Optimization and Performance Analysis of an FPGA-Based Structural Health Monitoring System for Aerospace Composite Structures

Abstract

Structural Health Monitoring (SHM) is a transformative technology in aerospace engineering, enabling real-time assessment of aircraft structural integrity to enhance safety, reliability, and maintenance efficiency. Traditional inspection methods, such as visual inspections and non-destructive testing (NDT), require manual labor and aircraft downtime. In contrast, a SHM system offers continuous and automated monitoring, thereby reducing operational costs and improving aircraft availability. This paper presents an optimization study of a previously developed FPGA-based SHM system for carbon fiber-reinforced polymer (CFRP) laminates. The system uses piezoelectric transducers (PZTs) to both excite and receive Lamb waves and employs signal processing and machine learning techniques for damage detection. In the original implementation, signal processing steps such as discrete wavelet transform (DWT) and feature extraction were executed in MATLAB. In this work, a fixed-point hardware implementation of the DWT was developed and integrated into the acquisition pipeline. The DWT stage, originally implemented in MATLAB, was migrated to hardware, resulting in a processing time reduction from 1.43 seconds to 5.68 microseconds, which led to over 250,000 times speedup. The classification stage, based on a support vector machine (SVM) enhanced with Mahalanobis distance, was preserved in hardware. A comparative analysis shows that the embedded version achieves the same classification accuracy as the softwarebased approach, with no loss of precision. The results demonstrate that feature extraction can be executed in real-time, advancing the system toward fully embedded SHM applications.