Abnormality Detection Model Generation Using Feature Importance Ranking

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing abnormality detection systems face challenges in accurately predicting anomalies in machines or equipment before actual operation, as they often lack sufficient abnormality/normality information for training models, leading to suboptimal predictive maintenance.

Innovation Solution

An abnormality detection system that includes a control operation part, an abnormality detection part, a state value storage part, and a model generation part. The model generation part generates feature values from collected state values, calculates their importance levels using multiple methods, and integrates these levels to determine rankings, enabling the creation of a high-accuracy abnormality detection model before actual operation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional abnormality detection methods are used, then the system can operate with simple data collection, but the detection accuracy is insufficient for predicting abnormalities before actual operation

Engineering Contradiction:

Improveabnormality detection accuracyVSAvoidmodel generation complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs preliminary data collection and model generation before actual operation begins. The model generation part creates the abnormality detection model using historical data and multiple evaluation methods in advance, so that when operation starts, the model is already ready to provide high-accuracy predictions without needing complex real-time processing

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The model generation part acts as an intermediary between raw state values and abnormality detection. It introduces multiple evaluation methods (information amount, variation, correlation) as intermediate processing steps that transform raw data into meaningful features, improving detection accuracy without requiring the end system to handle complex analysis directly

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If multiple evaluation methods are used to calculate importance levels, then the reliability of feature selection improves, but the computational burden increases

Engineering Contradiction:

Improvefeature selection reliabilityVSAvoidcomputational energy consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system uses multiple evaluation methods (information amount, variation, correlation) to calculate importance levels, which may seem excessive but ensures reliable feature selection. The model generation part performs these calculations once during model creation, not continuously during operation, so the computational burden is concentrated in the preparation phase rather than affecting ongoing energy consumption

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The system replaces manual or simple heuristic feature selection with automated computational evaluation using multiple mathematical methods. This substitution increases initial computational requirements but eliminates the need for continuous manual adjustment and ensures consistent, reliable feature selection based on objective criteria

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If feature values are generated from collected state values, then the model can capture complex patterns, but the processing time increases

Engineering Contradiction:

Improvepattern recognition accuracyVSAvoidmodel generation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The model generation part performs feature value generation and pattern analysis in advance before actual operation. By pre-processing the state values and generating comprehensive feature sets during the model creation phase, the system captures complex patterns without adding processing time to real-time operation

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The model generation process is segmented into distinct steps: collecting state values, generating feature values from those state values, calculating importance levels using multiple methods, and selecting final features. This segmentation allows each step to be optimized independently and enables parallel processing where applicable, reducing overall processing time

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS10795338B2Abnormality detection system, support device, and model generation method

Publication Date: 2020.10.06 OMRON CORP

AI summary

An abnormality detection system, support device, and model generation method for generating a more highly accurate abnormality detection model before an actual operation are provided. A model generation part includes a section for generating feature values from state values provided from a state value storage part; a section for calculating importance levels respectively for the generated feature values based on plural methods, wherein the importance levels indicating a degree that is effective for abnormality detection; and a section for integrating the importance levels calculated based on the plural methods for each of the generated feature values and determining rankings of the importance levels of the generated feature values.