Predictive analytics for failure detection
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.
1. A computer program product for failure prediction of a batch of manufactured objects, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform a method comprising:
receiving the batch of manufactured objects for the failure prediction by coupling the batch of manufactured objects to the computer;
generating manufactured object sample classifications for the batch of manufactured objects comprising integrated circuits by training, by the computer using a simulation, a K Nearest Neighbor (K-NN) classifier by classifying a set of samples with uniformly distributed parameter values using a NN-based classification approach and NN vector space;
determining a centroid of failing ones of the samples in the set, based on the manufactured object sample classifications, the centroid being a geometric mean of the failing ones of the samples in the set;
generating a new set of samples with a distribution around the centroid of the failing ones of the samples in the set, the new set of samples relating to the batch of manufactured objects;
populating the NN vector space using the new set of samples;
classifying, using the trained K-NN classifier, the new set of samples by performing a nearest neighbor search on the nearest neighbor vector space using a distance metric and
assigning weights to each of the samples in the new set based on shifted sampling locations relative to the centroid using an inherited weight scheme when a number of neighbors is equal to one and using a voting scheme when the number of neighbors is odd and greater than one; and
selectively accepting or physically discarding the batch of manufactured objects based on a failure probability estimate generated for the batch of manufactured objects, the failure probability estimate based on classifications for the new set of samples.
2. The computer program product of claim 1 , wherein said generating step uses mixed importance sampling to reduce classification variation by shifting a sampling location to the centroid of the failing ones of the sample in the set.
3. The computer program product of claim 2 , wherein the mixed importance sampling selectively drives an accuracy of and a speed in generating classifications for the new set of samples.
4. The computer program product of claim 1 , wherein said classifying step comprises weighing respective contributions of the new set of samples by the distance metric.
5. The computer program product of claim 1 , wherein each of the samples in the new set of samples is weighted by a value of 1/d, where d is a distance from a given one of the samples to the centroid.
6. A computing system, comprising:
a processor and a memory operably coupled to the processor, configured to
receive a batch of manufactured objects for failure prediction by coupling the batch of manufactured objects to the processor;
generating manufactured object sample classifications for the batch of manufactured objects by training a K-NN classifier by classifying, using a simulation, a set of samples with uniformly distributed parameter values;
determine a centroid of failing ones of the samples in the set, based on the manufactured object sample classifications, the centroid being a geometric mean of the failing ones of the samples in the set;
generate a new set of samples with a distribution around the centroid of the failing ones of the samples in the set, the new set of samples relating to the batch of manufactured objects comprising integrated circuits;
populate the NN vector space using the new set of samples; and
classify, using the K-NN classifier the new set of samples by performing a nearest neighbor search on the nearest neighbor vector space using a distance metric and assigning weights to each of the samples in the new set based on shifted sampling locations relative to the centroid using an inherited weight scheme when a number of neighbors is equal to one and using a voting scheme when the number of neighbors is odd and greater than one; and
selectively accept or physically discard the batch of manufactured objects based on a failure probability estimate generated for the batch of manufactured objects, the failure probability estimate based on classifications for the new set of samples.
7. The computing system of claim 6 , wherein at least a portion of the system is comprised in a cloud configuration.
8. The computing system of claim 6 , wherein said classifying step comprises weighing respective contributions of the new set of samples by the distance metric.
9. The computing system of claim 6 , wherein each of the samples in the new set of samples is weighted by a value of 1/d, where d is a distance from a given one of the samples to the centroid.