Aimed at the problem of extracting damage-sensitive features during the damage detection process of jacket platform, a damage detection model based on deep learning is proposed. In order to better fit the characteristics of the Convolutional Neural Network (CNN), this model first converts the one-dimensional time domain signal into a two-dimensional grayscale image; and then extracts the damage features existing in the two-dimensional grayscale image through the CNN which is used for damage detection. Through the tests on the jacket platform model, the effects of different grayscale image generation methods on the detection results are compared. The results of damage detection tests show that: the detection model can identify the damage types with an accuracy of 99.4%, and is of good damage detection capability; the detection model can identify the degree of damage with an accuracy of 96.3%, and it can be applied to damage early warning of jacket platforms.