Feature Vector Based Remaining Useful-Life Assessment in Mechanical Shock and Vibration for Leadfree Electronics
- Resource Type
- Conference
- Authors
- Lall, Pradeep; Thomas, Tony; Blecker, Ken
- Source
- 2022 IEEE 72nd Electronic Components and Technology Conference (ECTC) ECTC Electronic Components and Technology Conference (ECTC), 2022 IEEE 72nd. :248-259 May, 2022
- Subject
- Components, Circuits, Devices and Systems
Engineered Materials, Dielectrics and Plasmas
Engineering Profession
Photonics and Electrooptics
Vibrations
Fuses
Electric shock
Mission critical systems
Predictive models
Lead
Time measurement
PCB
Vibration
Drop and Shock
Strain Analysis
Principal Component Analysis (PCA)
FFT
Failure Prediction
Remaining Useful Life (RUL)
- Language
- ISSN
- 2377-5726
Electronics in harsh environments including automotive and defense may be subjected to shock and vibration in transport, storage, deployment and normal operation. High reliability operation requires the progressive-assessment of damage accrual and remaining useful life for assurance of uninterrupted operation over time-in-service. Prognostics Health Management methods have been proposed in this paper for feature vector-based assessment of damage initiation and damage progression in electronic systems. The methods can be used for early-indication of impending failure for mission critical electronics during operation or assessment of mission readiness prior to deployment. Prior work in the area focuses on the use of fuses and canaries for identification of impending failure. Prior implementation of data-driven methods often ignore the failure mechanics. Solution for progressive-assessment of evolving damage in complex systems involving nonlinear material behavior has been presented. Feature vector have been identified for prediction of remaining useful life for various acceleration levels of mechanical shock and vibration. Damage progression has been studied in test-vehicles fabricated from different solder-interconnects including SAC105 and SAC305 with the same assembly-geometry and architecture. The changes in the feature vectors based on the differences in the solder material were also studied. Feature-vector engineering has been pursued to construct meaningful data-vectors for correlation with the underlying progression of damage in pristine and aged assemblies. Feature vector selected from the time and frequency domain analysis of the strain signal is modeled based on Long Short-term Memory (LSTM) deep learning technique to predict the packages' remaining useful file during the drop. Validation cases have been presented for the feature vectors identified in the study by correlating the plastic work predicted by finite elements simulation with that predicted using LSTM. In addition, the remaining useful life prediction has been correlated with experimentally measured time to failure.