Local Fault Processing Using Knowledge Distillation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning models designed for fault processing require significant computing resources, making them unsuitable for local deployment, and while knowledge distillation can create smaller models for local use, their effectiveness is compromised compared to cloud models.
Innovation Solution
Deploying a first deep learning model on a local storage device to determine if faults can be solved locally, using a second model for local processing if possible, and sending complex issues to a cloud-based third deep learning model for resolution, all obtained through knowledge distillation from a larger cloud model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a deep learning model is deployed on a local storage device, then the response speed is improved, but the computing resources required increase significantly
Solution Approach 1:
The patent segments the fault processing system into multiple components: a local shallow neural network for quick environmental factor classification, a local knowledge base for common fault resolution, and a cloud-based deep learning model for complex fault analysis. This segmentation allows the system to achieve fast local response while offloading heavy computing resources to the cloud when needed.
Solution Approach 2:
The patent implements partial action by using a simplified shallow neural network locally that handles only environmental factor classification and common fault processing, rather than deploying the complete deep learning model. This partial deployment reduces local computing resource consumption while maintaining fast response for common scenarios.
2Use of energy by moving object
If knowledge distillation is performed on deep learning models for local deployment, then computing resources are reduced, but the model effectiveness deteriorates
Solution Approach 1:
Instead of attempting to distill the entire deep learning model into a single small model, the patent segments the functionality into a shallow neural network for environmental classification and a knowledge base for fault processing. This segmentation preserves the effectiveness of each component for its specific purpose while reducing overall computing resource requirements.
Solution Approach 2:
The patent introduces a knowledge base as an intermediary between the shallow neural network and the cloud-based deep learning model. The knowledge base stores pre-processed fault information and processing rules, enabling the local system to effectively handle common faults without requiring the full deep learning model, thus maintaining effectiveness while reducing resource consumption.
3Use of energy by moving object
If a shallow neural network is used for local fault processing, then computing resources are reduced, but the accuracy of fault diagnosis decreases
Solution Approach 1:
The patent segments the fault processing into two stages: environmental factor classification using a shallow neural network, and detailed fault analysis using a knowledge base and cloud-based deep learning model. This segmentation allows the shallow network to efficiently handle initial classification while the knowledge base provides accurate diagnosis for specific fault types.
Solution Approach 2:
The patent performs preliminary action by using the shallow neural network to classify environmental factors and identify potential fault types before detailed diagnosis. This preliminary classification narrows down the search space, allowing the knowledge base to quickly retrieve accurate diagnostic information without requiring the shallow network to perform complex analysis.
Data Source
AI summary
A method in an illustrative embodiment includes determining, by a first deep learning model on a storage device, whether a fault is caused by an environmental factor. The method further comprises: determining, by a second deep learning model on the storage device in response to determining that the fault is caused by the environmental factor, whether the fault can be solved locally in the storage device. The method further comprises: processing the fault according to a knowledge base in response to determining that the fault can be solved locally in the storage device. The method further comprises: sending the fault to a third deep learning model on a cloud device in response to determining that the fault cannot be solved locally in the storage device, the first deep learning model and the second deep learning model being obtained by model distillation of the third deep learning model.