Topology Optimization Design for Hybrid Uncertainties
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional reliability-based topology optimization methods fail to effectively handle multi-source uncertainties, leading to distorted uncertainty modeling and inconsistent optimal schemes due to the use of single-type mathematical models, and often require additional calculations and subjective factors, especially when dealing with multiple reliability-based constraints and gradient-based optimization frameworks.
Innovation Solution
A reliability-based topology optimization method is developed that considers bounded hybrid uncertainties by discretizing the design domain, decoupling probabilistic and interval uncertainties, and using a moving asymptote algorithm to determine worst working conditions and calculate reliability, thereby improving gradient analyticity and objective/constraint function gradients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional reliability-based topology optimization uses a single type of mathematical model to describe multi-source uncertainties, then the modeling process is simplified, but the uncertainty modeling becomes distorted and the optimal scheme fails
Solution Approach 1:
The patent segments the unified uncertainty modeling approach into distinct probabilistic and interval modeling components. By dividing the uncertainty representation into separate mathematical frameworks suited to different data types (sufficient samples vs. insufficient samples), the method maintains modeling accuracy while managing complexity through structured separation of uncertainty sources.
Solution Approach 2:
The patent employs a composite uncertainty modeling framework that combines probabilistic and interval mathematics, analogous to composite materials. This hybrid approach integrates two different mathematical paradigms to create a more robust and accurate uncertainty representation than either method could achieve alone, addressing the distortion problem while maintaining reasonable complexity.
2Device complexity
If the most probable point method is used to transform reliability constraints into deterministic constraints, then the optimization problem becomes simpler, but the most probable points from different constraints become inconsistent and require additional subjective trade-offs
Solution Approach 1:
The patent introduces a reliability indicator as an intermediary mechanism that mediates between multiple reliability constraints without requiring transformation to deterministic form. This intermediary allows direct handling of reliability-based constraints while maintaining consistency across different constraints, avoiding the subjective trade-off problem that arises from most probable point methods.
Solution Approach 2:
Instead of transforming reliability constraints into deterministic constraints through most probable point analysis, the patent inverts the approach by directly using reliability indicators within the optimization framework. This inversion maintains the probabilistic nature of constraints while achieving computational tractability, thereby preserving constraint consistency without subjective interventions.
3Measurement precision
If existing reliability expressions are constructed with additional condition discrimination, then the reliability analysis becomes more accurate, but the gradient calculation becomes difficult to apply in gradient-based topology optimization
Solution Approach 1:
The patent changes the parameter representation in reliability expressions by using reliability indicators that are directly differentiable with respect to design variables. This parameter transformation maintains the accuracy of reliability analysis while ensuring that the expressions possess the analyticity required for gradient-based optimization, thereby resolving the conflict between precision and ease of operation.
Data Source
AI summary
A reliability-based topology optimization design method for a part structure considering bounded hybrid uncertainties, includes the following steps: considering the uncertainties in the manufacture and service of the part structure, describing an external load with insufficient samples and a material property with sufficient samples as an interval variable and a bounded probabilistic variable respectively; discretizing a design domain of the part structure, setting the physical and geometric constraints, and establishing a reliability-based topology optimization design model; solving by a moving asymptote algorithm: decoupling the probabilistic and interval uncertainties, and determining the worst working condition by using gradients of constraint performance functions; defining a performance fluctuation under the worst working condition and calculating reliability of constraint performance; and finally, calculating the gradients of objective and constraint functions with respect to the design variables for iteration.