The Liquid Rocket Engine Experiment Data Quality Improvement Based on 3σ-LMBP

Abstract

During the ground test of liquid rocket engine, the complex on-site experimental environment will lead to abnormal sensor output, and affect subsequent data analysis. For the acceleration data's zero drift problem, the Levenberg Marquardt BP neural network is introduced in this paper to compensate and correct it, and the standard deviation method is used to eliminate the abnormal data, in which the 3σ-LMBP model constructed. The experimental data process shows that this method can improve the data quality, and ensure the follow-up data characteristic extraction. Finally, aiming at the internal causes of zero drift, the external circuit compensator design are proposed to comprehensively solve the problem of zero drift and improve the data quality.