| OCR Text |
Show low temperature, the reaction rate for the formation of N O x is slow. This contrasts to the operation of the older style burners that operate with a combustion temperature range where there is a substantial slope to the curve and with combustion temperatures where the reaction rate is much greater. In addition, the actual increase in flame temperature is less for the internal flue gas recirculation burner when the fuel hydrogen content is increased because of the additional mass in the primary combustion zone. Thus, increases in fuel hydrogen content have very little impact on the production of N O x in a well designed internal flue gas recirculation burner when contrasted to older style burners where the fuel hydrogen content is a significant factor in the quantity of N O x produced. Within an eighteen month period, Callidus built a data with over 2,500 data records that were used for generating correction curves. The data records included variation in operating parameters such as excess air, furnace temperature, combustion air temperature, fuel composition and fuel splits. Because of the large quantity of data for a single type of burner, it was possible to generate correction curves that provided reasonable accuracy. The correction curves were certainly a better correlation than the correlations for older types of burners that were developed using smaller data bases. However, problems that were associated with the correlations for the older style burners were also present with the correlations for the newer low emission burners. For example, the correction factor for preheated air is not necessarily the same when different fuels are used. Therefore, to be completely accurate, a series of curves must be developed that show N O x as a function of preheated air for various fuel compositions. The above is true for variations in other operating parameters. From the previous example it is easy to see that development of a set of curves to predict N O x levels would be very difficult, especially in light of the inability to generate a reasonable test program that examines all the possible combinations of variables. Recently, with an expanded data base of over 4,000 data records, Callidus Technologies utilized a software package called Process Insights that has enabled us to accurately consider the variation in many variables to predict N O x levels from the combustion process. Process Insights utilizes artificial neural networks which represent a set of very powerful mathematical techniques for modeling, control and optimization. Neural networks are an ideal technology for exploiting historical data to build prediction models. Process Insights "learns" the relationship of the various parameters such as excess air, combustion air temperature, fuel composition, etc. to N O x production. Because of the large data base available for the L E burner, Process Insights is the ideal tool to assist in the understanding of these relationships. The models generated with Process Insights makes it possible to predict N O x levels for the Callidus L o w Emission Burner for an almost infinite combination of operating conditions. The models can be used to construct a series of curves that predict how the N O x emission level will change as one parameter is changed and all other iv-21 |
