Facility Anomaly Analysis Using Parameter-Link Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for analyzing anomalies in electronic appliances, such as data centers or server farms, are inefficient and fail to quickly identify and rectify deviations from operational states, affecting reliability and availability.
Innovation Solution
A method and system using processor and memory circuitry to detect deviations by building models that link relevant parameters, allowing identification of mutable parameters whose changes can restore the system to an operational state, with periodic retraining of models based on collected data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional anomaly detection methods are used in facilities with electronic appliances, then the system can detect anomalies, but the analysis is inefficient and fails to quickly identify and rectify deviations from operational states
Solution Approach 1:
The system performs preliminary actions by continuously collecting and storing operational data from multiple parameters before anomalies occur. Models are pre-trained on this historical data to establish normal operational patterns, enabling rapid anomaly detection and root cause identification when deviations occur, thus improving efficiency and reducing response time.
Solution Approach 2:
The system transitions from traditional single-parameter anomaly detection to multi-dimensional analysis by collecting data from multiple parameters simultaneously and using machine learning models to analyze relationships across these dimensions. This holistic approach enables faster and more accurate identification of anomaly root causes by examining the system from multiple operational perspectives at once.
2Reliability
If complex monitoring systems are implemented to track multiple parameters, then anomaly detection capability improves, but the device complexity increases
Solution Approach 1:
The system implements self-service by using machine learning models that automatically learn from historical operational data and autonomously identify anomalies and their root causes without requiring complex manual configuration or expert intervention. The models continuously adapt to changing operational patterns, maintaining high reliability while keeping the system architecture relatively simple.
Solution Approach 2:
The system achieves universality by implementing a unified machine learning framework that can detect and analyze multiple types of anomalies across different parameters simultaneously. The same model architecture handles various operational parameters (temperature, pressure, flow rates, etc.), reducing overall system complexity compared to having separate specialized systems for each parameter.
3Ease of operation
If traditional anomaly response methods are used, then the system can identify anomalies, but it fails to automatically determine parameter modifications needed to restore operational state
Solution Approach 1:
The system implements feedback by continuously monitoring operational parameters, comparing current values against model predictions, and automatically determining the modifications needed to restore operational state. The models provide real-time feedback on which parameters should be adjusted and by how much, enabling automated corrective actions that improve ease of operation while maintaining high automation levels.
Solution Approach 2:
The machine learning models act as intermediaries between anomaly detection and corrective action implementation. Instead of directly controlling system parameters, the models analyze the anomaly, identify the root cause, and recommend optimal parameter adjustments, which are then implemented by the control system. This intermediary approach simplifies the overall control architecture while enabling sophisticated automated responses.
Data Source
AI summary
There is provided a system and method of analysing anomalies in one or more electronic appliances including at least one computer. The method comprises, by a processor and memory circuitry, upon detection of a deviation of a given parameter representative of the one or more electronic appliances from an operational state, providing a model associated with the given parameter, wherein the model links one or more other parameters to the given parameter, wherein the one or more other parameters affect the given parameter, and based at least on the model, identifying, among the one or more other parameters, at least one parameter Pj for which a change in its value allows bringing back the given parameter to the operational state.