Hybrid Module Generation Algorithm for Product Architecture Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current clustering algorithms for product architecture, such as Genetic Algorithms, face challenges with execution time and inability to account for complex component relationships, leading to inefficient module generation and increased computational resources.
Innovation Solution
A hybrid module generation algorithm that utilizes a square design structure matrix and a domain mapping matrix to group elements into clusters, incorporating the Reangularity metric and stochastic behavior to balance coupling and similarity-based views, thereby optimizing module formation and reducing computational time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If Genetic Algorithms are used for clustering in product architecture, then module generation can be performed, but execution time becomes excessively long and computational resources increase
Solution Approach 1:
The patent segments the clustering problem into two distinct matrix representations: a square design structure matrix for coupling-based clustering and a rectangular domain mapping matrix for similarity-based clustering. This segmentation allows each matrix type to be processed independently with optimized algorithms, avoiding the computational burden of trying to handle all clustering requirements simultaneously with a single complex algorithm like Genetic Algorithms.
Solution Approach 2:
The patent introduces a new dimension by using two different matrix formats (square and rectangular) rather than relying on a single square matrix approach. This dimensional change enables the system to capture both coupling relationships and similarity relationships simultaneously, allowing for more efficient clustering without requiring iterative optimization methods that consume excessive computational time.
2Adaptability or versatility
If traditional clustering algorithms are used, then module formation can be achieved, but the algorithms cannot account for complex component relationships including both coupling and similarity
Solution Approach 1:
The patent merges two previously separate clustering approaches (coupling-based and similarity-based) into a unified system that processes both types of relationships simultaneously. By integrating both matrix types and their respective clustering algorithms, the system achieves comprehensive adaptability to handle complex component relationships while maintaining computational efficiency through specialized processing for each matrix type.
Solution Approach 2:
The patent creates a universal clustering framework that can handle multiple types of component relationships (both coupling and similarity) through a multi-functional system. The square design structure matrix handles coupling relationships while the rectangular domain mapping matrix handles similarity relationships, allowing the system to be universally applicable to various product architecture scenarios without sacrificing efficiency.
3Device complexity
If clustering algorithms assume any component can form a cluster with any other component, then computational simplicity is maintained, but real-world constraints on component relationships are ignored
Solution Approach 1:
The patent applies local quality by allowing different clustering constraints to apply to different components based on their specific relationships. The square design structure matrix captures local coupling constraints for components that must work together, while the rectangular domain mapping matrix captures local similarity constraints for components that should be grouped together. This localized approach to constraints maintains algorithmic clarity while significantly improving the reliability of module formation by respecting real-world component relationships.
Data Source
AI summary
A computer program product and a computer-implemented method are disclosed. The method includes providing a square design structure matrix having a number of rows and a number of columns equal to the number of rows, wherein each row represents one of a plurality of elements, wherein each entry of the design structure matrix is a numerical quantity; providing a domain mapping matrix having a number of rows equal to the number of rows of the design structure matrix, each row having an identical representation of one of the plurality of elements as a corresponding row of the design structure matrix, the domain mapping matrix having any number of columns, wherein each entry of the domain mapping matrix is a numerical quantity; and using a computer processor to group one or more elements into a cluster arrangement, using a plurality of entries from each of the matrices.