Lifecycle Thread Clustering via Pre-computed Distance Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for finding and replicating lifecycle threads in complex projects like aircraft development are hindered by inconsistent task naming conventions and the computational intensity of large multi-dimensional matrices, making it labor-intensive and infeasible to interactively interrogate data.
Innovation Solution
A system and method that customizes clustering to collect necessary network elements for lifecycle threads by using a manifest with distance maps to reduce data storage and computational requirements, allowing for ordinal comparisons instead of square root and multiplication operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional clustering methods with standardized criteria are used, then the process is simple and automated, but the clusters cannot be customized to collect necessary network elements for lifecycle threads
Solution Approach 1:
The patent applies preliminary action by pre-defining distance maps that store computed distances between all unique values of selected attributes before the actual clustering process. This pre-computation allows the clustering algorithm to efficiently retrieve pre-calculated distances rather than computing them on-the-fly, enabling customized clustering criteria while reducing computational complexity during the interactive clustering phase.
2Ease of operation
If interactive handling of large datasets is performed using current approaches, then user control is maintained, but large multi-dimensional matrices are required making interactive interrogation infeasible
Solution Approach 1:
The patent extracts and stores only the essential distance information between unique attribute values in pre-computed distance maps, rather than maintaining complete multi-dimensional matrices of all task comparisons. This extraction of necessary information reduces data storage requirements while preserving the ability to perform interactive clustering operations on large datasets.
Solution Approach 2:
By pre-computing and storing distance maps before interactive clustering begins, the system prepares the necessary computational foundation in advance. This allows the interactive phase to proceed efficiently with reduced memory requirements, as the distance calculations are already available in optimized data structures rather than requiring on-demand computation during user interaction.
3Measurement precision
If distance computations involving multiple attributes are performed for each task comparison, then accurate similarity measurement is achieved, but computational intensity increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-computing distance values between all unique values of selected attributes and storing them in distance maps. During the clustering process, these pre-computed distances are retrieved and combined using weighted sums rather than performing full multi-attribute distance computations for each task pair. This significantly reduces computational power requirements while maintaining measurement precision through the use of pre-calculated accurate distance values.
Data Source
AI summary
A method is provided for processing data to establish lifecycle threads in the development of a structural product. The method includes defining a source lifecycle thread from process-related information for development of the structural product, and defining and matching a target lifecycle thread to the source lifecycle thread. The process, and source and target lifecycle threads are expressible as respectively a network and sub-networks of tasks described by a plurality of attributes. For defining either or each of the source lifecycle thread or target lifecycle thread, the method includes at least receiving user selection of an attribute of the plurality of attributes as user-selected criteria, performing a cluster analysis according to the user-selected criteria to produce a plurality of clusters of candidate tasks from the plurality of tasks, and selecting tasks from the plurality of clusters of candidate tasks for the source/target lifecycle thread.