Study And Application On Compressive Sampling Based On Adaptive Sparse Representation For Vibration Signal

Along with the rising of the people’s safety consciousness, structural health monitoring(SHM)has become a hot research subject in civil engineering nowadays. Especially for large-scale structure, lots of sensors are needed to acquire a large number of vibration signal in SHM system.So the sampling method based on Nyquist sampling theorem has brought great pressure to data storage and transmission.The compressed sensing(CS) theory is put forward in recent years. It provides a new approach to signal compressive sampling. The CS theory proves that sparse or compressible signal can be reconstructed from a small set of linear measurements with optimization algorithms.Therefore signal sampling frequency can be below Nyquist rate. On the basis of study on CS theory and vibration signal feature, a compressive sampling system based on adaptive sparse representation for vibration signal is designed in this dissertation. The main works of this paper are as follows:1. Study on CS theory systematically in terms of signal sparse representation, measurement matrix and signal reconstruction algorithms. And an adaptive sparse representation base which called adaptive AR basis is constructed based on vibration signal linear feature. Compared with fixed orthogonal bases like DCT and DFT, the adaptive AR basis can improve the sparse representation of vibration signal significantly to guarantee the accurate reconstruction.2. Three signal compressive sampling methods described in this paper are random filtering,pseudorandom sampling and uniform random sampling. And the corresponded measurement matrixes are derived and proved to be irrelevant to DCT basis. Finally the relation of compressive sampling methods and reconstruction quality is illustrated by simulation. Instead of Nyquist sampling, the proposed CS-based signal sampling methods can improve the reconstruction precision and reduce the samples.3. A compressive sampling system for vibration signal based on adaptive sparse representation is presented. The system employs the adaptive AR basis and uniform random compressive sampling method. The hardware circuit is designed based on ADuC7026 microcontroller to sample the vibration signal of cable-stayed bridge model. The real vibrationsignal is reconstructed completely later. The experiment shows this system proposed can compress the vibration signal effectively and reduce the complexity of hardware.Given all that, this thesis describes a compressive sampling system based on adaptive sparse representation for vibration signal. It focuses on the sparse representation basis matrix and compressive sampling techniques. Related experimental results demonstrate this system can effectively decelerate the sampling frequency and ensure the accurate reconstruction of vibration signal.