ML-Based Alarm Prioritization for Building Management Systems

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The issue of alarm fatigue in building management systems (BMS) due to numerous nuisance alarms, leading to missed critical alarms and inefficient manual sorting processes, resulting in high costs and disruptions.

Innovation Solution

An alarm monitoring and evaluation system utilizing machine learning (ML) to analyze alarm data, identify high confidence and nuisance alarms, and provide a unified user interface for prioritized alarm handling, reducing redundant alarms through automated work order generation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual sorting and investigation of alarms is performed, then critical alarms can be identified, but time consumption and labor costs increase significantly

Engineering Contradiction:

Improvealarm identification accuracyVSAvoidtime for alarm sorting

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces manual mechanical sorting of alarms with an automated electronic system that uses machine learning algorithms to classify and prioritize alarms. The system automatically processes alarm data, identifies patterns, and ranks alarms by importance, eliminating the need for manual investigation while maintaining high accuracy in identifying critical alarms.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The alarm management system performs self-service by automatically analyzing and prioritizing alarms without human intervention. The machine learning model continuously learns from alarm patterns and autonomously categorizes new alarms, reducing dependency on manual labor while improving response time and consistency in alarm identification.

Inventive Principle:

Principle #25Self-service

2Reliability

If all alarms are acknowledged and investigated individually, then no critical alarms are missed, but productivity decreases due to alarm fatigue

Engineering Contradiction:

Improvecritical alarm detectionVSAvoidoperational efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent segments the alarm stream into distinct categories based on priority levels, confidence scores, and pattern recognition. Critical alarms are separated from nuisance alarms and presented to operators in a prioritized sequence, allowing them to focus only on high-priority items while maintaining reliable detection of critical events.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different quality levels of alarm processing are applied to different alarm types. High-priority alarms receive full investigative attention, while low-priority nuisance alarms are automatically filtered or grouped. This localized quality approach ensures critical alarms are never missed while improving overall operational productivity by reducing unnecessary investigations.

Inventive Principle:

Principle #3Local quality

3Measurement precision

If alarm thresholds are set tightly to detect all issues, then measurement precision improves, but nuisance alarms increase causing alarm fatigue

Engineering Contradiction:

Improvealarm detection sensitivityVSAvoidnumber of alarms

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The system dynamically adjusts alarm processing based on learned patterns and contextual information. Rather than using fixed thresholds that generate nuisance alarms, the machine learning model adapts its sensitivity based on historical data, equipment status, and operational context, maintaining high detection precision while reducing the quantity of false or low-priority alarms.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The alarm evaluation system combines multiple data sources and analysis methods to create a composite assessment of each alarm's true importance. By integrating pattern recognition, contextual analysis, and confidence scoring, the system achieves high measurement precision in identifying critical issues while filtering out nuisance alarms, effectively reducing the total number of alarms requiring attention.

Inventive Principle:

Principle #40Composite materials

4Productivity

If mass acknowledgment of alarms is performed, then processing speed increases, but loss of information occurs as individual alarm details are not investigated

Engineering Contradiction:

Improvealarm processing speedVSAvoidalarm detail accuracy

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The system performs preliminary analysis and prioritization of alarms before they reach the operator. By pre-processing alarm data, identifying patterns, and ranking items by importance, the system enables operators to quickly acknowledge low-priority alarms while automatically flagging critical ones for detailed investigation, thus maintaining both speed and information accuracy.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250111282A1Alarm monitoring and evaluation systems and methods

Publication Date: 2025.04.03 CBRE INC

AI summary

At least one processor may receive vent data describing a plurality of events related to equipment managed by a building management system. The at least one processor may process the event data using at least one machine learning model. Outputs of the at least one machine learning model may include at least a priority label and a probability score for each respective event in the event data. The at least one processor may generate a user interface within the building management system. The user interface may indicate at least the priority label and the probability score for at least one of the events.