Edge Vibration Anomaly Detection for Low-Bandwidth Building Equipment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for monitoring building equipment struggle with efficiently analyzing high-frequency vibration data from building equipment, leading to increased communication bandwidth requirements and potential issues with unstable network connections, which can result in missed faults and inefficient use of analyst time.
Innovation Solution
A system that includes vibration sensors coupled with an edge device capable of detecting abnormalities in vibration data using machine learning models, performing signal processing, and communicating relevant data only when certain criteria are met, thereby reducing the need for continuous data transmission and enhancing fault detection efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If continuous high-frequency vibration data is transmitted to remote systems for analysis, then fault detection capability is improved, but communication bandwidth requirements increase and network instability causes missed faults
Solution Approach 1:
The system segments the data processing function by separating edge devices that perform local signal processing and anomaly detection from remote computing systems. The edge device processes high-frequency vibration data locally and transmits only processed results or alerts, dividing the overall monitoring function into distributed components that reduce network data transmission requirements
Solution Approach 2:
The edge device performs preliminary signal processing, feature extraction, and anomaly detection before data transmission to remote systems. By pre-processing the high-frequency vibration data at the edge and identifying potential faults locally, the system reduces the volume of data that needs to be transmitted while ensuring faults are detected, thereby improving reliability without proportionally increasing bandwidth requirements
2Reliability
If all vibration data is transmitted for analysis, then complete fault detection is achieved, but analyst time and processing resources are wasted on normal data
Solution Approach 1:
The edge device applies local quality processing by performing signal processing and anomaly detection specific to the local vibration data characteristics. The system tailors the processing to identify only relevant anomalies at each edge location, transmitting only data that requires further analyst review, thus reducing wasted analyst time on normal data while maintaining complete fault detection
Solution Approach 2:
The system implements feedback mechanisms where the edge device continuously monitors vibration data and automatically transmits alerts when anomalies are detected. This feedback loop enables the system to focus analyst attention only on actual problems rather than continuously reviewing all normal data, reducing analyst time loss while maintaining detection completeness
3Quantity of substance
If edge devices perform real-time signal processing and machine learning analysis, then data transmission is reduced, but device complexity increases
Solution Approach 1:
The system segments computational complexity across different levels: the edge device handles signal processing, feature extraction, and basic anomaly detection, while more complex machine learning model training and refinement are performed remotely. This segmentation allows edge devices to have sufficient processing capability for real-time operation without requiring full-scale complex AI infrastructure at each location
Solution Approach 2:
The edge device acts as an intermediary between vibration sensors and remote computing systems. It performs necessary signal processing and preliminary analysis locally to reduce data transmission volume, while serving as a bridge that connects to remote systems for more complex processing when needed, thus balancing local complexity requirements with overall system capabilities
Data Source
AI summary
A system includes a plurality of vibration sensors configured to be coupled to a unit of building equipment, an edge device arranged between the plurality of vibration sensors and a communications network. The edge device is programmed to detect an abnormality in vibration data from the plurality of vibration sensors by ingesting streams of vibration data from the plurality of vibration sensors, performing signal processing on the streams of vibration data to obtain inputs for a machine learning model, determining whether the abnormality is occurring by applying the inputs to the machine learning model, determining a trend associated with the abnormality based on outputs of the machine learning model, and causing a portion of the vibration data to be communicated to a first computing system via the communications network in response to the trend satisfying a criterion.