Invention Grant
- Patent Title: Frequency-domain impedance sensing system and method for neural network prediction of impending failure modes
-
Application No.: US16823811Application Date: 2020-03-19
-
Publication No.: US11461660B1Publication Date: 2022-10-04
- Inventor: Orion Davies
- Applicant: Rockwell Collins, Inc.
- Applicant Address: US IA Cedar Rapids
- Assignee: Rockwell Collins, Inc.
- Current Assignee: Rockwell Collins, Inc.
- Current Assignee Address: US IA Cedar Rapids
- Agency: Suiter Swantz pc llo
- Main IPC: G06N3/08
- IPC: G06N3/08 ; G01R31/316 ; G06N3/04 ; H03M1/12

Abstract:
A system and method for monitoring a circuit for impending failure includes measuring changes to frequency-domain impedance with a pair of analog-to-digital converters connected to a conductor while powering up, with or without interposing a series sense resistor for measuring applied current. Changes to frequency-domain impedance or power delivery voltage ratio are identified via a neural network trained to identify frequency-domain impedance associated with normal system behavior and general system failure mode. The neural network may be further trained to produce a predictive probability of specific failure mode types.
Information query