Building Equipment Fault Prediction with Cross-Device Model Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault detection systems in building management systems often rely on a robust set of historical data with multiple instances of different types of fault events, which is not always available in practice, limiting their effectiveness.
Innovation Solution
A system that augments data sets for fault prediction models using supplemental data from multiple devices, clusters devices based on characteristics, and trains fault prediction models to improve fault detection and diagnostics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fault prediction models are trained using only local device data, then model training is simple and fast, but the model accuracy is insufficient when historical fault data is limited
Solution Approach 1:
The patent combines data from multiple devices of the same type into a unified training dataset. The server collects data from a plurality of devices, merges them together, and uses this combined dataset to train fault prediction models, thereby improving model accuracy when individual device data is insufficient.
Solution Approach 2:
The patent creates a universal fault prediction model that can be applied across multiple devices of the same type. The model trained on aggregated data from multiple devices serves as a general solution that can predict faults for any device in the group, making the system multi-functional rather than device-specific.
2Measurement precision
If data from multiple devices is collected and merged, then model training accuracy improves, but data privacy and security risks increase
Solution Approach 1:
The patent introduces a server as an intermediary that handles data collection, merging, and model training. The server acts as a trusted mediator that processes sensitive device data without requiring direct access to individual device systems, thereby reducing security risks while enabling data aggregation for improved model accuracy.
3Reliability
If fault prediction models are trained with sufficient historical fault data, then prediction reliability improves, but the system cannot be deployed when such data is unavailable
Solution Approach 1:
The patent performs preliminary data aggregation and model training by collecting and merging data from multiple devices before deployment. The server proactively gathers data from various devices, merges them, and trains models in advance, so that when individual devices need fault prediction, the models are already trained and ready to use, eliminating the need for each device to have its own extensive historical fault data.
Data Source
AI summary
A system includes a plurality of devices of building equipment, an additional device of building equipment, and a computing system. The computing system is configured to process data from the plurality of devices to extract common features of the plurality of devices, train a global model based on the common features, obtain additional data from the additional device, adapt the global model for the additional device based on the additional data to obtain an adapted model for the additional device, predict a status of the additional device using the adapted model, and affect an operation of the additional device based on the status.