Multilevel Adaptive Near-Lossless Compression in Edge Collaborative Wireless Sensor Networks for Mechanical Vibration Monitoring

Abstract

To address the difficulties of severe lack of storage and computational resources and high delay of transmitting massive vibration data in wireless sensor networks (WSN) for mechanical vibration monitoring (MVM), this paper proposes a novel multilevel adaptive near-lossless compression in edge collaborative WSN (MANLC-ECWSN) for MVM, which could effectively solve the above problems. On the one hand, the sparse pattern of mechanical fault signals is analyzed. A multi-level adaptive near-lossless compression (MANLC) method is proposed to characterize mechanical fault feature information with high accuracy in low storage space, and the proposed method is implemented on the self-developed acquisition node (AN), which effectively improves the storage space and transmission efficiency. On the other hand, the edge computing (EC) technique is integrated into WSN, and data reconstruction and high-precision feature detection are efficiently implemented on the self-developed edge computing node (ECN), which effectively reduces the storage and computing pressure of the data center server. The comprehensive experiments demonstrate that high precision data reconstruction and feature detection could be achieved in the proposed approach from measurements that occupy a little storage space of the AN. The computational power and transmission efficiency of WSN are significantly improved, which provides a potential solution for practical engineering applications.