Sensor Health Modeling for Multi-Mode Building Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fault detection and diagnostics (FDD) systems in building management systems rely on hard-coded rules, which lack scalability and fail to detect subtle changes in sensor data, leading to inaccurate fault detection and inability to distinguish between bad sensors and abnormal equipment operation.
Innovation Solution
A sensor health system using machine learning models to classify sensor data into specific modes, generate mode-specific distributions, and compare these distributions to expected distributions to identify abnormalities, initiating corrective actions when deviations exceed thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If hard coded rules are used for fault detection, then the system is simple to implement, but the system lacks scalability and cannot detect subtle changes in sensor data
Solution Approach 1:
The patent replaces hard-coded mechanical rule-based systems with machine learning models that automatically learn patterns from sensor data. The ML models can detect subtle changes and adapt to different operating modes without requiring manual rule configuration, thereby improving scalability and detection capability while maintaining ease of implementation through automated training processes.
Solution Approach 2:
The patent changes the fundamental parameter of fault detection from fixed threshold values to dynamic, learned distributions. By using mode-specific distributions generated from training data, the system adapts to different operating conditions and detects anomalies relative to the expected behavior for each mode, rather than using universal hard-coded thresholds.
2Ease of operation
If a single threshold range is used for fault detection across multiple operating states, then the rule is easy to manage, but the system fails to detect abnormal operation in specific states
Solution Approach 1:
The patent segments the fault detection process by creating mode-specific distributions for different operating states. Each operating mode receives its own tailored distribution based on training data from that specific mode, allowing precise detection within each state while the overall system remains managed through a unified multi-modal framework.
Solution Approach 2:
The patent introduces dynamic adaptation by automatically selecting the appropriate mode-specific distribution based on the current operating state. The system dynamically adjusts its detection criteria to match the expected behavior of each mode, improving measurement precision without requiring manual management of multiple static rules.
3Device complexity
If conventional FDD systems use hard coded rules, then the system structure is simple, but the system cannot distinguish between bad sensors and abnormal equipment operation
Solution Approach 1:
The patent introduces mode-specific distributions as an intermediary layer between raw sensor data and fault detection decisions. These distributions serve as a reference model that captures normal behavior for each operating mode, allowing the system to distinguish between sensor failures and actual equipment anomalies by comparing against the expected multi-modal distribution.
Data Source
AI summary
A sensor health system for building equipment obtains sensor data measuring a variable state or condition affected by operating the building equipment, classifies the sensor data into a particular mode of a plurality of modes corresponding to a plurality of operating states of the building equipment, generates a mode-specific distribution of the sensor data corresponding to the particular mode of the plurality of modes, identifies the sensor data as abnormal by comparing the mode-specific distribution of the sensor data to an expected mode-specific distribution selected from a plurality of expected mode-specific distributions corresponding to the plurality of modes, and initiates a corrective action in response to identifying the sensor data as abnormal.