We proposed a generative design method for the unit cells of an ultra-wideband acoustic meta-silencer and experimentally validated the noise attenuation performance of a designed silencer. A novel internal structure of a unit cell was presented. Additionally, a shape optimization problem was formulated to maximize the width of its band gap. Various optimized unit cells were generated by solving the formulated optimization problem for different initial shapes. The target noise reduction of the acoustic meta-silencer at each frequency was determined considering the current frequency-dependent characteristics of the exhaust noise and the target overall noise reduction level. The noise attenuation performance of the generated unit cells was estimated using a logical evaluation algorithm with mathematical expressions. It quantitatively calculated the contribution of each generated unit cell to the overall noise reduction. Based on the estimated noise-attenuation performance of the optimized unit cells, twelve unit cells were sequentially selected and arrayed to form an ultra-wideband silencer. The acoustic behavior of each selected unit cell was investigated graphically. Moreover, the effect of the order of the unit cells on the noise-attenuation performance of the acoustic meta-silencer was discussed. An acoustic meta-silencer designed using the proposed generative design framework was fabricated, and its noise attenuation performance was experimentally validated. Its noise attenuation frequency range was over 5000 Hz in the frequency range of 40–6000 Hz with an overall noise reduction level of over 30 dB.
