CNC tool accessories are affected by cutting vibration and load impact for a long time, and are prone to fatigue damage. However, traditional optimization methods often ignore the synergy between life and performance, making it difficult to balance service life, structural stiffness, deformation control and other indicators. Therefore, this study proposes a multi-objective topology optimization method that integrates life prediction models. To address the correlation characteristics between vibration signals and life degradation, this study first preprocesses the vibration signals through interpolation method and 3σ criterion. It builds a complete integrated empirical mode decomposition-Fine-to-Coarse signal processing framework with adaptive noise, extracts multi-scale sensitive features, and then designs a CNN-BiLSTM life prediction model. Experiments show that the MAE and RMSE of this model are as low as 0.007 and 0.008, and the average prediction accuracy is 96.1%, which is better than the comparative method. In the test using milling machine fixtures as the object, the proposed multi-objective optimization method of BPNN-PSO converges after 90 iterations, the generation distance is 0.011, the proportion of remaining life after optimization is up to 0.96, and the deformation amount is smaller than other methods. Experimental data demonstrate that the proposed method realizes favorable coordination between precise remaining life prediction and structural performance improvement. Under tested working conditions, this approach attains a satisfactory trade-off between fixture service life and deformation control, offering an effective technical reference for the design of CNC tool attachments.
