Detecting Faulty Hardware Using Neural Networks TRIZ Cases
- Semiconductor Process Inspection Using Simulation-Based Defect Prediction
Comparing measured device data with virtual process simulations helps predict defect-causing factors faster with less analysis time and compute.
- RUL Estimation Using Censored Time-Series Ordinal Regression
Uses LSTM-RNN ordinal regression to estimate remaining useful life from failed and censored sensor time series when failure labels are scarce.
- Hardware Failure Prediction With Local Diagnostic Confirmation
Telemetry-based failure alerts are verified by local diagnostics, reducing premature hardware replacement and unexpected downtime.
- Multi-Sensor Neural Monitoring Circuit for Compact Abnormal Detection
A shared multi-sensor circuit uses analog neural processing to classify abnormal conditions while cutting IoT device size and design cost.
- NPU Built-In Self-Test for Runtime Defect Isolation
Runtime scan testing lets an NPU detect minor defects in active components and isolate faulty blocks before erratic AI behavior spreads.
- NPU Runtime Scan Testing for Faulty Component Isolation
Runtime scan chains let an NPU detect latent defects in memory and logic blocks, then isolate faulty components to prevent erratic AI behavior.
- NPU Runtime Function Testing for Defect Isolation
Runtime function and scan testing helps NPUs detect latent defects, isolate faulty components, and avoid erratic AI behavior.