Fault Response TRIZ Cases
- PLC Virtual Modules for Process Instrument Anomaly Detection
Server-based diagnosis linked to PLC virtual modules detects process instrument anomalies and applies default data to maintain control accuracy.
- Context-Aware Feature Generation for Asset Fault Prediction
Generated features link sensor data to operating states and sensor interdependencies, improving anomaly scoring and earlier fault prediction.
- Static Violation Root Cause Analysis Using Repository Pruning
Preprocessed static-checker data and repository pruning identify SoC design violation root causes faster while reducing manual debug effort.
- CPU Fault Diagnosis Line for Partitioned Server Boards
Pulled-up fault indication lines and switch control simplify CPU fault detection in partitioned server boards while reducing cabling complexity.
- Lithography Predictive Maintenance Using Transfer Entropy Causality
Transfer entropy and quality-weighted context data improve lithography fault diagnosis and alert prioritization for predictive maintenance.
- Path Planning Error Detection Using Human-Driven Reference Paths
Comparing predicted vehicle paths with human-driven reference paths captures corner-case errors while storing only relevant context data.
- Sensor Index Selection Using PCA Migration Distance
Partitioned sensor blocks and PCA migration distance reveal local-feature indexes, improving failure prediction with fewer redundant signals.
- CPS Anomaly Detection Using Forecast Error Thresholds
Forecasted CPS feature values and total error thresholds enable earlier anomaly detection and source identification in hazardous cyber-physical systems.
- Abnormality Detection Support Using Representative Sensor Time Series
Automatically links overall abnormality scores to representative sensor time series, cutting manual screening in complex system monitoring.
- Failure Probability Evaluation Using Tail-Specific Density Estimation
Extreme-value tail modeling is combined with central distribution fitting to improve mechanical failure probability evaluation accuracy.
- Multimodel Fault Detection Using Constrained Kalman Filtering
A constrained Kalman filter supervises submodels across operating regimes to cut false alarms and improve fault isolation under noise.
- Display Image Verification Using Split-Light Feedback
A split-light verification path checks displayed image data against processed sensor data to catch visual output errors in critical displays.
- Failure Prediction Sampling Around a Failure Centroid
Failure-prone regions are sampled around a computed centroid, cutting high-dimensional analysis time while preserving detection accuracy.
- Edge Connection Control for Secure Abnormality-Triggered Remote Access
Remote access stays closed during normal plant operation, then an edge unit opens a secure monitoring link only when a qualifying abnormality is detected.
- Failure Prediction Sampling Around Failing-Case Centroids
Targets high-dimensional yield analysis by resampling near failing-sample centroids and using nearest-neighbor classification to predict failures faster.
- Performance Indicator Correlation for Incident Lead-Time Prediction
Correlated performance indicators estimate time to a computing incident, filtering weak signals and reducing false alarms for faster maintenance action.
- Machine State Diagnosis Using Normalized Autoencoder Features
Statistical feature normalization lets autoencoders learn low-amplitude displacement and frequency signals for more accurate machine diagnosis.
- Forecasting Models for Data Stream Anomaly Detection
Multiple periodic forecasting models score incoming data streams to cut false positives and flag true anomalies as patterns change.
- HVAC Fault Awareness Through Intentional Comfort Setting Changes
A detected HVAC fault triggers a deliberate comfort shift that prompts occupants to check the interface, reducing damage and service delay.
- Inference Workflow for Sensor-Aware Manufacturing Defect Analysis
Question classification switches between sensor data and user input to speed manufacturing defect inference without added setup cost.
- Abnormality Detection Model Generation with Virtual Feature Data
Virtual feature data expands limited pre-operation training sets, improving abnormality detection accuracy for predictive maintenance.
- Production Failure Information Sharing for Timely Countermeasures
Networked distribution of failure and avoidance information helps production managers act early to prevent downtime and defects.
- Multi-Period Forecast Scoring for Data Stream Anomaly Detection
Incoming stream values are scored against models with different periodicities to reduce false positives and flag anomalies faster.
- Unsupervised Machine Failure Prediction From Sensor Patterns
Real-time unsupervised analysis of machine sensor data detects anomaly patterns early, helping cut downtime and avoid premature maintenance.
- Vehicular Telematics Fault Detection Using Expected Communication States
Expected-versus-actual communication monitoring pinpoints vehicular network faults in real time while avoiding heavy telematics log processing delays.
- Deep Belief Network Learning for Equipment Failure Signatures
Failure signatures and unified sensor data help deep belief network agents predict equipment faults across similar assets despite siloed data.
- Time-Series Input Correction for Defect Detection and Prediction
Comparing predictions from raw and estimated time-series data helps detect input defects and improve analysis reliability.
- Gaze-Tracked Automation Display for Pilot State Awareness
Gaze tracking checks whether a pilot has seen automation state changes and their trigger conditions, then issues alerts to prevent missed mode awareness.
- Neural Network Anomaly Detection for Equipment Health Monitoring
Automated neural network monitoring compares actual and nominal sensor data to flag equipment anomalies early and cut manual model-building effort.