BMS Alarm Classification Using Dual AI Consensus Filtering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Building Management Systems often generate false alarms, diverting operator attention and wasting time, as existing systems lack effective methods to automatically distinguish between true and false alarms.

Innovation Solution

Implementing a dual AI model system comprising a false alarm AI Model and a true alarm AI Model to classify incoming alarms, with automatic classification based on consensus or manual override when models disagree, and continuous training on databases of classified alarms.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a Building Management System raises alarms for detected abnormalities, then operator awareness of potential issues is improved, but operator time and attention are wasted due to false alarms

Engineering Contradiction:
Improvealarm accuracyVSAvoidoperator response time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

An AI classification system is introduced as an intermediary between the alarm detection mechanism and the operator. This intermediary automatically analyzes incoming alarms, compares them against historical data and patterns, and classifies them as true or false alarms before presenting them to operators, thereby filtering out false alarms and preventing operator time waste

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs self-diagnosis and self-classification of alarms using trained AI models that automatically evaluate alarm characteristics, metadata, and historical patterns to determine whether an alarm is genuine or false without requiring immediate operator intervention for every alarm

Inventive Principle:
Principle #25Self-service

2Measurement precision

If manual alarm classification is used, then accurate distinction between true and false alarms is achieved, but operator workload and response time are increased

Engineering Contradiction:
Improvealarm classification accuracyVSAvoidoperator efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system creates digital copies of historical alarm data and operational patterns, then uses AI models to analyze and compare incoming alarms against these copied historical patterns. This allows automated classification without requiring operators to manually review every alarm, maintaining high accuracy while improving efficiency

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The manual mechanical process of operator review and classification is replaced with an automated electronic AI-based classification system. The AI models process alarms electronically, substituting human cognitive effort with automated algorithms that can rapidly classify alarms with consistent accuracy

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If automated AI classification is implemented, then operator time is saved by filtering false alarms, but system complexity increases

Engineering Contradiction:
Improvealarm processing efficiencyVSAvoidsystem architecture complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The alarm classification function is segmented into separate AI models - one specialized for detecting false alarms and another for detecting true alarms. These segmented models can be independently trained, deployed, and maintained, reducing the complexity of a single monolithic classification system while improving overall processing efficiency

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4700737A1Systems and methods for mitigating false alarms in a building management system
Publication Date: 2026.02.25 HONEYWELL INTERNATIONAL INC
  • EP4700737A1 patent drawingFigure 1
  • EP4700737A1 patent drawingFigure 2A
  • EP4700737A1 patent drawingFigure 2B

AI summary

A false alarm Artificial Intelligence, AI, Model (14) is trained using metadata associated with alarms classified as false alarms and a true alarm Artificial Intelligence, AI, Model (18) is trained using metadata associated with alarms classified as true alarms. An incoming alarm is received. The false alarm AI Model and the true alarm AI Model are both applied to the incoming alarm and both models classify the incoming alarm as either a false alarm classification or a true alarm classification. When the false alarm AI model and the true alarm AI Model agree, the incoming alarm is automatically classified accordingly. When the models do not agree, the incoming alarm is presented to an operator console (22) of the BMS, and a manual classification of the incoming alarm is received from the operator console.