2014 年 80 巻 811 号 p. DR0051
A smart sensor is required if we are to popularize health monitoring using the vibration of a structure. It is used to evaluate the vibration characteristics of the structure. To detect the variance in the vibration characteristics from the vibration data of a structure, a time-frequency analysis technique that can detect the change of the frequency components of the vibration data over time is needed. In a previous report, we have proposed a real-time data processing method with arrayed narrowband band-pass filters based on a single degree of freedom vibration system for the evaluation of structural vibration characteristics. This method is applicable to processors with low computation loads and memory capacities. In this paper, the relation between the parameter setting and the accuracy of the evaluation result when the proposed method is applied to several vibrational environments is examined. Concretely, accuracy and the stability of the estimated result of the natural frequency of the structure were examined when a steady state noise is included to the response of the structure. Furthermore, a similar examination was conducted for the measurement data of the microtremor response of the pier model where the unsteady input is applied. Moreover, to obtain the criterion of the evaluation accuracy of the proposed method, the estimate was compared with AR model. As a result, it is clarified how each parameter of the proposed method was related to the estimation accuracy of the natural frequency. The proposal method showed that the influence on the noise was less than the AR model.
