Anomaly Detection Analysis for Identifying Fault-Causing Variables

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current anomaly detection methods, such as PCA, fail to identify specific anomalous variables causing anomalies in data samples, leading to ambiguous results and misidentification due to dimensionality reduction and limited applicability to other models.

Innovation Solution

The method involves using a combination of anomaly detection models like PCA, independent component analysis, or autoencoders with imputation models to iteratively process data samples, identifying anomalous variables through exhaustive or greedy algorithms, which can be applied to any anomaly detection model, reducing computational intensity and improving accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If PCA model is used for anomaly detection, then anomaly detection capability is improved, but ability to identify specific anomalous variables deteriorates

Engineering Contradiction:

Improveanomaly detection capabilityVSAvoididentification of anomalous variables

Core Design Contradiction:

ReliabilityVSLoss of information

Solution Approach 1:

The patent segments the anomaly detection process into two distinct phases: first detecting anomalies using PCA, then identifying specific anomalous variables through a separate variable contribution analysis step. This segmentation allows each phase to specialize - PCA handles detection while the contribution analysis handles identification, resolving the contradiction between detection capability and variable identification.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary mechanism (variable contribution analysis) that bridges the gap between anomaly detection and anomalous variable identification. This intermediary step processes the anomaly scores and variable relationships to identify which specific variables are responsible, without compromising the original PCA detection capability.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Loss of information

If contribution analysis method is used to identify anomalous variables, then variable identification is attempted, but result accuracy deteriorates due to PCA dimensionality reduction

Engineering Contradiction:

Improvevariable identification capabilityVSAvoidaccuracy of anomalous variable identification

Core Design Contradiction:

Loss of informationVSMeasurement precision

Solution Approach 1:

The patent performs preliminary action by preserving the original variable-space relationships before dimensionality reduction occurs. The variable contribution analysis is designed to work with the PCA-transformed data while accounting for the dimensionality reduction effects, allowing accurate identification despite the loss of information in the transformation process.

Inventive Principle:

Principle #10Preliminary action

3Loss of information

If contribution analysis method is used, then anomalous variables can be identified, but method versatility deteriorates due to PCA dependency

Engineering Contradiction:

Improvevariable identification capabilityVSAvoidapplicability to different anomaly detection models

Core Design Contradiction:

Loss of informationVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal variable contribution analysis framework that can work with multiple types of anomaly detection models, not just PCA. The method is designed to be model-agnostic, allowing it to be applied to different dimensionality reduction and anomaly detection techniques while maintaining its ability to identify anomalous variables.

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Measurement precision

If exhaustive variable combination processing is used, then identification accuracy is improved, but computational complexity increases

Engineering Contradiction:

Improveidentification accuracyVSAvoidcomputational complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies partial action by processing variable combinations selectively rather than exhaustively. The method prioritizes processing of variable combinations that are more likely to contain anomalous variables based on initial anomaly scores and variable relationships, achieving good identification accuracy without the full computational burden of exhaustive processing of all possible combinations.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentEP4062246B1Measurement result analysis by anomaly detection and identification of anomalous variables

Publication Date: 2023.11.01 ELISA OYJ

Figure 1~2
Figure 3A
Figure 3B

AI summary

A computer implemented method of analyzing measurement results of a target system, such as an industrial process or communication network. The method comprises receiving (301) a data sample comprising a plurality of variables representing the measurement results, detecting (302) that the data sample is an anomalous sample using an anomaly detection model, processing (303) the data sample by applying an imputation model to selected subsets of variables of the data sample to obtain imputed samples, and applying the anomaly detection model to the imputed samples, determining (304) anomalous variables of the data sample based on results from the processing of the data sample, and outputting (305) the anomalous variables of the data sample for management operations in the target system.