Adaptive Weighting in Geophysical Joint Inversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current joint inversion methods for geophysical data often get trapped in local minima due to the use of constant weight values, leading to inefficient convergence and incomplete models, especially when dealing with multiple independent data sets.
Innovation Solution
An adaptive weighting scheme is introduced, where weights are dynamically adjusted during the inversion process to emphasize different data types at various stages, allowing the algorithm to escape local minima and converge to a global minimum by modifying the objective function based on convergence criteria.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If constant weight values are used in joint inversion, then the inversion process is simpler to implement, but the algorithm gets trapped in local minima and convergence is incomplete
Solution Approach 1:
The patent applies dynamics by transforming the static constant weight values into dynamic adaptive weight values that automatically adjust during the inversion process. The weight for each data type is updated iteratively based on the relative reduction in misfit, allowing the algorithm to escape local minima and achieve better convergence while maintaining a relatively simple implementation framework.
Solution Approach 2:
The patent changes the parameter of weight values from fixed constants to adaptive variables that evolve during inversion. The weight adjustment mechanism modifies the contribution of each data type dynamically based on its misfit reduction performance, enabling the system to transition from simple constant weights to sophisticated adaptive weights without fundamentally changing the inversion algorithm structure.
2Productivity
If constant weight values are used for all data types, then the computational process is more efficient per iteration, but multiple different inversion runs are required to find an optimal solution
Solution Approach 1:
The patent ensures continuity of useful action by implementing continuous weight adjustment throughout the inversion process. Instead of performing multiple separate inversion runs with different constant weights, the algorithm continuously adapts the weights during a single continuous inversion process, maintaining computational efficiency while progressively improving the solution quality without interruption.
Solution Approach 2:
The patent applies preliminary action by establishing the adaptive weight adjustment mechanism before the inversion begins. The weight update rules are pre-defined based on misfit reduction principles, allowing the algorithm to automatically navigate toward the optimal solution without requiring multiple preliminary runs with different constant weight configurations.
3Adaptability or versatility
If arbitrary weight choices are made to balance data type contributions, then the inversion can proceed, but the solver moves through model space inefficiently with increased iterations required
Solution Approach 1:
The patent implements feedback by using the misfit reduction performance of each data type to inform the weight adjustment for the next iteration. The algorithm continuously monitors how much each data type contributes to reducing the overall misfit and automatically adjusts its weight accordingly, creating a closed-loop feedback system that optimizes the balance between data types dynamically rather than relying on arbitrary predetermined choices.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Method for adaptive weighting of geophysical data types in iterative joint inversion to speed convergence and aid escape from local minima of the penalty (objective) function. Two or more geophysical data sets (11) representing a region of interest are obtained, and are jointly inverted to infer models of the physical properties that affect the particular types of data used. The misfit for each data type is a weighted term in the penalty function (13). The invention involves changing the weights (51) as the iteration cycles progress when the iteration convergence criteria are satisfied (15), to see if they remain satisfied (52) with the modified penalty function.