Prognostic Pattern-Recognition System for Sensor Failure Resilience
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing enterprise computer systems face downtime and vulnerability to incipient anomalies when sensors fail, as current retraining processes require collecting new data, leading to 'blackout periods' and increased risk of undetected issues.
Innovation Solution
A technique for training a prognostic pattern-recognition system to detect anomalies by initially gathering and storing telemetry data from all sensors, then training additional models to operate with missing sensors, allowing seamless switching to pre-trained models when failures occur, thus avoiding downtime.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the system retrain the prognostic pattern-recognition model after sensor failure by collecting new training data, then the model can operate with fewer sensors, but the system enters a blackout period during which predictive analytic techniques cannot be used
Solution Approach 1:
The system pre-trains multiple prognostic models during the initialization phase, each model trained to operate with different combinations of missing sensors. When a sensor failure occurs, the system can immediately switch to a pre-trained model that matches the current sensor configuration, eliminating the need for retraining and avoiding blackout periods.
Solution Approach 2:
The system divides the prognostic modeling into multiple specialized models, where each model is trained for a specific sensor failure scenario. This segmentation allows the system to have dedicated models ready for different failure modes, enabling immediate switching without general retraining.
2Measurement precision
If the system uses threshold-based warnings to detect operational issues, then the system can identify problems when they occur, but the warnings are reactive and the system performance is already severely affected by the time the threshold is exceeded
Solution Approach 1:
The prognostic pattern-recognition model analyzes telemetry data to predict future system states and detect incipient anomalies before they manifest as threshold violations. By performing preliminary analysis on historical and current data, the system provides early warnings that enable proactive problem prevention rather than reactive response.
Data Source
AI summary
The disclosed embodiments relate to a technique for training a prognostic pattern-recognition system to detect incipient anomalies that arise during execution of a computer system. During operation, the system gathers and stores telemetry data obtained from n sensors in the computer system during operation of the computer system. Next, the system uses the telemetry data gathered from the n sensors to train a baseline model for the prognostic pattern-recognition system. The prognostic pattern-recognition system then uses the baseline model in a surveillance mode to detect incipient anomalies that arise during execution of the computer system. The system also uses the stored telemetry data to train a set of additional models, wherein each additional model is trained to operate with one or more missing sensors. Finally, the system stores the additional models to be used in place of the baseline model when one or more sensors fail in the computer system.