US 12,131,269 B2
Time-series fault detection, fault classification, and transition analysis using a k-nearest-neighbor and logistic regression approach
Dermot Cantwell, Sunnyvale, CA (US)
Assigned to Applied Materials, Inc., Santa Clara, CA (US)
Filed by Applied Materials, Inc., Santa Clara, CA (US)
Filed on Feb. 14, 2020, as Appl. No. 16/792,021.
Application 16/792,021 is a continuation of application No. 15/269,530, filed on Sep. 19, 2016, granted, now 10,565,513.
Prior Publication US 2020/0210873 A1, Jul. 2, 2020
This patent is subject to a terminal disclaimer.
Int. Cl. G06N 7/01 (2023.01); G06F 11/07 (2006.01); G06N 20/00 (2019.01)
CPC G06N 7/01 (2023.01) [G06F 11/0727 (2013.01); G06F 11/0751 (2013.01); G06F 11/0787 (2013.01); G06F 11/079 (2013.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving historical time-series data, the historical time-series data having been generated by one or more sensors during excursion behavior of one or more semiconductor processing processes via semiconductor processing equipment;
generating training data comprising a plurality of randomized data points associated with an expected range for the excursion behavior of the historical time-series data; and
training, by a processing device, a logistic regression classifier based on the training data to generate a trained logistic regression classifier, wherein the trained logistic regression classifier is associated with a logistic regression that indicates a location of a transition pattern from one or more first data points indicating the excursion behavior to one or more second data points indicating non-excursion behavior of the one or more semiconductor processing processes via the semiconductor processing equipment, wherein the transition pattern reflects about a reflection point located on the transition pattern, the trained logistic regression classifier being capable of indicating a probability that new time-series data generated during a new execution of the one or more semiconductor processing processes matches the excursion behavior corresponding to the historical time-series data.