Industrial Component Similarity Search with Feedback-Based Interchangeability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for searching and replacing industrial components lack adaptability and flexibility, as they rely on theoretical models and do not effectively incorporate user feedback, limiting their ability to handle unforeseen situations and providing little to no impact on subsequent applications.
Innovation Solution
A computer-implemented method that uses similarity embeddings and user feedback to create interchangeability classes, combining legacy replaceability data with empirically induced clusterization, allowing for the generalization of replaceability to interchangeability through the use of neural networks and vectorization techniques like autoencoders and Bidirectional Encoder Representations from Transformers (BERT).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If theoretical models are used for component similarity search, then the search method is simple and implementable, but the adaptability to unforeseen situations is poor and flexibility is limited
Solution Approach 1:
The patent transforms the static theoretical model into a dynamic system by continuously updating similarity embeddings with user feedback. The embedding model learns and adapts its parameters based on interaction data, allowing the system to evolve from a fixed theoretical framework to a flexible, adaptive system that improves over time while maintaining implementability.
Solution Approach 2:
The patent implements a feedback loop where user interactions with search results are captured and used to retrain the embedding model. This feedback mechanism allows the system to learn from actual usage patterns and improve its similarity assessments, resolving the contradiction between simple implementation and adaptability by incorporating learning from feedback.
2Adaptability or versatility
If user feedback is incorporated as a polishing step, then user feedback is considered, but the feedback does not affect the search itself or subsequent clustering
Solution Approach 1:
The patent merges the feedback mechanism with the core search functionality by integrating user feedback directly into the embedding computation process. Instead of treating feedback as a separate polishing step, the system combines feedback signals with the original search query to generate updated embeddings, ensuring that feedback affects both the current search results and future clustering operations.
Solution Approach 2:
The patent performs preliminary action by pre-computing similarity embeddings that incorporate feedback signals before the actual search execution. This allows the search system to use pre-prepared, feedback-informed embeddings, ensuring that user feedback is already integrated into the search foundation rather than being applied as an afterthought.
3Measurement precision
If strict replaceability criteria are used, then the search results are precise and reliable, but the system lacks flexibility and cannot handle fuzzy contexts
Solution Approach 1:
The patent changes the parameter representation from discrete, binary replaceability criteria to continuous similarity scores derived from embedding vectors. This allows the system to maintain precision through vector similarity measurement while gaining flexibility by allowing gradual transitions and fuzzy matching, enabling components to be partially similar rather than strictly identical or completely different.
Data Source
Figure 1~3
Figure 4~6
Figure 7~8
AI summary
A computer implemented method for improving a similarity search of an industrial component model (12) comprises: • providing a set of industrial component models, each having associated attributes and a similarity embedding (10) being an embedding of at least some of said associated attributes; • receiving a similarity request using a given industrial component model as an input, the output of said similarity request being a first subset of industrial component models selected from the set of industrial component models based on the comparison between their similarity embeddings (10) and the similarity embedding of the input industrial component model; • receiving a second subset of industrial component models selected by a user from said first subset of industrial component models based on an interchangeability criteria of the input industrial component model with any industrial component model of said second subset of industrial component models; • associating a similarity attribute to the input industrial component model (ICM), said attribute being a unique ID defined by said similarity request and/or said second subset of industrial component models (ICMS2); and • computing a new set of similarity embeddings (10).