2012 年 68 巻 2 号 p. I_751-I_759
This paper reports on the accuracy improvement in damage detection of bridge joints based on the passing sound of a vehicle using Chaos prediction analysis and frequency analysis. Herein, the factors such as sampling rate, frequency band, supervised data and criteria for judgment were examined. It was found that sampling rate is not significant for POA (partial overall value based on frequency analysis) but slightly affects the accuracy of error prediction. It was also clarified that the index of max by deviation for threshold shows better accuracy than that of Kurtosis for both evaluation methods. As for the frequency band in POA, the band of 210-230Hz and 450-650Hz exhibits the most accurate recognition. Finally it was found that detection performance by POA slightly exceeds that by prediction error.