Reverse Time Migration Wavefield Storage via Co-Processor Offloading
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for reverse time migration of seismic wavefield data face significant computational challenges due to the need for simultaneous access to forward and backward propagated wavefields, leading to inefficiencies in data storage and retrieval, which limits the processing speed and increases computational burden.
Innovation Solution
The system utilizes co-processor resources to offload core computational work, allowing for efficient data compression and decompression, and implements a parallel programming model with asynchronous communication to optimize the Reverse Time Migration algorithm, reducing the need for repeated computations and improving data transfer efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If forward and backward propagated wavefields are stored and retrieved simultaneously at each time step, then image quality is improved, but computational burden and storage requirements increase significantly
Solution Approach 1:
The patent extracts only the essential wavefield data needed for imaging by implementing selective storage of wavefield snapshots at specific time steps rather than storing all intermediate data. This extraction approach reduces storage requirements while maintaining the quality of subsurface images by preserving critical information needed for the imaging condition.
Solution Approach 2:
The patent performs preliminary forward propagation of the wavefield to compute and store wavefield snapshots at predetermined time steps before the actual imaging process. This preliminary action allows the system to have pre-computed wavefield data ready for correlation during reverse-time migration, eliminating the need to re-compute or store all intermediate wavefields during the imaging phase.
2Reliability
If forward wavefield is propagated repeatedly to the n-th time step, then data availability is improved, but computational time increases
Solution Approach 1:
The patent performs preliminary forward propagation of the wavefield to compute and store wavefield snapshots at predetermined time steps before the actual imaging process. This preliminary action allows the system to have pre-computed wavefield data ready for correlation during reverse-time migration, eliminating the need to re-compute or store all intermediate wavefields during the imaging phase.
Solution Approach 2:
The patent creates copies of wavefield data at specific time steps during forward propagation and stores them for later retrieval. Instead of repeatedly propagating the wavefield to the n-th time step, the system uses these pre-created copies during the imaging process, significantly reducing computational time while maintaining data availability for correlation.
3Quantity of substance
If data compression techniques are applied to reduce storage burden, then storage requirements are reduced, but additional computational burden is imposed
Solution Approach 1:
The patent applies partial data compression by storing wavefield snapshots at selectively chosen time steps rather than compressing all intermediate data. This partial action approach reduces storage requirements significantly while imposing minimal additional computational burden, as compression is only applied at discrete time points rather than continuously throughout the propagation process.
4Power
If CPU resources are used for compression and decompression, then co-processor resources are freed for core computation, but overall processing speed may be reduced
Solution Approach 1:
The patent segments the computational workflow into distinct phases: forward propagation phase where wavefield snapshots are computed and stored, and imaging phase where these snapshots are retrieved and correlated with reverse-time wavefields. By separating compression operations from core imaging computations, the system optimizes resource utilization without significantly impacting overall processing speed.
Solution Approach 2:
The patent performs data compression during the forward propagation phase as a preliminary action, so that when the imaging phase begins, the data is already compressed and ready for efficient retrieval and correlation. This timing strategy ensures that CPU resources are available during the imaging phase for the computationally intensive correlation operations, while compression occurs when co-processor resources are temporarily less utilized.
Data Source
AI summary
A system and method for generating images of a subsurface region of interest is provided. In one embodiment, a computer-implemented method of generating images related to a subsurface region of interest includes: accessing, via a central processing unit (CPU), seismic data and an earth model related to the subsurface region of interest; forward propagating a source wavefield using the earth model at a first time interval via at least one external co-processor coupled to the CPU; transferring, at a second time interval, the forward propagated source wavefield to the CPU for compression and external storage; backward propagating the seismic data at the first time interval via the external co-processor to derive backward propagated receiver wavefield; and transferring, at the second time interval, the backward propagated receiver wavefield to the CPU. Via the CPU, the method further includes retrieving the stored forward propagated source wavefield; decompressing the retrieved forward propagated source wavefield; and applying imaging conditions to the decompressed forward propagated source wavefield and backward propagated receiver wavefield to construct image data representative of the subsurface region of interest.