pressure drop difficulties, are not included here because of the qualitative nature of these parameters. Pairwise scatterplots of all variables were also generated to establish nonlinear trends and to prevent misleading interpretations due to the effects of outliers. As Table 2 reveals, some inherent imbalance in the test matrix is evident. The as-fired P20S concentration displays a strong positive correlation with oxygen level. Strong correlation also exists between sulfur, CaO, Fe 203 , Si0 2 , and K20 and all ASTM fusion temperatures. These relationships are consistent with coal characteristic impacts on |
