Gradient-Based Sensor Identification for Failure Precursor Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current failure prediction systems in cyber-physical systems face challenges in accurately predicting failures due to the lack of descriptive information about failure precursors, making it difficult to identify and address underlying causes effectively.
Innovation Solution
A method and system that utilize a neural network to determine a prediction index from sensor time series data, detect failure precursors, and identify associated sensors by calculating gradients, allowing for timely corrective actions to prevent or minimize failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional failure prediction methods are used without descriptive information about precursors, then the system can operate with limited data, but the prediction accuracy and ability to identify underlying causes deteriorates
Solution Approach 1:
The patent introduces an intermediary representation called 'failure precursor signatures' that mediates between raw sensor data and failure prediction. These signatures act as a bridge, encoding descriptive information about failure precursors in a structured format that can be learned from historical data and applied to predict future failures with higher accuracy
Solution Approach 2:
The system performs preliminary action by pre-computing and storing failure precursor signatures from historical failure data before actual failure prediction is needed. This allows the system to have descriptive information about precursors ready in advance, improving prediction accuracy when failures are anticipated
2Reliability
If all sensors are monitored equally for failure prediction, then the system is comprehensive, but the complexity of identifying relevant sensors and their contributions increases
Solution Approach 1:
The patent replaces manual or rule-based sensor identification with an automated machine learning approach. The system uses trained models to automatically determine which sensors are relevant to specific failure precursors and quantifies their contributions, eliminating the need for complex manual analysis while maintaining reliability
Solution Approach 2:
The system changes the parameter of sensor relevance from a static, manual classification to a dynamic, data-driven metric. By using gradient-based methods and model predictions, the system continuously adjusts which sensors are considered relevant based on the specific failure mode being predicted, reducing identification complexity
3Ease of manufacture
If the system waits for failures to occur before learning from them, then the data collection process is simple, but the system cannot provide accurate predictions for rare or unique failure modes
Solution Approach 1:
The system performs preliminary action by pre-processing and storing failure precursor signatures from historical data in a structured format. This allows the system to have failure patterns ready for comparison against current sensor readings, enabling predictions for rare failure modes without requiring those specific failures to have just occurred
Solution Approach 2:
The patent uses copying by creating representations (signatures) of failure precursors from historical data. These copied signatures capture the essential characteristics of failure patterns and can be reused to predict similar failures, improving generalizability to rare failure modes while keeping data collection simple
Data Source
AI summary
Methods and systems for predicting failure in a cyber-physical system include determining a prediction index based on a comparison of input time series, from respective sensors in a cyber-physical system, to failure precursors. A failure precursor is detected in the input time series, responsive to a comparison of the prediction index to a threshold. A subset of the sensors associated with the failure precursor is determined, based on a gradient of the prediction index. A corrective action is performed responsive to the determined subset of sensors.