Statistical Process Control for Multiple Batch Effects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing statistical process control methods fail to effectively qualify manufacturing processes with multiple batch effects, leading to incorrect qualification of compliant and non-compliant items, which can result in quality issues and increased testing costs.
Innovation Solution
A method and system that utilize a linear mixed model to determine statistical process control parameters, such as process capability index and control limits, accounting for multiple batch effects by measuring a quantifiable property of items, developing a statistical process control standard deviation, and computing parameters like Cpk* to ensure compliance and reduce sampling costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If known statistical process control techniques are used in the presence of multiple batch effects, then the process qualification and control charting can be performed, but incorrect items are rejected or qualified leading to quality degradation
Solution Approach 1:
The patent modifies the statistical parameters used in process control by developing new estimators for process standard deviation that explicitly account for multiple batch effects. The process capability index is recalculated using these adjusted parameters, transforming the statistical approach to accommodate batch variability without compromising quality decisions.
Solution Approach 2:
The patent introduces batch effect as an intermediary factor in the statistical model. By incorporating batch effects into the mixed-effects model, the methodology mediates between the observed variability and the true process capability, preventing incorrect qualification or rejection of items due to batch-related variations.
2Loss of energy
If traditional statistical process control methods are applied without accounting for batch effects, then testing costs can be reduced through process qualification, but the qualification results become unreliable
Solution Approach 1:
The patent changes the parameter estimation methodology by using mixed-effects models to separate batch-related variability from true process variability. This allows for more accurate process capability assessment, enabling reduced sampling plans that lower testing costs while maintaining qualification reliability.
Solution Approach 2:
The patent segments the total variability into distinct components: batch effects and within-batch variability. By separating these sources of variation, the methodology can accurately assess true process capability independent of batch effects, enabling cost-effective reduced sampling while maintaining reliability.
3Device complexity
If process capability index is calculated without considering batch effects, then the calculation is simple, but compliant items are incorrectly rejected or non-compliant items are incorrectly qualified
Solution Approach 1:
The patent extends the traditional process capability index calculation by incorporating batch effect parameters into the standard deviation estimate. While this increases calculation complexity, it dramatically improves the accuracy of compliance determination by accounting for the source of variability.
Solution Approach 2:
The patent introduces batch effect as a mediating factor in the capability index calculation. By including this intermediary component, the calculation accurately reflects true process capability while filtering out batch-related noise, leading to correct compliance decisions.
Data Source
AI summary
Techniques for qualifying for use in an overall manufacturing process items produced by a bulk manufacturing process that has a plurality of batch effects are presented. The techniques can include obtaining a collection of items produced by a bulk manufacturing process that has a plurality of batch effects; measuring a quantifiable property of a sample of items from the collection of items; developing a linear mixed model for the quantifiable property based on the measuring; determining a statistical process control standard deviation for the collection of items based on the linear mixed model; computing a statistical process control parameter from the statistical process control standard deviation; determining that at least a portion of the collection of items conform to the statistical process control parameter; accepting at least a portion of the collection of items; and using at least a portion of the collection of items in the overall manufacturing process.