Time-Series Diagnosis Model Updating for Fast Adaptive Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing time-series data diagnosis systems face challenges in achieving high accuracy and efficient updating of learned models due to computational constraints and the need for speedy processing during operation, particularly in handling changes in device conditions over time.
Innovation Solution
A time-series data diagnosis device with storage, difference-degree calculation, diagnosis-result-information generation, evaluation information acquisition, and updating means, which calculates and updates learned-sample subsequence data based on diagnosis results and evaluation feedback to maintain high accuracy with reduced computational load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complete accuracy is achieved in generating sample subsequence data from scratch, then diagnostic accuracy is improved, but time consumption and computational resources increase significantly
Solution Approach 1:
The system performs preliminary learning offline to generate initial sample subsequence data before actual operation. This pre-computation stores learned patterns in advance, allowing rapid online diagnosis without repeating the full learning process during time-critical operations.
Solution Approach 2:
Instead of performing complete re-learning during operation, the system performs partial updates by selectively adding new sample subsequence data when diagnostic accuracy deteriorates. This partial action maintains adequate accuracy while minimizing computational overhead during operation.
2Adaptability or versatility
If the learned model is updated frequently to adapt to changing operating conditions, then adaptability is improved, but computational load increases
Solution Approach 1:
The system monitors diagnostic accuracy as feedback and compares it against thresholds. Updates to the learned model are triggered only when accuracy deteriorates below acceptable levels, creating a feedback-controlled update mechanism that balances adaptability with computational efficiency.
Solution Approach 2:
The system dynamically adjusts the update frequency and scope of the learned model based on actual performance needs. Rather than fixed periodic updates, the model evolves adaptively in response to changing operating conditions and diagnostic accuracy requirements.
3Reliability
If traditional cluster-based updating methods are used to incorporate new time-series data, then learning completeness is improved, but processing speed decreases due to large calculation amounts
Solution Approach 1:
The system extracts only the essential new information from incoming time-series data that is not already represented in the existing sample subsequence data. By taking out only the novel patterns rather than processing all data through complete clustering, the system maintains learning completeness while reducing computational load.
Solution Approach 2:
The system changes the update parameter from complete cluster re-computation to selective addition of new sample subsequence data. This parameter change in the updating strategy maintains the essential learning function while dramatically reducing the computational complexity from O(n²) clustering to linear-time data addition.
Data Source
AI summary
The time-series data diagnosis device includes a storage, a difference-degree calculator, a diagnosis-result-information generator, an evaluation information acquirer, and an updater. The difference-degree calculator calculates a difference degree between diagnosis target time-series data and a learned-sample subsequence data item stored in the storage. The diagnosis-result-information generator generates, based on the difference degree calculated by the difference-degree calculator, diagnosis result information about the diagnosis target time-series data. The evaluation information acquirer acquires evaluation information indicating evaluation as to whether a diagnosis result indicated by the diagnosis result information is valid. The updater, when the evaluation information acquired by the evaluation information acquirer indicates evaluation that the diagnosis result is not valid, updates, based on the diagnosis target time-series data corresponding to the diagnosis result information, the learned-sample subsequence data item stored in the storage.