Learning Model Design System for Bio-Inspired Analogies
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing design systems face challenges in generating diverse and creative solutions for complex design tasks due to limitations in data availability, manual curation costs, and the limited diversity of analogies generated by large language models.
Innovation Solution
A design system that utilizes a learning model with an analogy pipeline to identify and generate mechanisms from natural processes, leveraging a prompt transformer to construct a diverse mechanisms dataset and interactively adapt analogies for design tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If expert-curated training data is used to train LLMs, then solution quality for complex design tasks is improved, but data curation cost and scarcity increase
Solution Approach 1:
The system enables self-service by allowing the LLM to autonomously browse the web, search for natural analogs, and generate training data without requiring expert curators. The model performs its own data collection and curation tasks, transforming from a system that needs expert intervention to one that self-suffs its training needs.
Solution Approach 2:
The patent creates synthetic training data by copying and adapting information from web sources and natural analogs. Instead of relying on unique expert-curated data, the system generates multiple synthetic examples that replicate the quality and structure of expert annotations, thereby reducing dependency on scarce expert resources.
2Extent of automation
If LLMs generate analogies for bio-inspired designs, then solution generation is automated, but conceptual diversity is limited
Solution Approach 1:
The system implements a universal analogy generation framework that can draw from multiple domains (biology, physics, nature) and apply them across various design tasks. The LLM is enhanced with tools to search different knowledge sources and generate analogies from diverse natural phenomena, not just a fixed dataset, thereby achieving both automation and diversity.
Solution Approach 2:
The patent introduces a new dimension to analogy generation by enabling the LLM to access and process information from the web and natural domains beyond traditional training data. This expands the search space from static training corpora to dynamic, multi-source knowledge, resulting in more diverse and creative analogies.
3Reliability
If manual controls are used in natural-analogical systems, then data quality is maintained, but system scalability is reduced
Solution Approach 1:
The system replaces manual data quality control with self-service mechanisms where the LLM autonomously evaluates, selects, and curates training data from web sources. The model performs its own quality assessment and data selection tasks, eliminating the need for manual intervention while maintaining data quality standards and enabling scalable operation.
Solution Approach 2:
The patent substitutes the mechanical process of manual data curation with an automated LLM-based system. Instead of human operators manually reviewing and selecting data, the system uses computational methods and language models to automatically curate and quality-check training data, thereby achieving both high data quality and system scalability.
Data Source
AI summary
Systems, methods, and other embodiments described herein relate to identifying and generating mechanisms from natural processes by a learning model for accelerating design development. In one embodiment, a method includes identifying mechanisms for a design task using a prompt transformer with seeds from biological processes, and the prompt transformer forms a taxonomy tree using the mechanisms. The method also includes generating functional solutions that expand sparse branches of the taxonomy tree for the mechanisms using the prompt transformer. The method also includes clustering the mechanisms using text embedding for the design task. The method also includes inspecting the mechanisms with the prompt transformer to select a solution associated with the design task.