Thermal Anomaly Management via Deep Learning and Bayesian Causal Hierarchy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying and diagnosing thermal anomalies in data centers is challenging due to varying formats and structures of monitored variables, complex patterns, and the need for domain expertise, leading to potential false causal models and ineffective anomaly recognition.
Innovation Solution
A computer-implemented method using a deep learning system to identify thermal anomalies from recorded environment parameter data, combined with a Bayesian network to identify relationships between environment parameters and develop a causal explanation hierarchy, enabling real-time intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional thermal mapping and data-driven approaches are used to detect thermal anomalies, then anomaly recognition can be achieved to some extent, but the approaches rely on specific models proving accurate predictors of real-world data centre behaviour and require significant domain expertise
Solution Approach 1:
The patent introduces an intermediary layer between raw sensor data and anomaly detection that automatically learns and adapts to the specific data centre environment. This intermediary consists of trained machine learning models that translate complex sensor patterns into meaningful thermal anomaly indicators, eliminating the need for domain experts to manually create and validate complex thermal models while maintaining high detection accuracy.
2Reliability
If detailed domain knowledge is used to identify causal relationships between environment parameters, then more accurate causal models can be developed, but this requires significant domain expertise and may lead to false causal models if understanding is incorrect or incomplete
Solution Approach 1:
The system employs machine learning models that automatically learn causal relationships between environment parameters from historical data without requiring domain expertise. The models self-service by autonomously identifying patterns, correlations, and causal structures in the data, eliminating the need for human experts to manually construct and validate causal models while improving reliability through data-driven insights.
3Loss of information
If all environment parameter data is used to provide causal explanation for thermal anomalies, then comprehensive analysis can be achieved, but this increases processing time and computational resources
Solution Approach 1:
The patent extracts and focuses only on the most relevant environment parameters that contribute to thermal anomalies using machine learning feature selection techniques. By identifying and extracting the key causal parameters from the full set of sensor data, the system provides comprehensive causal explanations for anomalies while significantly reducing processing time and computational resource requirements by excluding irrelevant parameters.
Data Source
AI summary
Some embodiments relate to a method of managing thermal anomalies in an environment is described. A deep learning system is trained to identify thermal anomalies from recorded environment parameter data. A Bayesian network is also trained to identify relationships between environment parameters, and the identified relationships are used to develop a causal explanation hierarchy. Using these trained systems, environment parameters are measured first to identify a thermal anomaly, and on identification of a thermal anomaly, and then to provide a causal explanation hierarchy for the thermal anomaly. This enables a real-world intervention to address the thermal anomaly. A suitable system to perform this method is also described.