Abstract: Objectives To improve the diagnostic performance of marine engine fault diagnosis models under cross-operating conditions, an intelligent fault diagnosis framework based on industrial foundation model technology for marine engines is proposed. Methods A deep learning model, ResNet-18, is used as the feature extractor for engine faults. Initially, a pre-trained model is constructed based on source domain data, and then a fine-tuning classifier strategy is employed to transfer the fault features from the pre-trained model to the target domain, enabling cross-operating condition fault diagnosis. Results The effectiveness of the proposed framework is validated using fault data from the lubrication and intake/exhaust systems of marine engines. Experimental results demonstrate that the framework achieves diagnostic accuracies of more than 94% and more than 98% for the lubrication and intake/exhaust systems, respectively, in fault diagnosis tasks under cross-operating conditions. Additionally, a comparison with the VGG-11 model, used as a feature extractor, highlights the superiority of the proposed framework. Conclusions The proposed intelligent fault diagnosis framework for marine engines, based on industrial foundation model technology, effectively enhances the cross-operating condition fault diagnosis performance of the fault diagnosis model, significantly improving its generalization capability.