Personalized Event Prediction for Sensor-Monitored Physical Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing event prediction systems for physical systems, such as heart failure in humans or engine failure in airplanes, rely on data from multiple similar systems to find common characteristics, leading to inaccurate predictions due to diverse patient populations or systems, and require large data collections.
Innovation Solution
A method that creates a personalized statistical model based solely on data from individual physical systems using sensors, analyzing time-series data to detect deviations from normal behavior and predict events like failure, without relying on data from similar systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data from multiple similar physical systems is collected to find common characteristics for event prediction, then the system can build predictive models, but the prediction accuracy deteriorates due to diversity among systems (e.g., patient populations)
Solution Approach 1:
The patent segments the prediction problem by creating separate predictive models for each individual physical system rather than using a single model for all systems. Each system's data is processed independently to capture its unique characteristics, thereby maintaining high prediction accuracy despite system diversity. This is achieved through personalized statistical models that analyze individual time-series data patterns.
Solution Approach 2:
Instead of the conventional approach of aggregating data from multiple systems to find common patterns, the patent inverts the approach by focusing on individual system data and creating personalized models. This inversion allows the system to accommodate diversity while maintaining accuracy by treating each system's unique characteristics as the basis for prediction rather than as noise to be filtered out.
2Reliability
If data from multiple similar physical systems is collected to build predictive models, then models can be created, but large amounts of data are required
Solution Approach 1:
The patent segments the data collection and modeling process into individual system-specific analyses. Each system requires only its own historical data to build a personalized predictive model, eliminating the need to aggregate large volumes of data from multiple systems. This segmentation reduces the total data collection burden while maintaining or improving model quality through personalized approaches.
Solution Approach 2:
Each physical system serves itself by generating its own predictive model using only its collected data, without requiring data from other systems. This self-service approach allows individual systems to become self-sufficient in terms of prediction, reducing the overall data infrastructure requirements and enabling decentralized predictive analytics.
3Adaptability or versatility
If population-based predictive models are used for diverse systems, then generalization is attempted, but prediction accuracy suffers due to difficulty in identifying common profiles
Solution Approach 1:
The patent inverts the traditional generalization approach by abandoning the search for common profiles across diverse systems. Instead, it creates specific personalized models for each individual system, accepting that systems are too diverse to share common characteristics. This inversion transforms the problem from finding similarities to embracing differences, thereby maintaining high accuracy across diverse populations.
Solution Approach 2:
The patent applies local quality by tailoring the predictive model to each individual system's specific characteristics rather than applying a uniform model across all systems. Each system receives a customized model that reflects its unique operational patterns and data characteristics, ensuring locally optimized prediction accuracy rather than globally averaged performance.
Data Source
AI summary
A physical system receives and records measurements from a plurality of sensors for the physical system over a period of time and creates therefrom a statistical model of normal behavior of the physical system. The statistical model is applied to monitor the physical system for events or significant changes. The method detects events or significant changes in the operation or behavior of the physical system responsive to the monitoring of the physical system for events or significant changes. Events or significant changes in the operation or behavior of the physical system may cause a notification or alert, which can be sent a supervisory system.