The Study Of BP Networks's Application In The Plume Thermometry For The Rocket

The solid propellant rocket engine (SPRE)'s main function and operation life depend on its propellants. Plume temperature of the rocket is an important physics parameter that reflects the propellant's combustion condition, the temperature measurement for SPRE plays a dominant role in the total diagraph. Through the temperature measurement of the rocket engine spout, we can judge the function of the related engine equipment and its propellants; also it is important for choosing the adiabatic material of the spout, researching the directions for producing propellants and improving the design of explosive pillar and the research of increasing the impulse ratio; as well it can prove the correctness of the supposed condition and the main data., The emissive platform's combustion eclipse at high temperature in the system which adopting the high power propellants is more prominent than the system which adopting the medium power propellants. The influence of plume temperature should also be taken into account in the vector control technique of various missile weapons. Measuring qualitatively or quantitatively the flame temperature and its distribution plays an important role in observing and understanding the engine's combustion process and flow field ,the intrinsic characteristic of combustion products, setting up reasonable combustion model, carrying accurately on the computer imitative analysis. However the study of plume temperature measurement for SPER is proposed basely on the reliability of flight in the manufacture experiment and reservation experiment Studying the real plume temperature measurement for SPER by using neural networks technique is important for improving our country's diagnostic technique of plume temperature for SPRE. It's of important strategic meaning and huge economy value for defining operation life of the rocket scenically. Spectral thermometry's essence is measuring the single-wavelength radiation strength. Its wavelength is directly proportional to brightness and related to radiator's absolute temperature. Brightness thermometry at high temperature measurement depend on that the blackbody radiation is graduated in the same wavelength. As objective single-wavelength radiation's strength is different to the absolute blackbody radiation's strength, however, this thermometry is influenced by spectral emission efficiency. The result in the high temperature measurement is not the true body's temperature, they are different. The solid propellant's emission efficiency can be accurately given, but spout flow's emission efficiency only be approximately given. Metallic solid propellant contains twenty percent aluminum, yet combustion products contain thirty-five percent coagulation aluminum oxide. The mixed solid propellant combustion temperature is 3000-3500K, spout temperature is 1500~2000K, and coagulation aluminum oxide's fusion temperature is 3200K. To increase the accuracy of the temperature measurement need to measure the temperature of the spout flame's temperature. Multispectral thermometry utilizing the information of multispectral radiation measurement through the data processing abstains spectral emission efficiency of objective true temperature material, which is abstained dependently on the information of objective spectral emission efficiency, while it is relative to the material's components, facial statement, measurement angle, and wavelength. So it is a problem using radiation thermometry which needs to solve the linear projection relationship between objective radiation and true temperature. Neural networks imitate the human's brain by specific topological structure and algorithm. Its strong ability of learning and the prominent input-output projection characteristic has been valued early. Theory of neural networks is extensively applied to signal processing, system modeling, pattern recognition, data compression, system recognition, neural controller, artificial intelligence, biomedical engineering and so on. BP neural network is the most extensive network model and is global network theoretically. Using theory of neural networks for processing the nonlinear relationship existing in flame measurement overcame the shortcoming of supposing...