Published-test-benchmarked and uncertainty-aware data-physics formula for blast-induced displacement of clamped steel-concrete composite panels

Clamped steel-concrete (SC) and steel-concrete-steel (SCS) composite panels are widely used as protective wall and slab components, but rapid prediction of blast-induced displacement remains difficult when near-field loading and nonlinear core-faceplate interaction produce large deformation. A published-test-benchmarked and uncertainty-aware data-physics framework is developed for the maximum inward displacement of clamped concrete-core steel composite panels under hemispherical surface-burst loading. Published SC/SCS blast and pressure-pulse tests are screened into explicit evidence tiers. The detailed nonlinear finite-element (FE) workflow is checked against permanent-displacement evidence and two external first-peak load cases, with present-FE peak deviations of +11.1% and +2.8% at elastic and inelastic response levels. A 109-case nonlinear FE database provides a controlled computational expansion over Z = 0.405–5.426 m/kg 1/3 , t c = 350–1200 mm, t s = 6–28 mm, f c = 30–70 MPa, and f y = 345–550 MPa. The dimensionless analytical formula achieves a root-mean-square log error (RMSE log ) of 0.250, R 2 = 0.992, and 108/109 factor-of-two coverage. A Gaussian-process residual layer reduces the cross-validated RMSE log to 0.123 and decreases the conditional multiplicative surrogate-error factor, evaluated relative to deterministic FE outputs, from 1.63 to 1.27. Monte Carlo, partial rank correlation coefficient (PRCC), and Sobol analyses show that pressure demand dominates full-domain variance, while scaled distance becomes a strong conditional driver for Z < 1.5 m/kg 1/3 . The framework is intended only for preliminary screening within the calibrated FE domain; reported errors quantify surrogate-to-FE agreement, not general physical prediction accuracy or a total physical 95% prediction interval, and transfer to new concrete or panel systems requires case-specific validation.