Research on Filtering Algorithms for LMS-based Pressure Testing Systems

Abstract

Pressure testing systems are widely used in industrial, aerospace, and other fields, where environmental noise is a key factor affecting the accuracy of testing systems. To address the problems of noise interference in pressure testing systems and the drawbacks of the classical noise filtering algorithm LMS, such as slow convergence speed and insufficient time-varying signal tracking capability, this paper proposes a denoising method based on an improved adaptive filtering algorithm. Specifically, the Root Mean Square Propagation adaptive learning rate adjustment mechanism is introduced into the fixed step-size module of the LMS algorithm to optimize the weight update process of the algorithm. By leveraging the ability of RMSProp to adaptively assign learning rates to different parameters during the iteration process, the problem of convergence stagnation caused by premature learning rate attenuation is effectively avoided. Meanwhile, an attenuation coefficient is incorporated into the improved algorithm to suppress oscillations during gradient updates, thereby significantly enhancing the stability and robustness in time-varying noise environments. Experimental results show that the improved algorithm increases the convergence speed by 69.5% and improves the signal-to-noise ratio by 17.7%, which enhances the signal processing accuracy and anti-interference capability, and satisfies the denoising requirements of ground tests for pressure testing systems.