Study on identification of dam safety monitoring data based on Pettitt's non-parametric test method

In order to cope with the difficulty of identifying abrupt and gradual change features in dam safety monitoring data, the study proposes a novel identification method for dam safety monitoring data by comprehensively applying the standard normal homogeneity test, linear function test and non-parametric test. The experimental results show that compared with the traditional static statistical analysis method and the existing dam safety monitoring data identification techniques, the research method performs best in identifying the mutation features in dam safety monitoring data. The average absolute error, mean square error and average absolute percentage error are only 12.08%, 10.11% and 9.87%, respectively. In addition, the regression equations and regression coefficients of the trend mutation point detection for the horizontal displacement-deformation monitoring series of the 8# monitoring points showed good significance and passed the significance level test of α=0.05. Through this novel method, the study identified the specific dates when the asymptotic trend within the horizontal displacement deformation monitoring sequence of the 8# monitoring points changed abruptly as 3 January 2004 (t-49), 12 March 2008 (t-99), 16 October 2016 (t-181) and 16 August 2023 (t-203), respectively. The identification of these dates provides precise time stamps for dam safety management and helps in further analysing and taking necessary maintenance measures.