Joint data and physics model driven full-waveform inversion using CMP gathers and well-logging data

We propose a joint data- and physics-model-driven fullwaveform inversion (FWI) method based on semisupervised learning framework, which uses well-logging data, pseudo labels produced from conventional FWI and common mid-point (CMP) gathers to train neural network. Neural network builds mapping relationship between several adjacent CMP gathers and the vertical profiles of the subsurface model. The conventional model-driven FWI is used to produce pseudo label datasets to train neural network, which can reduce the reliance on massive datasets. At the same time, FWI is also a physical model constraint on the neural network to make the inversion result more physical interpretable. Using our method, the neural network is able to utilize information from both logging data and physical models, thus it can take advantage of well-logging data to make the inversion more accurate and stable. Synthetic tests on the Marmousi2 model and Overthrust model show that compared with the conventional model-driven FWI, our method can get more accurate and stable inversion result in the situation of lacking low-frequency data and bad initial model.