A MODAL-TEST-UPDATED GEOMETRICALLY NONLINEAR REDUCED-ORDER MODEL FOR PREDICTION OF RESPONSES UNDER INTENSE ACOUSTIC LOADING OF A COMPOSITE STIFFENED PANEL 1)

Accurate and efficient prediction of acoustically induced vibration is essential for the acoustic fatigue-resistant design of aircraft thin-walled structures. To address geometric nonlinearity under intense random acoustic loading and discrepancies in dynamic characteristics between nominal finite element models and test structures, a nonlinear reduced-order modeling (NLROM) method calibrated using modal test data is proposed. The primary innovation lies in an integrated framework comprising low-amplitude modal-test calibration, renewed model reduction of the updated model, and independent validation under intense acoustic loading. First, reduced-order dynamic equations incorporating quadratic and cubic nonlinear stiffness terms are formulated through projection onto modal coordinates, enabling the bending–membrane coupling and nonlinear modal interactions to be represented in a low-dimensional form. Second, a parameter-calibration strategy based on the frequency sensitivity of the first dominant mode is proposed. Low-amplitude modal tests conducted under installed boundary conditions are employed to identify sensitive equivalent parameters. Following parameter updating, the modal basis and nonlinear stiffness coefficients are recalculated, thereby incorporating information from the low-amplitude tests into the prediction of responses under intense acoustic loading. The response data measured under 163 dB acoustic loading are not used for parameter identification and are reserved exclusively for independent validation. A flat-plate example demonstrates that the proposed NLROM reproduces the principal resonance peaks of the full-order finite element model. The relative errors in the displacement root mean square (RMS) values at measurement points A and B are 4.0% and 0.73%, respectively, while the computational efficiency is improved by approximately 14.2 times. In the 163 dB acoustic test of a composite stiffened panel, the updated-parameter NLROM yields relative errors of 12.6% for the acceleration RMS at measurement point 6 and 0.46% for the strain RMS in the 0° direction at measurement point 5, outperforming both the initial-parameter model and the updated-parameter linear model. Comparisons among the four models indicate that modal-parameter updating primarily improves the dominant frequencies and overall response levels, whereas the incorporation of geometric nonlinearity further improves the prediction of response energy under intense acoustic loading. Within the investigated structure, the loading frequency range of 50–600 Hz, and the validated response-amplitude range, the proposed method achieves a favorable balance between predictive accuracy and computational efficiency, providing a practical reference for the rapid assessment of intense acoustically induced responses in thin-walled structures.