Multivariate Time-Series Event Classification for Rare Asset Patterns

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Traditional data analytics and digital transformation in industrial environments require significant human interaction and are inefficient, with limited time spent on modeling, leading to suboptimal use of computing resources and challenges in data-driven automatic labeling of events, particularly rare events in multivariate data.

Innovation Solution

A system employing pattern recognition in multivariate time series data for event classification, utilizing data preprocessing, data augmentation, machine learning classifiers, and user interfaces for labeling and retraining, to provide real-time insights and dashboard visualizations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional data analytics methods are used with human interaction, then modeling can be performed, but computing resource efficiency deteriorates and time consumption increases

Engineering Contradiction:

Improvedata analytics efficiencyVSAvoidtime spent on modeling

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system enables automatic event classification through machine learning models that autonomously analyze multivariate time series data without requiring human experts for each classification task. The models self-train on historical data and automatically label events, freeing human resources for higher-value activities while dramatically reducing modeling time.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

Manual engineering analysis and human expert interaction are replaced with automated machine learning systems. The mechanical process of human data analysis is substituted with computational algorithms that process multivariate time series data efficiently, improving productivity while reducing time loss.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If manual event classification is performed, then accuracy can be maintained, but computing resource requirements increase and scalability decreases

Engineering Contradiction:

Improveevent classification accuracyVSAvoidcomputing resource requirements

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the classification task into distinct machine learning models trained for specific event types. Each model specializes in particular patterns within multivariate time series data, maintaining high classification accuracy while distributing computational load efficiently across multiple specialized components rather than requiring one complex general-purpose system.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system optimizes computing resource usage by dynamically adjusting model parameters and processing configurations based on data characteristics. Machine learning models adapt their computational requirements based on the specific patterns being analyzed, maintaining accuracy while reducing unnecessary resource consumption through parameter optimization.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If extensive modeling is performed to classify rare events, then classification accuracy improves, but time consumption and resource usage increase

Engineering Contradiction:

Improverare event classification accuracyVSAvoidmodeling throughput

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The system performs preliminary actions by pre-training machine learning models on historical multivariate time series data including rare events. This advance preparation creates ready-to-use classification capabilities that can quickly and accurately identify rare events in real-time without requiring extensive modeling efforts when events occur, thus maintaining accuracy while improving productivity.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The machine learning models continuously learn and refine their classification capabilities through ongoing training on new data. This continuous useful action maintains high accuracy for rare event classification while improving the system's ability to detect and classify rare events efficiently over time, increasing productivity without sacrificing precision.

Inventive Principle:

Principle #20Continuity of useful action

Data Source

PatentUS12554559B2Classification of events by pattern recognition in multivariate time series data

Publication Date: 2026.02.17 HONEYWELL INTERNATIONAL INC

AI summary

Various embodiments described herein relate to classification of events by pattern recognition in multivariate time series data associated with one or more assets. In this regard, a request to classify events associated with one or more assets is received. The request includes an asset descriptor describing the one or more assets. In response to the request, aggregated multivariate data associated with the one or more assets is obtained based on the asset descriptor. Furthermore, one or more events associated with the aggregated multivariate data is labeled based respective defined data signatures for respective defined events associated with a defined event attribute. A dashboard visualization is then provided to an electronic interface of a computing device. The dashboard visualization includes data associated with the one or more events.