2D Trace Data Reshaping for CNN Fault Detection in Panel Manufacturing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting faulty conditions in display panel manufacturing processes using trace data are inefficient due to raw trace data not being in a suitable format for machine learning models, leading to inadequate pattern recognition and classification of faults.
Innovation Solution
Converting one-dimensional time-series sensor data into a two-dimensional data array to enhance the performance of machine learning models, specifically convolutional neural networks, by preprocessing the data through techniques such as mapping, merging, and rearranging, allowing for improved fault detection and prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If raw trace data is used directly as input to machine learning models, then the data processing is simple, but the pattern recognition accuracy and fault detection reliability are insufficient
Solution Approach 1:
The patent transforms one-dimensional time-series sensor data into two-dimensional data arrays by introducing a spatial dimension through techniques such as sliding window segmentation, multi-sensor fusion, and temporal-spatial mapping. This dimensional transformation enables convolutional neural networks to effectively capture both temporal patterns and spatial correlations in manufacturing trace data, significantly improving fault detection accuracy while providing a systematic preprocessing framework.
2Reliability
If complex data preprocessing is performed to improve pattern recognition, then the fault detection accuracy improves, but the processing time and computational resources increase
Solution Approach 1:
The patent implements preliminary data preprocessing operations including normalization, denoising, and feature extraction before model training. By preparing the data in advance and organizing it into standardized 2D array formats, the system reduces computational burden during real-time inference, thereby improving fault detection reliability without excessive processing delays during production.
Solution Approach 2:
The patent segments continuous manufacturing trace data into discrete time windows or batches, processing them in manageable units. This segmentation approach enables efficient batch processing and parallel computation, reducing overall processing time while maintaining high detection reliability through comprehensive analysis of each segment.
3Adaptability or versatility
If traditional classification methods are used for fault detection, then the implementation is straightforward, but the ability to detect complex patterns and correlations is limited
Solution Approach 1:
The patent replaces traditional mechanical classification algorithms with data-driven deep learning models, specifically convolutional neural networks. This substitution enables the system to automatically learn complex patterns and correlations in manufacturing data without manual feature engineering, significantly enhancing pattern recognition capability and adaptability to various fault types while the standardized preprocessing framework manages model complexity.
Data Source
AI summary
A method for detecting a fault includes: receiving a plurality of time-series sensor data obtained in one or more manufacturing processes of an electronic device; arranging the plurality of time-series sensor data in a two-dimensional (2D) data array; providing the 2D data array to a convolutional neural network model; identifying a pattern in the 2D data array that correlates to a fault condition using the convolutional neural network model; providing a fault indicator of the fault condition in the one or more manufacturing processes of the electronic device; and determining that the electronic device includes a fault based on the fault indicator. The 2D data array has a dimension of an input data to the convolutional neural network model.