| OCR Text |
Show The blue chemiluminescence from excited CH radicals in the flame zone provides a method for obtaining images of flames in furnaces where background radiation renders the flame practically invisible to the naked eye. In this technique, the intensified CCD camera was used with a filter centered at 431 om and having a bandwidth of 10nm (full width at half maximum). The resulting image provided a time-averaged flame envelope that may be used to qualitatively define flame length and regions of high reaction rate, e.g. the flame front. Design of the Validation Tests Prior to proceeding with the follow-on measurements, a number academic and commercial modeling groups were surveyed to provide feedback on the kind of measurements and information needed to aid in validation of multi-dimensional combustion models. Information obtained in this survey was then incorporated into the construction of the second set of measurements and the planned validation document. In addition, feedback was solicited from the same groups on an initial release of the validation document based on the original measurements. This feedback resulted in additional improvements in the overall document and test plan. Predictions were completed with Babcock & Wilcox's combustion model, COMO (Fiveland and Jessee, 1994, and Jessee and Fiveland, 1995a) prior to testing and during the follow-on measurements to provide input to the test plan (Kaufman and Fiveland, 1995). Two-dimensional (2-D) axisymmetric and three-dimensional (3-D) predictions were compared with the original data of Sayre et al (1994) to assess the use of the hot- and cold-wall cases for validation before beginning the additional measurements. A burner and furnace geometry that could be used to evaluate both axisymmetric and full 3-D predictions was desired, and the initial modeling indicated that the IFRF burner and BERL geometry worked well for both types of applications. Discrepancies between the 2-D predictions, 3-D predictions, and the data were then used to determine where additional measurements were required. These assessments lead to the additional traverses near the quarl and through the internal recirculation zone. Based on observed differences in predicted downstream mixing between the 2-D and 3-D models, as shown by the predicted oxygen contours in Figure 7, several profiles were also added just past the main combustion zone, with additional profiles in the post-flame region to evaluate downstream mixing and further establish the external recirculation zones. The LDV traverses at the quarl exit were added to provide additional definition of the velocity field as it left the burner. The modeling results, together with the existing data, were used to define the desired level of measurement resolution across the flame at each traverse. A preliminary assessment of the low excess air condition was also completed using axisymmetric models prior to testing. The models were also used to evaluate the change from a 5-spool BERL configuration to a 6-spool configuration. The 5-spool furnace configuration had been used for the original measurements. However, the use of the existing 6-spool BERL configuration at the time of the additional measurements would save considerable effort and expense, while providing more time for collecting data. Comparisons between 5-spool and 6-spool cold-wall predictions showed only minor changes in the downstream flow patterns, with no significant effects on the nearflame region. This evaluation was borne out by the subsequent measurements. APPLICA TION TO MODEL VALIDATION Model Description The COMO model was used throughout this effort to predict the natural gas flames. COMO is a modular, fundamentally based multi-dimensional flow and combustion model. Both structured, orthogonal and unstructured body-fitted versions are currently available. Gas-phase equations are solved using a collocated, finite volume formulation for the Eulerian equations. The flow algorithm is built around a cell-vertex finite volume formulation of the steady, incompressible Navier-Stokes equations (Jessee and Fiveland, 1996). Turbulent flow is predicted using a standard two-equation k-£ turbulence model (Launder and Spalding, 1974). A solution the radiation heat transfer in the system is found using the discrete ordinates method of Fiveland and Jessee, 1995, while the combustion model is similar to that described in Fiveland and Jessee (1994). Solutions are obtained using algebraic mUltigrid techniques to enhance convergence. The model has been validated in a step-wise manner as described in Fiveland, et aI, 1996. 6 |
