Abstract: As the performance metrics and technical requirements in aerospace engineering become increasingly stringent, traditional methods face numerous challenges in addressing large-scale, multi-condition finite element model updating problems. The rapid development of deep learning technologies provides new perspectives for finite element model updating. The frequency response function (FRF) contains all the dynamic information of a structure, making FRF-based methods a popular topic in current finite element model updating research. However, traditional FRF-based model updating methods suffer from several issues, such as the selection of frequency points, numerical instability in the sensitivity matrices, poor noise resistance, and difficulties in handling coupled modes. To address these challenges, we propose a robust model updating method based on the feature map of frequency response functions (FMFRF). This approach treats the model updating as a forward problem by using a Bayesian Convolutional Neural Network (BCNN). The FRFs of multiple sensors are combined and transformed into feature maps, which are then used as inputs to the BCNN. The output of the BCNN is a vector composed of uncertain model parameters. The Bayesian framework helps mitigate overfitting in neural networks when dealing with small datasets, thus enhancing training robustness and improving the accuracy of mapping the relationship between FMFRFs and model parameters. This method retains the advantages of traditional FRF-based methods, such as eliminating the need for mode identification and mode matching, and allows for simultaneous updating of structural parameters and damping. Moreover, it significantly improves both efficiency and noise immunity. This paper first demonstrates, through numerical simulation, that the proposed method achieves higher accuracy and stronger noise immunity compared to the classical frequency response-based iterative method, and is more efficient in updating under multiple operating conditions. Then, the method is experimentally validated using a titanium alloy wall plate with six reinforcing ribs, proving its effectiveness in complex engineering structures.