Non-linear Beamforming Tuple Solver with Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The computational expense of solving non-linear beamforming (NLBF) sub-problems in seismic data processing is high due to the need for optimizing a large number of move-out function parameters across millions of spatial and temporal sub-regions using non-linear optimizers like genetic algorithms.
Innovation Solution
A method that involves discretizing seismic data into NLBF sub-problems, solving a subset with a non-linear optimizer, and periodically training a machine-learned model to generate intermediate NLBF tuple predictions, which can be used as initial values or accepted as final NLBF tuples, thereby reducing computational burden.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If non-linear optimizers like genetic algorithms are used to solve NLBF sub-problems, then the accuracy of move-out function parameters is improved, but the computational cost and time increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-training a machine learning model on a subset of NLBF sub-problems before processing the full dataset. This pre-computed knowledge is then reused during online processing, eliminating the need to solve every sub-problem from scratch and significantly reducing computational time while maintaining parameter accuracy.
Solution Approach 2:
The patent uses copying by training a machine learning model to replicate the optimization process. The model learns from solved sub-problems and generates predictions for unsolved ones, creating a computational copy of the optimization knowledge that can be rapidly applied without re-running the full non-linear optimization algorithm.
2Reliability
If non-linear optimizers like genetic algorithms are used to solve all NLBF sub-problems, then the quality of NLBF tuples is improved, but the computational expense becomes prohibitive
Solution Approach 1:
The system performs preliminary training on a representative subset of data to build a machine learning model that captures the essential patterns. This pre-computation phase invests computational resources once, enabling rapid online processing with significantly reduced computational expense while maintaining tuple quality through the model's learned predictions.
Solution Approach 2:
The patent changes the computational approach from directly solving optimization problems to using machine learning predictions. By transforming the problem from numerical optimization to pattern recognition, the system achieves comparable quality results with substantially lower computational expense during online operation.
3Measurement precision
If machine learning model training is performed frequently, then the accuracy of intermediate NLBF tuple predictions is improved, but the processing time for training increases
Solution Approach 1:
The patent implements periodic action by training the machine learning model at scheduled intervals rather than continuously. The model is trained on accumulated data periodically, balancing the need for accurate predictions with the cost of training time. This periodic retraining maintains prediction accuracy while avoiding the overhead of continuous training.
Solution Approach 2:
The system maintains continuity of useful action by using the trained model for online predictions between training periods. The model continuously processes incoming sub-problems with high speed, and training occurs periodically in the background, ensuring that useful processing continues without interruption while periodically refreshing the model's accuracy.
Data Source
AI summary
A method for determining a non-linear beamforming (NLBF) tuple is disclosed. The method includes receiving a seismic data set and discretizing the seismic data set into a plurality of NLBF sub-problems. The method includes solving a subset of the NLBF sub-problems with a non-linear optimizer to create final NLBF tuples. The method further includes periodically training a machine-learned model with a subset of the NLBF sub-problems and final NLBF tuples data and obtaining intermediate NLBF tuple predictions from the trained machine-learned model. The intermediate NLBF tuple predictions may be used as initial values in a non-linear optimizer to create final NLBF tuples or may be accepted as final NLBF tuples. The method includes storing the final NLBF tuples.