| Compared to the traditional cable machinery condition monitoring system, the mechanical vibration wireless sensor network could improve flexibility, maintainability, and scalability of the monitoring system, while enable vibration monitoring of key mechanical components in the sealed or rotary environment. However, due to the high frequency of mechanical vibration signals, mechanical condition monitoring usually adopt a higher sampling frequency, which requires wireless sensor network to achieve higher accuracy of synchronous acquisition. This paper research on clock synchronization of wireless sensor network which is the base of synchronization acquisition and then raising the synchronization acquisition accuracy of mechanical vibration wireless sensor network. The contents of paper is as follows.The clock synchronization of single-hop wireless sensor network is divided into two processes: clock offset compensation and crystal oscillator drift compensation. Traditional clock offset compensation algorithm is on the base of MAC layer timestamp exchange, and have the delay of interrupt handling, coding/decoding and byte alignment in timestamp exchange process, which would affect the accuracy of clock offset compensation. The paper adopts the method of SFD physical layer timestamp exchange to avoid the delay in timestamp exchange process except radio transmission and enhance the accuracy of clock offset compensation. For crystal oscillator drift compensation, after analyzing the relationship of quadratic polynomial between crystal skew of common node and the clock of reference node, the clock synchronization algorithm(LSCS) is proposed based on least squares curve fitting crystal oscillator drift compensation. Experiments validate that the average precision of clock synchronization of LSCS is 3.89μs in single-hop networks.As transmission range for wireless sensor nodes is limited, when the domain of nodes goes beyond single-hop transmission range, multi-hop network should be deployed.For multi-hop network clock synchronization, in order to reduce the number of communications between nodes and avoid receiving the synchronization package repeatedly, cluster-tree network model has been selected and then optimizing the hops of network. After hop optimization, the times of cluster-tree network communication is just half of that of the mesh network and the average receiving times of synchronization packet is one-eighth of mesh network. Based on single-hop clock synchronization, LSCS use the method of one-hop clock synchronization step by step to fulfill the clock synchronization of cluster tree wireless sensor network. The results of experiment indicate that the precision of clock synchronization of LSCS in every single jump is 2.75μs.Based on clock synchronization, a solution is proposed to solve synchronous triggering and sampling interval control in synchronization acquisition of mechanical vibration wireless sensor networks. PAN coordinator sets the initial time of acquisition to realize network synchronization acquisition, when PAN coordinator calculates initial acquisition time, it will with consider the latency of receiving and processing acquisition command of end-node. In multi-hop network, this paper use beacon collision avoidance mechanism based on routing depth to avoid the phenomenon that terminal nodes cannot receive collection command due to the conflict of beacons.Using the method of dynamic correction trigger frequency of the Timer which control acquisition could achieve sampling interval control. Using the signal from signal-source and mechanical vibration respectively, the synchronization acquisition accuracy of the method this paper proposed has been verified. The experiments results show that single hop synchronization acquisition accuracy is 4.38μs and double jump synchronization acquisition accuracy is 6.52μs under signal-source, and single hop synchronization acquisition accuracy is 5.35μs under mechanical vibration signal. Finally, compared the two machinery vibration signal gathered by two different nodes in time-frequency domain, the results of experiments show no obvious phase difference in time domain waveform and no significant difference in the distribution of the spectrum. The proposed synchronous acquisition method based on clock synchronization can meet the requirement of mechanical vibration signal synchronous acquisition. |
PDF Full Text Request