DATA RECOVERY FROM DE-QUANTIZED COMPRESSED MEASUREMENTS FOR STRUCTURAL MODAL ANALYSIS
Fundamental Research in Structural Engineering: Retrospective and Prospective, Vols 1 and 2(2016)
摘要
Ever-increasing aging civil engineering structures are in demand for Structural Health Monitoring Technologies, outputs of which provide suggestion to the maintenance of structures in a cost-effective manner. Large amount of sample data are required for to infer the state of structures, and this is a challenge in energy expenditure, signal transfer stability and capability for wireless sensor networks. Compressive Sensing (CS) as a new theory for signal processing has potential to handle this challenge. In this work the vibration signal reconstructed by CS are used for modal analysis. In order to minimize data transmission, the vibration signals are converted to compressed measurements by observation matrix with 50% off. And the compressed measurements are further compacted by quantized at very low resolution (4bit and 2bit) comparing with the normal resolution (12-16bit). The data size with the compressed measurement is only less than 20% of conventionally sampled signal. For better application of the de-quantized compressed measurement to reconstruct vibrational signal, we develop a Bayesian de-quantization (BDQ) algorithm under the framework of the block sparse Bayesian learning. This algorithm not only decreases the reconstruction error caused by quantization but also avoid the limitation of the sparse dictionary. Modal parameters are identified by Natural Excitation Technique (NExT) and the Ibrahim Time Domain Technique (ITD). We test the performance of our recovery algorithms for modal analysis. The effects of quantization to modal parameter identification are also compared.
更多查看译文
关键词
compressive sampling,de-quantized sampling,Bayesian learning algorithm,modal analysis
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络