Abstract: Cavitation erosion is widespread in marine vessels and pipeline transportation, posing a serious threat to equipment service life. Mitigating cavitation erosion has long been a research focus. This paper employs an experimental method by pre-setting air bubbles on the wall to control the direction of cavitation jets, thereby achieving the effect of preventing cavitation erosion. The study focuses on the coupled dynamics between pre-set single bubbles, double bubbles, and cavitation bubbles, as well as the patterns of jet direction alteration. A Multilayer Perceptron (MLP) model is constructed to intelligently predict the protective range of cavitation for different cavitation bubble arrays. The results indicate that the collapse modes of a cavitation bubble near pre-set bubbles can be classified into three types: the free-surface dominated mode, the solid-wall dominated mode, and the weak interaction mode. In the free-surface dominated mode, influenced by the bubble, the direction of the collapse jet is away from the solid wall, significantly reducing the degree of cavitation erosion. Consequently, the region where this mode occurs is defined as the effective protective region of the pre-set bubble, which is crescent-shaped for a single pre-set bubble. The dual-bubble configuration exhibits a stronger coupling mechanism during the bubble collapse process. However, increasing the number of bubbles does not enhance the protective intensity in the direction perpendicular to the wall. Nevertheless, due to the increased area covered by the bubbles, the cavitation protective region expands linearly. The MLP model demonstrates good classification accuracy and generalization ability, enabling accurate prediction of cavitation protective regions for various bubble arrays. This provides an intelligent, data-driven solution for the evaluation and optimization of protective effects under complex practical conditions.