Abstract: ZrO₂ materials are widely used in various fields due to their excellent properties. However, when used as coating materials in combination with metal substrates, corrosion cracks are prone to occur. Constructing a model for ultrasonic guided wave detection of corrosion cracks in ZrO₂ coatings using finite element software, and analyzing the feasibility of ultrasonic guided wave detection of corrosion cracks in ZrO₂ coatings, where the ZrO₂ coating is set as an inner coating on the inner side of the metal substrate. Collect ultrasonic signals of corrosion cracks at different depths based on simulation results, analyze the corresponding relationship between the depth of corrosion cracks and defect signals, and use SPWVD time-frequency analysis method to analyze the corrosion crack signals, extracting characteristic quantities that can characterize corrosion cracks from the signals. Sample 200 sets of crack signals, construct a dataset of signal feature quantities and corrosion crack depth, and use convolutional neural network algorithm to identify and classify the corrosion crack depth. The experimental results show that the recognition and training of the collected crack ultrasonic guided wave signals using convolutional neural network algorithm achieved a recognition rate of 96.7%, providing some reference for the detection and recognition of corrosion cracks in ZrO₂ inner coatings.