Improved hybrid wavelet neural network methodology for time-varying behavior prediction of engineering structures

摘要: An improved neuro-wavelet modeling (NWM) methodology is presented, and it aims at improving prediction precision of time-varying behavior engineering structures. The proposed distinguishes from the existing NWM by featuring distinctive capabilities constructing optimally uncoupled dynamic subsystems in light redundant Haar wavelet transform (RHWT) optimizing neural network. In particular, two techniques imitating packet RHWT reconstructing major crests power spectrum analyzed data are developed with aim data. resulting make underlying law more tractable than raw scale subwaves arose RHWT, they provide a platform for multiscale via individual subsystem level. Furthermore, on each subsystem, technique optimal brain surgeon conjunction new mechanism refreshing employed to optimize network, recombination outcomes every constitutes overall behavior. NMW offers feasible framework due its flexibility, adaptability rationality, particularly useful applications As an illustrative example, applied model forecast dam deformation, results show that possesses positive advantages over single techniques. promising valuable safety monitoring extreme event warning