A Novel Outlier Detection Method of Long-Term Dam Monitoring Data Based on SSA-NAR

Outlier generally exists in dam monitoring data which may seriously affect the accuracy of safety evaluation results. Aiming at outlier detection data, a novel dynamic method based on SSA-NAR is proposed. This combined does not depend effect quantity and influence relationship traditional theory only uses time series to mine variation, can avoid impact missing or abnormal quantity. The Nonlinear Autoregression (NAR) classical neural network widely used engineering field. However, prediction NAR greatly affected by selection model parameters, Sparrow Search Algorithm (SSA) parameter solution be with derive optimal parameters model. identified through analysis residual distribution between predicted measured data. case study shows that when original contain outliers, high. When included, proposed has good robustness little effect. It effectively detect provide reliable basis for evaluation.