Probabilistic Networks TRIZ Cases
- Perception Result Display Using Context Instead of Confidence Scores
Low-confidence perception outputs are filtered, while contextual cues replace raw confidence scores so non-technical users can judge result reliability.
- Actuator Control Parameter Tuning via Stationary Distributions
Uses stationary probability distributions instead of finite-horizon predictions to tune actuator control parameters for more reliable long-term adaptation.
- Virtual Network Allocation Using Deep RL for High-Dimensional States
Deep reinforcement learning improves virtual network allocation accuracy by handling continuous, high-dimensional network states and user demand.
- Microalloyed Steel Property Prediction Using Carbonitride Precipitation
A globally additive model predicts microalloyed steel properties from carbonitride precipitation, cutting physical tests and R&D time.
- Data Center Climate Control With Predictive Cooling Adjustment
Climate sensors and workload prediction adjust cooling flow and device placement to cut data center energy use while maintaining safe temperatures.
- Aircraft Trajectory Uncertainty Evaluation Using Polynomial Chaos
Polynomial chaos expansions quantify aircraft trajectory uncertainty and sensitivity faster than Monte Carlo for real-time air traffic support.
- Engineering Autonomy Robustness Scoring With Stratified Sampling
Optimal stratified sampling and mixed-fidelity models quantify component robustness across multidimensional design spaces with lower computation.
- Hierarchical Driving Policy Switching for Abrupt Phase Transitions
A high-level RL policy switches among imitative driving modes to handle near-accident phase transitions with less state-space exploration.
- Sensor-to-Expertise Translation for Machine Fault Diagnosis
Machine learning combines sensor signals with expert audio, video, and notes to preserve tacit knowledge and improve fault diagnosis.
- Artificial Hierarchical Memory for Real-Time Operator Intent Prediction
Hierarchical memory and temporal pattern learning filter sensor data to predict operator commands in real time during dynamic machine operation.