An efficient, adaptive algorithm for large‐scale random vibration analysis

An efficient, adaptive and robust algorithm is proposed to reduce the cost of large-scale stationary and transient random vibration analysis of structures excited by multiple partially correlated nodal and/or base excitations. The cost saving is accomplished by computing integrals selectively, and yet attempting to maintain a level of accuracy desired by the analyst. Recently, proposed closed-form solutions for fully coherent propagating band-limited white-noise excitation are used to rank approximately the terms in the modal covariance matrix. Terms are then evaluated starting from the most important one, and the computations are terminated in such a way that the accuracy level requested by the user is satisfied in an approximate sense.