Application Studies Of Fault Detection Technology For LRE Startup In Ground Test

The turbine pump and gas generator combined system of a LOX/Kerosene staged combustion cycle rocket engine is investigated in this paper. The mathematic models describing the startup transient process are provided. The test sample database consists of hot test data and numerical test results. On the base of the comparison and analysis for variant fault detection algorithms of liquid rocket engine, the Envelope Algorithm (EA), the Back Propagation algorithm (BP) and the Radial Basic Function algorithm (RBF), which are different Neural Network algorithms, can be adopted to detect the fault during startup process for this engine system. These three algorithms have outstanding advantages in real-time and accuracy. In this paper, princeple of these algorithms is explained in detail. EA?BP and RBF have been used in Matlab. Eight parameters are chosen and dived into two groups to detect fault in EA, ten input parameters and eight outputs were adopted to combine network in BP, nine inputs and six outputs are taken to form network in RBF. The fault detection decision principle is to show warning signal when a fault is detected by at least two algorithms of EA?BP and RBF. It is validated by hot test results and numerical test data that the developed fault detecting platform of turbine-pump and gas generator combined power system of LOX/Kerosene liquid rocket engine is accurate and available.