Intelligent first arrival screening chromatographic inversion method and system based on deep learning

Abstract

The invention provides an intelligent first arrival screening chromatographic inversion method and system based on deep learning. The method comprises that original earthquake data is obtained; an earthquake data training set and an earthquake data to-be-measured set are extracted from the originality earthquake data; an initial model of deep learning is trained based on the earthquake data training set, a 2D time-space image form is used to process multichannel single-shot records in the earthquake data training set in the image aspect, the initial model of deep learning is trained to learn space information among the multiple channels, and a first arrival pickup model is obtained; first arrival is picked up from the earthquake data to-be-measured set according to the first arrival pickupmodel; and a first arrival pickup result is used to carry out chromatographic inversion to obtain a chromatographic inversion result. Via the method and system, high-quality first arrival informationcan be extracted effectively, a shallow-layer high-speed abnormal body can be effectively represented by chromatographic inversion based on the high-quality first arrival information, and a higher-quality of underlying stratum imaging result of the high-speed abnormal body can be obtained.

Description

technical field [0001] The present application relates to the technical field of tomographic inversion in seismic exploration, in particular to a tomographic inversion method and system for intelligently screening first arrivals based on deep learning. Background technique [0002] At present, the tomographic inversion technology is becoming more and more mature. This technology obtains a more accurate shallow surface velocity model through multiple iterations, and finally obtains an ideal inversion imaging result. In igneous rock development areas, obtaining a relatively accurate igneous rock velocity model is very important for inversion imaging; if there is no more accurate velocity model, the final imaging results based on the velocity model will be far from the real results. [0003] In the prior art, someone proposes a tomographic inversion method using an artificial neural network to pick up first arrivals. However, it is necessary to calculate the original shot set ...

AI Technical Summary

Problems solved by technology

However, this method based on the subjective factors of the processor will ignore other unknown information related to the first arrival in the shot set data while emphasizing some attributes, which will affect the picking accuracy of the first arrival

In addition, this kind of method will consume a lot of computing cost and affect the collection efficiency because of the large amount of data attribute calculation work before the original collected shot data is input into the network.

Images