Research On Loose Particle Recognition And Location Method For Rocket Engines

As the power plant of aerospace system, rocket engine’s reliability is a critical factor in determining the success or failure of the space launch missions. Rocket engine has the characteristics of large volume and complex pipeline, the loose particle which produced in production and assembly process is an important factor affecting the reliability of the engine. Due to the presence of loose particle inside the engine, the major aerospace powers have lessons from the failure of the launch mission. Therefore, it is significant to detect loose particle to ensure the reliability of the engine. This dissertation has made an intense study of detection device and detection method for loose particle detection of rocket engine.Considering that the shape, weight and structural characteristics of the rocket engine, a excitation generator consist of inverter and induction motor was designed, which drives the rocket engine to rotate uniformly, and ensures the effective activation of the loose particle and the safety of the engine. Based on particle impact noise detection(PIND) theory, a sensor placement scheme and signal amplifier and acquisition circuits with eight channels are designed, which ensures the sampled signal can fully reflect the characteristics of the original signal. Providing the hardware foundation for the no blind zone detection and location of loose particle.Considering that the electromagnetic interference caused by inverter and sensor cable bundles, the variable frequency drive system model was established based on Simulink, analyzing the common mode voltage in mechanism and characteristics, then a design method of second order passive filter was proposed. The nine cores cable model was established based on Maxwell, investigating the influence of shield and cable spacing on distributed parameter, providing a reference for the structural design of the sensor cable bundle.Considering that the misjudgment of loose particle caused by inherent mechanical signals, three threshold endpoint detection method was applied to decompose PIND signals into pulse sequences, and pedigree clustering algorithm was designed to cluster pulse sequences according to the correlation matrix. Based on the time domain difference between inherent mechanical signals and loose particle signals, analyzing the standard deviation involved in pulse energy, peak voltage and pulse length, a combined standard deviation method was proposed to identify category attributes. The method can determine whether the loose particle is present, and determine the number of inherent mechanical signal types simultaneously.Considering that the difficulty in finding the position of loose particle, a location method which combined channel location and coordinate location was proposed, this method transforms the complex three-dimensional space location problem into one dimensional linear location problem. Based on the attenuation characteristics of acoustic emission(AE) signals, average energy and peak voltage were applied in channel location. Baded on the chaos characteristics of PIND signals, calculating chaotic parameters involves correlation dimension, Lyapunov exponent and Kolmogrov entropy. Those parameters were adopted in p article swarm optimization BP neural network algorithm to determine the coordinate of particle. The experimental results show that the accuracy rate of channel location is 85.12%, the maximum absolute error of coordinate location is 5.25 cm, the mean absolute error of coordinate location is 2.34 cm.