Data Analytic Engine for Anomaly Detection in Complex Physical Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex physical systems generate vast amounts of time series data with nonlinear and heterogeneous characteristics, posing challenges for anomaly detection and system dynamics modeling without prior knowledge of the underlying system.
Innovation Solution
The method involves extracting features through sliding window segmentation and linear or nonlinear subspace decomposition, generating a system evolution model from an ensemble of models based on data properties, and determining a fitness score to detect anomalies in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If domain specific techniques are used to extract knowledge from measurement data, then understanding of system properties is improved, but human involvement and complexity increase
Solution Approach 1:
The system performs self-management by automatically extracting knowledge from measurement data using domain-independent analytic tools. The analytic engine autonomously processes time series data from sensors, generates system evolution models, and detects anomalies without requiring continuous human intervention or domain expert involvement, thereby reducing complexity while maintaining effective knowledge extraction
Solution Approach 2:
The patent employs domain-independent solutions that can be applied across different complex physical systems without requiring system-specific customization. The analytic engine uses general-purpose tools such as similarity-based approaches and support vector machines that work universally across diverse systems, eliminating the need for domain-specific expertise while maintaining effectiveness
2Extent of automation
If domain independent solutions are used to extract knowledge from data, then human involvement is reduced, but large amount of historical data and computational resources are required
Solution Approach 1:
The patent segments the analysis process into distinct components: data collection from sensors, feature extraction, system evolution model generation, and anomaly detection. By dividing the complex analysis into manageable segments, the system reduces computational burden at each stage while maintaining automation. The sliding window approach segments time series data into manageable chunks for processing
Solution Approach 2:
The system performs preliminary feature extraction and creates system evolution models from historical data before actual anomaly detection is needed. This preliminary modeling phase prepares the analytic engine to quickly detect anomalies in real-time without requiring extensive computational resources during operational monitoring, thus reducing the computational burden during critical detection phases
3Measurement precision
If massive sensors are deployed to monitor physical systems, then measurement data collection is improved, but data analysis complexity and computational expense increase
Solution Approach 1:
The patent extracts only the most relevant features from the massive amount of measurement data collected by sensors. Rather than analyzing all raw sensor data, the system identifies and extracts key features that are most indicative of system state and potential anomalies. This selective extraction reduces data analysis complexity while maintaining monitoring precision
Solution Approach 2:
The system introduces an analytic engine as an intermediary layer between the sensors and the monitoring system. This analytic engine processes the raw measurement data, performs feature extraction, generates system evolution models, and presents processed information for anomaly detection. The intermediary analytic engine simplifies the complexity by handling data processing tasks automatically
Data Source
AI summary
Systems and methods for anomaly detection in complex physical systems, including extracting features representative of a temporal evolution of the complex physical system, and analyzing the extracted features by deriving vector trajectories using sliding window segmentation of time series, applying a linear test to determine whether the vector trajectories are linear, and performing subspace decomposition on the vector trajectory based on the linear test. A system evolution model is generated from an ensemble of models, and a fitness score is determined by analyzing different data properties of the system based on specific data dependency relationships. An alarm is generated if the fitness score exceeds a predetermined number of threshold violations for the different data properties.