The invention provides a local
artificial intelligence earthquake sand
liquefaction deformation prediction method and
system, and relates to the technical field of
geological disaster prediction, and the method comprises the steps: obtaining multi-source geological data; 3-5Hz
soil mass frequency domain features are extracted through
wavelet packet
decomposition, a DEM and a historical
liquefaction thermodynamic diagram are fused to divide a sedimentary
facies control unit, a three-dimensional
voxel model is generated by adopting a Bayesian constrained improved
Kriging algorithm, and a four-dimensional space-time
tensor is constructed in combination with real-time seismic oscillation
dominant frequency offset; inputting the
tensor into a local adaptive graph convolutional network, dynamically adjusting the weight of a
convolution kernel through a space-time attention mechanism, and synchronously outputting a prediction result; triggering
incremental learning based on MEMS acceleration and FBG strain data, correcting a predicted value of an unliquefied region, and feeding back the predicted value to the
voxel model; and finally, generating a probabilistic disaster chain scene tree through spatial
pattern matching. The technical
bottleneck that a traditional model is poor in adaptability in a complex sedimentary environment and cannot be updated in real time is solved, and intelligent early warning support is provided for
engineering safety of an earthquake active area.