The invention discloses a gross error elimination method for dam displacement data. The gross error elimination method comprises the steps of setting weight coefficients for four off-group point judging criteria, namely a Dixon criterion, a Grubbs criterion, a Chauvenet criterion, and a quartile criterion, and setting a gross error judging threshold value; then, for each measured value, carrying out judgment respectively by using above four criteria, and if a judgment result under a certain criterion is a suspected gross error, accumulating the weight coefficient corresponding to the criterion; and at last, obtaining a weight coefficient sum of the measured value, and comparing with the gross error judging threshold value, wherein the measured value with the weight coefficient sum up to or exceeding the gross error judging threshold value is judged as a gross error. According to the gross error elimination method disclosed by the invention, multiple off-group point data identification methods are considered comprehensively, the possibilities of misjudgment and judgment omission are weighed, and a scientific and reasonable method is provided to eliminate the gross error in the dam displacement data.