OCR Text |
Show The utility tests confirmed our expectations that the phenomenon does exist, and we are able to extract sensible, meaningful, and repeatable information from flame frequency spectra. Our observations have allowed us to make a number of important, sometimes unexpected, conclusions. F or example, we detennined that one of the most important practical features of our system (which is probably the most attractive for a field engineer) is its capability to rapidly identify out-of-tune or unstable burner conditions. As a result of our utility tests, we received answers to a number of important questions concerning flame fluctuations for coal-fired low NOx burners and, first of all, questions pertaining to the feasibility and the validity of the proposed concept. Does the temporal flame frequency spectra demonstrate repeatable and reproducible characteristics which reflect changes in the controlled flame? Or, as some skeptics claim, the flame spectra reflect only the chaotic nature of the combustion process and thus contain no useful information about individual burner flames (Ref. 3). Is it possible to extract good representative information out of the chaotic flame signals ? The answer to these questions is positive. A properly processed flame signal provides a repeatable and reproducible response to changes in burner flames, caused by variations in stoichiometry, mixing rate and burner load. If indeed we can extract useful representative information from the chaotic movements of the raw flame, it leads to the next important question. The resulting flame signal will be a function of many variables affecting burner flame, such as burner load, fuel-to-air ratio, swirl( s) intensity, flame stability, etc. Is it possible that various combinations of these factors may produce the same result thus preventing us from finding the only optimal solution, or the only optimal combination of these factors? In theory, in a multivariable system with many contributing factors there might be a situation when the same flame spectrum corresponds to different combinations of variables. However, in practical terms, it is probably not important. Burner adjustment based on frequency spectra can be used in the fine-trim mode, after preliminary coarse adjustments of fuel and air flows. In addition, each particular type of burner would have no more than one or two adjustable variables . for achieving the fine-tuned status. For a definite answer to this question, we have to test a single well instrumented burner and to look at all related flame parameters, including 02, CO, NOx and unburned carbon. The next question is related to burner balancing. Our approach to burner balancing can be explained as follows: if two burners of the same design, operating on the same fuel and at the same conditions are equipped with identical flame sensors which are installed in a similar way with the same mounting hardware and aimed at the same zone of the flame, we expect their flame signals to exhibit similar radiational characteristics and could be used for burner balancing and adjustment. This approach is illustrated in Figures 4 - balancing by the frequency flame traces, and Figure 5 - balancing by the calculated statistical values. However, we should remember that there are always some non-flame related factors which may cause deviation between the two signals. For example, defects in the optical system in one of the sensors, an obstruction of view in one sensor, etc. These factors, of course, could cause significant differences which should be corrected in the course of regular maintenance. However, even in this respect our suggested approach is attractive. First, we propose to work with the outputs from conventional existing flame scanners for which maintenance procedures are well developed and organized. Second, we suggest to operate with the AC component of the flame which is always less affected by the non-flame related factors . Third, we 6 |