AI Embedding Vector Calibration for DUT Tuning Without Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing test and measurement systems require time-consuming retraining of neural networks to accommodate new device under test (DUT) models with different tuning parameters, limiting their flexibility and scalability.
Innovation Solution
Implement an AI embedding model within a vector database system, using pretrained neural networks and transformers to convert tensor images into vectors, allowing for efficient storage and retrieval of optimal tuning parameters without retraining, and iteratively updating the database with new DUT models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional neural network systems are used for DUT tuning, then accurate predictions can be made for trained DUT models, but time-consuming retraining is required for new DUT models with different tuning parameters
Solution Approach 1:
The system performs preliminary embedding of DUT characteristics and tuning parameters into a vector database during manufacturing. This pre-stored embedding enables rapid retrieval and prediction for new DUT models without requiring retraining, as the embedded representations capture the essential relationships between device characteristics and optimal tuning parameters.
Solution Approach 2:
Instead of retraining neural networks for each new DUT model, the system creates embedded vector representations (copies) of DUT characteristics and tuning relationships. These embedded copies are stored in the vector database and can be rapidly queried and matched, providing prediction capabilities without the computational overhead of retraining.
2Manufacturing precision
If traditional neural network systems are used for DUT tuning, then specific DUT models can be tuned accurately, but system flexibility and scalability to new DUT models are limited
Solution Approach 1:
The vector database system provides a universal framework that can store and retrieve embedded representations for multiple DUT models with different tuning parameters. The embedding process creates a common representation space that accommodates various device types, enabling the system to adapt to new DUT models without requiring model-specific training procedures.
Solution Approach 2:
The system transforms DUT characteristics and tuning parameters into embedded vector representations, changing the parameter space from raw device measurements to compressed embedding vectors. This transformation enables flexible querying and matching across different DUT models by operating in the embedding space rather than the original parameter space.
3Reliability
If traditional neural network systems are used for DUT tuning, then trained models provide reliable predictions, but extensive retraining is needed to maintain reliability for new DUT models
Solution Approach 1:
The system performs preliminary embedding of DUT characteristics and optimal tuning parameters during the manufacturing process. This pre-computed embedding captures the reliable relationships between device characteristics and tuning settings, storing them in the vector database for rapid retrieval. New DUT models can immediately leverage these pre-embedded representations without requiring time-consuming retraining to maintain prediction reliability.
Data Source
AI summary
A test and measurement system has a test and measurement instrument that includes a connection to a device under test (DUT); one or more analog-to-digital converters (ADCs) to receive and convert a signal from the DUT to one or more digital waveforms; and one or more processors to: receive the one or more digital waveforms corresponding to one set of tuning parameters applied to the DUT; build one or more image tensors of the one or more digital waveforms; use an artificial intelligence embedding model that generates one or more text strings from metadata and embeds the metadata and the one or more image tensors into a vector; access a vector database; receive a set of indexes having a number of indexes corresponding to a number of matches; use the set of indexes to find one or more sets of optimal tuning parameters; and validate operation of the DUT.