SVM-Based Device State Classification for Fault Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Modern devices with numerous sensors face challenges in accurately determining their operational state and identifying fault causes due to complex relationships between operating parameters, often obscuring relevant values within a large number of measurements.

Innovation Solution

A system utilizing a support vector machine trained with labeled operating points in an n-dimensional parameter space to classify device states, generating boundaries between normal and error states, and predicting future error occurrences by analyzing operating parameter trajectories.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If a large number of sensors and operating parameters are used to monitor device state, then measurement completeness is improved, but measurement precision deteriorates because relevant parameter values become hidden among many measurements

Engineering Contradiction:
Improvenumber of operating parametersVSAvoiderror detection accuracy
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent extracts and isolates the most relevant operating parameters from the large set of measurements using support vector machine analysis. The SVM identifies support vectors and normal vectors that highlight critical parameters, separating them from the noise of irrelevant measurements, thus restoring measurement precision without reducing the quantity of collected data.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms the raw operating parameters into a new parameter space using support vector machines. By projecting the high-dimensional parameter space onto a lower-dimensional subspace defined by normal vectors, the system changes the representation of parameters to make relevant information more distinguishable, thereby improving measurement precision while maintaining comprehensive monitoring.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If manual analysis by experts is used to determine device state and fault causes, then analysis accuracy is improved, but device complexity and time consumption increase

Engineering Contradiction:
Improvestate determination accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements self-service by enabling the system to automatically determine device states and identify fault causes without requiring expert intervention. The support vector machine performs automated classification and analysis of operating parameters, replacing manual expert analysis with an autonomous computational system that maintains high accuracy while reducing complexity and time requirements.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent substitutes the mechanical process of manual expert analysis with a computational machine learning system. The support vector machine algorithm automatically processes operating parameters, generates classification boundaries, and identifies fault causes, replacing the need for human experts to manually superimpose curve representations and analyze data, thereby reducing system complexity and operational time.

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

3Measurement precision

If support vector machine classification is implemented, then state determination precision is improved, but device complexity increases due to the need for training and parameter processing

Engineering Contradiction:
Improveclassification accuracyVSAvoidalgorithm complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by performing offline training of the support vector machine with labeled training data before actual device operation. The classification boundaries and support vectors are pre-computed during the training phase, so that during runtime, the system only needs to evaluate new operating points against the pre-established boundaries, significantly reducing the computational complexity during actual state determination while maintaining high classification accuracy.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3811170B1Determining of states of a device using support vector machines
Publication Date: 2022.07.06 SIEMENS AG
  • EP3811170B1 patent drawingFigure 1~2
  • EP3811170B1 patent drawingFigure 3
  • EP3811170B1 patent drawingFigure 4

AI summary

The invention relates to a system and to a method for determining a state of a device by means of a trained support-vector machine. According to the invention, an operating parameter space is divided into classification volumes, at least one of which indicates a normal state and at least one other of which indicates a fault state of the device. A current state of the device can therefore be determined by determining where a current operating parameter point is to be arranged in the operating parameter space. The invention further relates to methods and to variants of the system in order to facilitate a cause evaluation and to determine particularly relevant operating parameters for the fault determination.