The Impulse Signal Detection Method Based On Wavelet Networks

This paper studies the issue that how to detect transient impulse using wavelet network in the context of strong vibration signals, and this method is tested to be effective by using actual vibration signals.Individual member loosing or plate and shell crack in truss structure is one of main fault that led to system failure, in the state monitoring of large-scale structures and bridges. In early stage of structural failure, the area of partial damaging within the system will appear unsteady, and it will lead to reallocate of load in an instant, when external load large change. The process of the load reallocate in the internal structure will be performance in external structure: can be observed impulse intermittent signal in the system response as the load changing, by using test vibration acceleration of structure. Therefore, monitoring the impulse phenomenon in system response can be found the damage of system. However, the signal that is collected in the actual condition monitoring system is the response that stochastic work load produce, the abnormal impulse response that is caused by member instability or plate and shell crack immersed normal stochastic signals which is caused by work load, this phenomenon is very difficult direct-viewing recognition in the time-domain waveform. But the impulse signal and the stochastic response all has the very wide frequency band, and the frequency overlap mutually, it is also very difficult direct-viewing resolution in the frequency range.The structure itself is a multi-degree of freedom second order linear system in the response processing that large-scale structures and bridges and so on mechanics systems under external force. Comparing with the impulse response produced by member loose or plate and shell crack, the response nature which the structure bear because of transportation, wind power, temperature change and earth pulsation and so on is continuously and relatively slow. Therefore, this article uses the second generation of wavelet the multi-scale and the good time frequency characteristic, as well as with the information entropy responded the signal the statistical distribution characteristic, processing includes the impulse vibration signal, finally indicated: With the second generation of wavelet entropy analysis method can accurate from the strong background noise determine the impulse signal appears time.The paper first chapter is an introduction, introduced the research work background and the significance; Second chapter introduced the wavelet analysis basic concept and the theory, finally introduce the wavelet network basic concept; In third chapter discussion wavelet network structure, study algorithm as well as wavelet network wavelet base choice; Finally wavelet network application in vibration signal denoising; Fourth chapter has discussed the second generation of wavelet entropy, and will be weak vibrates the impact signal localization.