Transformer Concept Generation Network for Cross-Field Design Reasoning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to automatically generate conceptual schemes in product design due to the complexity and volume of cross-field design knowledge, making it difficult to reason, transfer, and reorganize design knowledge effectively.
Innovation Solution
A Design Concept Generation Network (DCGN) is constructed using a Transformer-based model with a word importance constraint mechanism, incorporating a Transformer encoder, decoder, importance constraint matrix, and cross-attention layer to adaptively learn and generate conceptual schemes from text data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual experience and design heuristics are used to reason and reorganize design knowledge, then design quality can be maintained, but design efficiency deteriorates due to the increasing difficulty of handling large amounts of design data
Solution Approach 1:
The patent replaces the mechanical manual reasoning and reorganization process with an automated neural network system (DCGN). The Transformer-based DCGN automatically reasons, transfers, and reorganizes design knowledge from unstructured text data, substituting human manual effort with an intelligent system that handles large-scale design data efficiently while maintaining design quality through learned patterns from training data
Solution Approach 2:
The DCGN system enables self-service by automatically generating conceptual schemes without requiring manual intervention. The system trains on design data and then autonomously performs knowledge reasoning, transfer, and reorganization tasks, serving the design process independently and significantly improving design efficiency while maintaining consistent quality standards
2Adaptability or versatility
If design knowledge from diverse fields is extensively applied to expand design space, then innovation potential is improved, but knowledge complexity increases making it difficult to reason and reorganize
Solution Approach 1:
The DCGN system provides a universal framework that handles multiple types of design knowledge (functions, structures, scientific effects, cases) from diverse fields through a single unified neural network architecture. The Transformer-based model processes various knowledge types uniformly, enabling the system to adapt to different design domains and expand design space without requiring separate specialized systems for each knowledge type
Solution Approach 2:
The patent transforms unstructured design knowledge text into structured numerical representations through embedding layers and attention mechanisms. By converting textual knowledge into numerical vectors and processing them through learned parameters, the system manages knowledge complexity effectively while maintaining the ability to reason and reorganize cross-field design knowledge for innovative conceptual schemes
3Ease of manufacture
If deep learning models generate design concepts in the forms of images or spatial shapes, then visual design exploration is improved, but the concepts become too abstract or too detailed to be suitable for conceptual scheme design
Solution Approach 1:
The patent applies local quality by generating design concepts at the appropriate level of abstraction for conceptual scheme design. Rather than producing overly abstract images or overly detailed spatial shapes, the DCGN generates text-based conceptual schemes that contain the right amount of specific information and detail, making them directly suitable for early-stage design exploration and further development
Data Source
AI summary
A method for constructing a design concept generation network (DCGN) and a method for automatically generating a conceptual scheme are provided. A DCGN includes a Transformer encoder, a Transformer decoder, an importance constraint matrix generation module, an importance constraint embedding layer, a cross-attention (CA) layer, and an optimization module. A word importance constraint is ingeniously introduced based on an attention mechanism of a Transformer to record input word constraint information contained in a generated text sequence. This can effectively ensure the reliability and effectiveness of a generated conceptual scheme and is conducive to capturing potential semantic importance information and implementing semantic knowledge reasoning.