The invention provides a crack width variation test method based on strain gauge. A crack width variation test device used in the method comprises the first strain gauge and the second strain gauge, the cross-joint section of the first strain gauge is provided with a protruding part which does not generate strain along with the change of the crack width, and the cross-joint section of the second strain gauge generates strain along with the change of the crack width. When the method is used for testing, the strain difference value delta epsilon 21 of the first strain gauge and the second straingauge is measure and substituted into the formula that delta f equals to l2delta epsilon 2f and equals to l1delta epsilon 21, and the variation delta f of the crack width can be obtained, wherein l2fis the length of a sensitive grid of the cross-joint section of the second strain gauge, epsilon 2f is the strain capacity of the cross-joint section of the second strain gauge, and l1 is the lengthof a sensitive grid of the first strain gauge. It is known from the formula that the crack width variation obtained through the method can exclude size change values, generated due to stress, of components on the two sides of a crack, and therefore the measurement accuracy is improved.