Failure Prediction Sampling Around Failing-Case Centroids
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current statistical yield calculation methods do not scale well for problems with many parameters, making it difficult to effectively detect failures in manufactured objects.
Innovation Solution
A computer-implemented method that classifies samples with uniformly distributed parameter values, determines the centroid of failing samples, generates new samples around the centroid, and uses a nearest neighbor vector space for classification, enabling efficient failure prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current statistical yield calculation methods are used, then failure detection is possible, but the method does not scale well for problems with many parameters (hundreds or more)
Solution Approach 1:
The patent segments the high-dimensional parameter space into multiple lower-dimensional subspaces by identifying and analyzing critical parameter combinations separately. This allows the system to handle hundreds of parameters by breaking them into manageable groups that can be processed independently, thus scaling the failure detection capability to high-dimensional problems.
Solution Approach 2:
The patent transforms the analysis from direct high-dimensional sampling to a two-stage approach: first identifying critical parameter pairs (reducing dimensionality), then using those insights to guide comprehensive analysis. This dimensional transformation enables efficient handling of problems with hundreds of parameters by working through intermediate lower-dimensional representations.
2Reliability
If sampling in dimensions of many variables is performed, then statistical yield calculation can be done, but the computational cost increases significantly
Solution Approach 1:
The patent performs preliminary analysis by first examining pairwise parameter interactions to identify critical combinations before conducting full statistical yield calculation. This preliminary step filters out non-critical parameters and combinations, reducing the overall computational burden while maintaining accuracy in the final statistical yield assessment.
Solution Approach 2:
The patent applies different analysis strategies to different parameter combinations based on their criticality. Critical parameter pairs receive more detailed analysis while non-critical parameters are handled more efficiently, optimizing the allocation of computational resources across the parameter space to reduce overall computation time.
3Measurement precision
If refined models with hundreds of parameters are analyzed, then more accurate failure prediction is achieved, but the analysis time becomes unreasonable
Solution Approach 1:
The patent segments the analysis of refined models with hundreds of parameters by identifying and prioritizing critical parameter interactions. This segmentation allows the system to focus computational effort on the most influential parameter combinations while using efficient approximation methods for less critical parameters, achieving accurate failure prediction within reasonable timeframes.
Data Source
AI summary
A computer-implemented method and computing system are provided for failure prediction of a batch of manufactured objects. The method includes classifying, by a processor using a simulation, a set of samples with uniformly distributed parameter values, to generate sample classifications for the batch of manufactured objects. The method further includes determining, by the processor, a centroid of failing ones of the samples in the set, based on the sample classifications. The method also includes generating, by the processor, a new set of samples with a distribution around the centroid of the failing ones of the sample in the set. The method additionally includes populating, by the processor, a nearest neighbor vector space using the new set of samples. The method further includes classifying, by the processor, the new set of samples by performing a nearest neighbor search on the nearest neighbor vector space using a distance metric.