Machine Learning Disk Inspection for Storage Quality Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing capacity and density of hard disk drive storage devices lead to prolonged quality inspection times, resulting in inefficiencies and increased costs due to the need for thorough, time-consuming checks of each disk area for defects before deployment.
Innovation Solution
Implementing machine learning methods, such as convolutional neural networks, to perform partial scans of disks, allowing for the prediction and classification of defects with reduced scanning time, potentially scanning only 10-20% of the disk surface, thereby reducing inspection time by up to 90%.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional quality inspection methods are used to check each area on every disk, then defect detection accuracy is improved, but inspection time increases significantly
Solution Approach 1:
The inspection process is segmented into two phases: a rapid partial inspection covering only critical areas (spindle hole, landing zones, data zones) followed by a comprehensive full inspection. This segmentation allows the system to achieve quick initial defect detection while maintaining the option for thorough verification, thus resolving the contradiction between speed and accuracy.
Solution Approach 2:
The system performs partial inspection of only the most critical disk areas (spindle hole, landing zones, and data zones) rather than inspecting the entire disk surface. This partial action approach identifies the majority of defects quickly, reducing inspection time while maintaining sufficient detection accuracy for deployment decisions.
2Quantity of substance
If storage drive capacities are increased with more disks and higher data density, then storage capacity is improved, but inspection time increases
Solution Approach 1:
The inspection system applies local quality by differentiating inspection intensity across different disk regions. Critical areas such as the spindle hole, landing zones, and data zones receive focused inspection attention, while less critical areas are inspected more rapidly or selectively. This allows high-capacity drives with multiple disks to be inspected efficiently without requiring uniform thorough inspection of every surface.
Solution Approach 2:
The inspection process is divided into rapid partial inspection of critical zones followed by selective full inspection only when needed. This segmentation enables the system to handle increased storage capacity with proportionally less inspection time, as the majority of defects are detected in the initial rapid phase.
3Reliability
If traditional thorough inspection methods are used, then defect detection completeness is improved, but manufacturing efficiency decreases
Solution Approach 1:
The system performs partial inspection of critical areas sufficient to detect the majority of defects, accepting that not every single defect will be caught in the rapid phase. This partial action approach maintains manufacturing efficiency while achieving adequate defect detection completeness for deployment, with full inspection available as a follow-up if needed.
Solution Approach 2:
The inspection system uses feedback mechanisms where results from the rapid partial inspection inform whether a subsequent full inspection is necessary. This feedback loop ensures that defect detection completeness is maintained by triggering comprehensive inspection only when the rapid phase identifies potential issues or when deployment requirements demand higher certainty.
Data Source
AI summary
Methods are provided for utilizing machine learning operations configured for use in processing missing pieces of visual data in image data to predict potential location of defects and/or damage in storage device disks. These predictions can allow for a sufficient ability to categorize disks during storage device quality inspections. This can allow for quality inspections to conclude before all areas of the disk surface are scanned. Because less surface area of the disks within the storage device require scanning, the time required for quality inspection scanning prior to deployment can be greatly reduced. Additionally, the partial scans occurring prior to deployment may be supplemented or updated after deployment through the performance of a dense scan. These secondary scans can be configured to scan all previously unscanned areas during storage device downtimes or in response to an environmental trigger such that the storage device will not exhibit any loss in performance.