Study on noise reduction method for bridge temperature signal using adaptive parameter selection and improved wavelet threshold function

To address the persistent challenge of environmental noise interference in bridge structural health monitoring, this study proposes a denoising method combining an improved wavelet threshold function with adaptive parameter selection. First, a Coati Optimization Algorithm-Variational Mode Decomposition (COA-VMD) collaborative optimization framework is established, marking the first application of the COA in bridge monitoring. This framework achieves adaptive precision matching of variational mode decomposition parameters (k, α), overcoming inherent limitations of traditional empirical parameter selection in complex environmental signal processing. Second, a novel modal selection theory based on harmonic parameter confidence intervals is proposed. A comprehensive criterion parameter P is constructed through the fusion of approximate entropy and permutation entropy, enabling adaptive inference of noise modes and establishing new standards for modal separation of nonlinear non-stationary signals. Subsequently, an improved wavelet threshold function is designed to resolve the technical bottleneck of effective signal distortion in traditional threshold processing. Validation through both simulated signals and real bridge temperature monitoring data demonstrates: In 15 dB simulation experiments, the signal-to-noise ratio (SNR) improves by 48.7 % with 37.9 % reduction in mean square error (MSE); In practical applications, the noise mode (NM) reaches optimal values while maintaining signal energy ratio (SER) over 99.7 %. This methodology achieves deep integration of bio-inspired algorithms with bridge signal decomposition theory, establishing a new paradigm of “self-optimizing parameters-precise modal identification-gradual noise filtration” for bridge monitoring data denoising.