Dam break grading early-warning index extraction method for homogeneous earth dam

The invention discloses a dam break grading early-warning index extraction method for a homogeneous earth dam. The method comprises the steps that a sliding window method is used for dividing a value measuring sequence, a earth dam monitoring physical quantity sliding window matrix is extracted based on principal component analysis (PCA) or kernel principal component analysis algorithm (KPCA), and the evolution of the principal component number/principal component matrix norm with earth dam conditions on the basis of 85% contribution rate is determined. Actual dam break tests indicate that the enlargement and the sudden changes of the two indexes correspond to the condition deterioration and the ultimate break of an earth dam respectively, and it is full demonstrated that the principal component number or the principal component matrix norm can be used as the first-grade early-warning index and the second-grade early-warning index for the safety condition of the earth dam; mutual checking can be conducted between the principal component number and the principal component matrix norm, wherein the first-grade early-warning index corresponds to the earth dam condition deterioration, and the second-grade early-warning index corresponds to the earth dam break.