Failure Prediction Device Using Segmented Calculation Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing failure prediction technologies are unable to accurately predict failures in multiple components simultaneously and lack clarity on the basis of their predictions, leading to invalidity in assessment.
Innovation Solution
A failure prediction device and method that utilize multiple calculation models, each generated based on specific operation information related to distinct failure types, allowing for the input of operation data from devices to predict failure details with improved accuracy by identifying the most relevant factors contributing to each failure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single failure prediction model is used for multiple component types, then the device complexity is reduced, but the prediction accuracy for each specific failure type deteriorates
Solution Approach 1:
The patent divides the failure prediction system into multiple specialized models, each dedicated to predicting a specific type of failure (e.g., CPU failure, memory failure, storage failure). This segmentation allows each model to focus on learning the unique patterns and characteristics of its specific failure type, thereby improving prediction accuracy without requiring a single overly complex model to handle all failure types simultaneously.
Solution Approach 2:
Each failure prediction model is trained on operation information specifically related to its target failure type, giving it specialized knowledge tailored to local characteristics of that failure mode. This local quality approach ensures that each model develops expertise in detecting subtle patterns specific to its designated failure type, improving overall system accuracy.
2Measurement precision
If multiple specialized calculation models are used for different failure types, then the prediction accuracy for each failure type is improved, but the device complexity increases
Solution Approach 1:
The patent employs a universal data collection and processing framework that serves all specialized failure prediction models. The operation information acquisition unit, data preprocessing unit, and result aggregation unit function across all models, providing multi-functionality that reduces the overall complexity increase despite having multiple specialized models. This universal infrastructure allows the system to scale to multiple failure types without proportionally increasing complexity.
3Productivity
If operation information is analyzed without failure-specific filtering, then the data processing efficiency is maintained, but the ability to identify specific failure causes deteriorates
Solution Approach 1:
The patent segments the operation information into failure-type-specific datasets before feeding them to corresponding prediction models. This segmentation ensures that each model receives only the relevant information needed to predict its specific failure type, preventing information loss while maintaining processing efficiency through targeted data flow.
Solution Approach 2:
The patent introduces a data preprocessing unit that acts as an intermediary between raw operation information and the prediction models. This intermediary filters, transforms, and organizes the operation information according to the specific requirements of each failure type, ensuring that relevant failure cause information is preserved and presented in the appropriate format for each specialized model.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A failure prediction device (1) includes: a plurality of failure prediction model units (10) that output a failure prediction result according to a value input to calculation models generated for different failure details based on operation information on operation up to the occurrence of the respective failures from operation history of a predetermined device; and an operation information input unit (20) that inputs operation information acquired from a device subjected to failure prediction to the plurality of failure prediction model units (10).