Batch Sequencing by Yield Data for Root Cause Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying the root cause of yield loss in manufacturing processes is challenging, especially in early stages, due to weak correlations between yield loss and specific steps or tools, requiring analysis of large data sets and multiple batches, which is time-consuming and costly.
Innovation Solution
Sequencing batches by output data instead of chronological order to correlate specific detrimental factors, allowing for the identification of modality in output data linked to specific tool subsets, thereby reducing the need for extensive testing and improving correlation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If chronological sequencing of batches is used to identify yield loss causes, then the manufacturing process maintains standard operational procedures, but the correlation accuracy between yield loss and specific tools/steps remains weak
Solution Approach 1:
The patent inverts the traditional chronological sequencing approach by sorting batches based on output data (yield values) in descending order. This inversion allows batches with similar yield characteristics to be grouped together, making it easier to identify patterns and correlations between specific tools/steps and yield loss. The system sequences batches by their output performance rather than by manufacturing time, enabling faster detection of detrimental factors.
Solution Approach 2:
The patent changes the sorting parameter from chronological time to output data (yield value). By using yield-based sequencing, the system transforms how batches are organized and analyzed, allowing for more effective identification of yield loss patterns. This parameter change enables the correlation analysis to focus on batches with similar performance characteristics rather than those produced in sequence.
2Measurement precision
If analysis of large data sets from many production batches is performed to clarify root cause, then the accuracy of identifying detrimental factors improves, but the time and cost required for analysis increases significantly
Solution Approach 1:
The patent performs preliminary sequencing of batches by output data before conducting detailed correlation analysis. This preliminary action organizes the data in a way that makes subsequent analysis more efficient, allowing the system to identify patterns and detrimental factors without needing to analyze every batch in detail. The pre-sequencing step reduces the effective search space for root cause identification.
Solution Approach 2:
The patent creates a sequenced representation of batch data organized by output performance rather than requiring analysis of the original chronological data sets. This copying approach allows the system to work with a transformed version of the data that preserves the essential relationships while enabling faster pattern recognition and correlation analysis.
3Reliability
If multiple tool-subsets are analyzed at each process step to identify detrimental factors, then the completeness of the analysis improves, but the device complexity and difficulty of analysis increases
Solution Approach 1:
The patent segments the analysis by first grouping batches based on their output data characteristics, then analyzing tool-subset correlations within these segmented groups. This segmentation approach breaks down the complex multi-tool analysis into more manageable segments, making it easier to identify which specific tool-subsets are associated with yield loss in each performance category.
Data Source
AI summary
The disclosed embodiments include systems and methods of manufacturing a product. The system may include a non-transitory computer readable medium comprising computer readable program code for performing the method. The method may include manufacturing batches of the product according to steps of a process flow, determining output data for each batch, sequencing the batches by output data, determining a plurality of modes of output data based on grouping the batches, identifying a detrimental factor to output data in a process flow step based on a correlation between the process flow step and a mode of the plurality of modes, and correcting the detrimental factor.