BMS Vibration Analytics for HVAC Predictive Maintenance

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing building management systems (BMS) rely on manual initiation and rule-based approaches for analyzing vibration data from HVAC equipment, which are inefficient and do not utilize large amounts of historical data to fine-tune advanced models, leading to suboptimal maintenance and performance.

Innovation Solution

A method involving training machine learning models with historical data for various components of HVAC equipment, using vibration datasets that include metadata and time waveforms, to generate condition scores and automatically schedule maintenance or adjust setpoints based on these scores.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If manual initiation and rule-based approaches are used for analyzing vibration data, then the system is simple to implement, but the analysis efficiency and accuracy are insufficient

Engineering Contradiction:
Improveanalysis efficiencyVSAvoidsystem complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent replaces manual initiation and rule-based approaches with automated machine learning models that process vibration data. The system uses trained ML models to automatically analyze vibration datasets, extract features, and generate condition scores without manual intervention, thereby improving analysis efficiency while managing complexity through automation.

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

Solution Approach 2:

The system enables self-service through automated maintenance scheduling. The machine learning models autonomously analyze vibration data, assess component conditions, and schedule maintenance events without human intervention. This self-service capability improves productivity by eliminating manual analysis while the modular architecture manages system complexity.

Inventive Principle:
Principle #25Self-service

2Measurement precision

If large amounts of historical data are used to fine-tune advanced models, then the measurement precision and reliability improve, but the data processing time and computational resources increase

Engineering Contradiction:
Improvecondition assessment accuracyVSAvoiddata processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-training machine learning models with historical vibration data before deployment. The models are trained offline on extensive historical datasets to learn patterns and relationships, then deployed for real-time or near-real-time analysis. This approach achieves high measurement precision through comprehensive historical data utilization while minimizing data processing time during actual operation.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system dynamically adapts to different data processing needs by selecting appropriate analysis depths. The machine learning models can operate in different modes depending on available computational resources and time constraints, adjusting the level of detail in analysis while maintaining acceptable accuracy. This dynamic approach balances measurement precision with data processing time.

Inventive Principle:
Principle #15Dynamics

3Reliability

If automated machine learning models are deployed for vibration analysis, then maintenance optimization and asset lifetime extension improve, but the device complexity and implementation difficulty increase

Engineering Contradiction:
Improvemaintenance optimizationVSAvoidimplementation complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the vibration analysis system into distinct functional modules: data collection, feature extraction, machine learning model application, condition scoring, and maintenance scheduling. Each module performs a specific function and can be independently developed, tested, and deployed. This segmentation reduces implementation complexity while enabling automated maintenance optimization through coordinated operation of specialized components.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system introduces an intermediary layer of machine learning models that bridge raw vibration data and maintenance decisions. These models translate complex vibration patterns into interpretable condition scores and maintenance recommendations, making the system more manageable and easier to implement while achieving reliable maintenance optimization.

Inventive Principle:
Principle #24Intermediary (Mediator)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach enables automated, data-driven maintenance and performance optimization of HVAC equipment, improving insight, performance, and extending the lifetime of building assets by leveraging advanced analytics and machine learning techniques.

Implementation Method 1

vibration data contained in the vibration dataset is collected using a triaxial accelerometer

Methodology Applied
Scientific EffectVibration: Vibration

Implementation Method 2

calculating a fast Fourier transform (FFT) of each of the time waveforms, and calculating a machine speed based on a frequency domain information obtained from the FFT

Methodology Applied
Scientific EffectFast Fourier Transform:

Data Source

PatentEP3701338B1Building management system with automated vibration data analysis
Publication Date: 2024.02.14 JOHNSON CONTROLS TYCO IP HLDG LLP
  • EP3701338B1 patent drawingFigure 1
  • EP3701338B1 patent drawingFigure 2
  • EP3701338B1 patent drawingFigure 3

AI summary

A method performed by a Building Management System (BMS) includes receiving a vibration dataset associated with a machine controlled by the BMS, The vibration dataset includes machine metadata, machine operating conditions, and one or more time waveforms. The method further includes assembling a feature vector comprising one or more features of the vibration dataset for input to a machine learning model. The machine learning model is associated with a component of the machine. The method further includes applying the feature vector to the machine learning model in order to generate a condition score for the component. The method further includes scheduling a maintenance event for the machine or changing a setpoint associated with the machine depending on the condition score.