Artificial Intelligence-Based Methods for Assessment of Accrued Damage and Remaining Use-Life in Automotive Underhood Electronics
- Resource Type
- Conference
- Authors
- Lall, Pradeep; Thomas, Tony; Mehta, Vishal
- Source
- 2024 Pan Pacific Strategic Electronics Symposium (Pan Pacific) Pan Pacific Strategic Electronics Symposium (Pan Pacific), 2024. :1-12 Jan, 2024
- Subject
- Aerospace
Bioengineering
Communication, Networking and Broadcast Technologies
Components, Circuits, Devices and Systems
Computing and Processing
Engineered Materials, Dielectrics and Plasmas
Engineering Profession
Fields, Waves and Electromagnetics
General Topics for Engineers
Photonics and Electrooptics
Power, Energy and Industry Applications
Robotics and Control Systems
Signal Processing and Analysis
Transportation
Vibrations
Fuses
Mission critical systems
Predictive models
Time measurement
Plastics
Long short term memory
Prognostics Health Management
Reliability
Leading Indicators of Failure
Electronic Packaging
Automotive Electronics
Artificial Intelligence
- Language
To ensure the smooth and uninterrupted operation of electronic systems over their lifespan, it is essential to assess progressive damage accrual and remaining useful life in operation. This paper proposes Prognostics Health Management methods that utilize feature vector-based assessment to detect damage initiation and progression in such systems. The presented method can effectively detect realtime impending failures in mission-critical electronics and assess mission readiness before deployment or during operation. Previous research in this field has mainly relied on fuses and canaries for failure detection, but this paper presents a method that assesses evolving damage in complex systems with nonlinear material behavior. Identification of meaningful state-vectors for correlation with damage progression, feature-vector engineering is pursued. The study identifies feature vectors for various mechanical shock and vibration levels that predict remaining useful life, even for different solder interconnect materials. The selected feature vector is modeled using the Long Short-term Memory (LSTM) deep learning technique to predict the remaining useful life during drops. The study validates the feature vectors' accuracy by correlating LSTM's prediction of plastic work with that predicted by finite element simulation and experimentally measured time to failure.