Frequency Domain Decomposition (FDD): An Output-Only System Identification Technique in Civil Engineering

Frequency Domain Decomposition (FDD) represents a crucial system identification technique in civil engineering, particularly favored in structural health monitoring applications. Its primary advantage lies in identifying modal parameters using only output response data, eliminating the need for hard-to-measure input excitation data, making it a practical tool for real-world engineering applications.

FDD analyzes frequency response data from multiple output channels by constructing a power spectral density matrix and performing singular value decomposition to extract modal parameters including natural frequencies, damping ratios, and mode shapes. The implementation workflow typically involves three key stages: data preprocessing (such as detrending and filtering), frequency domain transformation, and modal parameter extraction. In MATLAB environments, automated analysis can be performed by reading multi-channel time history data (such as acceleration records) stored in Excel files, with accompanying example files helping users quickly understand data format requirements and algorithm deployment methods.

The technique employs singular value decomposition of the power spectral density matrix at each frequency line, where peaks in the singular value plots correspond to structural modes. The first singular vector at each resonant frequency provides the mode shape estimate. MATLAB implementation typically involves functions like fft for Fourier transformation, svd for singular value decomposition, and pwelch for power spectral density estimation.

This technology is particularly suitable for long-term monitoring of large civil structures, such as vibration characteristics assessment of bridges and high-rise buildings, providing data support for structural safety diagnosis and maintenance decisions.