The invention relates to the field of
rail transit and steel rail flaw detection, in particular to a multi-period steel rail damage
trend prediction method based on
data mining, which comprises the following steps: S1, training and judging a position mark A display waveform based on a
deep learning model, and establishing a corresponding position mark
signal; s2, combining position mark signals with geographic information such as mileage, rice blocks and GPS, and carrying out coarse alignment on all the position marks; s3, determining ultrasonic
signal sequences C and Q which need to be aligned through the aligned position marks; s4, performing fine alignment on the ultrasonic
signal sequences C and Q through a
dynamic time warping method; s5, tracing the damaged position back according tothe aligned ultrasonic wave sequence; and S6, quantifying the development trend of the same damage in different periods, and analyzing the development trend of the damage. According to the method, based on a
big data mining
algorithm, the periodic change of various damage type characteristic data can be captured, a reliable damage
growth data model is formed, and the damage condition of the railis effectively estimated and judged. Based on the prediction model, effective
predictive maintenance work can be performed on the damage
growth point appearing in the specific
station section rail ina targeted manner, so that gradual change is prevented, and unnecessary
periodic maintenance cost is saved to a great extent.