Topology Optimization Shape Transformation for Binary Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for handling binary variables and equality constraints in topology optimization are inflexible and complicated, particularly when using gradient-based linear problem solvers, as they require conversion to real value variables with inequality constraints, making it difficult to enforce constraints like the sum of binary variables being 0 or 1.
Innovation Solution
A method involving the transformation of a K-dimensional unit cube into a transformed shape using a shape function, which shifts undesirable vertices, ensuring that the design search space is reduced and the sum of binary variables is either 0 or 1, by iteratively executing a topology optimization algorithm and projecting values between the normal and transformed coordinate systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If binary variables with equality constraints are used in topology optimization, then the design search space is well-defined and combinatorial choices are clear, but gradient-based linear problem solvers cannot handle them directly
Solution Approach 1:
The patent introduces a shape function as an intermediary transformation that maps the K-dimensional unit cube (containing binary variables with equality constraints) into a transformed search space. This mediator allows gradient-based solvers to operate in the transformed space while the shape function ensures that when projected back, the binary variable constraints are satisfied. The shape function acts as a bridge between the discrete combinatorial problem and the continuous gradient-based optimization.
2Ease of operation
If binary variables are converted to real value variables with inequality constraints, then gradient-based solvers can handle the problem, but the method becomes inflexible and complicated to implement
Solution Approach 1:
The patent transforms the problem parameters by applying a shape function that changes the coordinate system from the normal K-dimensional unit cube to a transformed shape. This parameter transformation allows the use of real value variables in the transformed space while maintaining the equivalence to the original binary variable problem. The transformation modifies the search space geometry rather than the fundamental nature of the variables, simplifying implementation while preserving constraint satisfaction.
3Adaptability or versatility
If the full K-dimensional unit cube is used as the design search space, then all possible combinations are available, but undesired combinations that do not satisfy constraints are included
Solution Approach 1:
The patent extracts or removes undesired combinations from the design search space by applying a shape function that transforms the K-dimensional unit cube into a modified shape where vertices representing invalid combinations (those not satisfying the equality constraint) are eliminated or shifted. This extraction process maintains the desired combinations while removing the undesired ones, ensuring constraint satisfaction without sacrificing design flexibility among valid options.
Data Source
AI summary
Methods for combinatorial choice in topology optimization are disclosed. In one embodiment, a method of performing combinatorial choice in a non-linear programming problem for designing a structure includes establishing, by one or more processors, a set of K variables, wherein the set of K variables defines an K-dimensional unit cube in a normal coordinate system. The K-dimensional unit cube includes a plurality of desired vertices representing a plurality of desired combinations that satisfy a constraint and a plurality of undesired vertices representing a plurality of undesired combinations that do not satisfy the constraint. The method further includes transforming the K-dimensional unit cube into a transformed shape in a transformed coordinate system using a shape function. Values of the set of K variables cannot have a combination represented by the plurality of undesired combinations.