Plant State Monitoring Using Frequency-Domain Fault Estimation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for monitoring production systems lack sufficient detection reliability and precision for predictive error detection, leading to inefficiencies in maintenance planning and increased downtimes.

Innovation Solution

A method involving time-frequency transformation to create a system model based on machine learning, using reference, normal, and disruptive parameters to generate an operating state estimator, which outputs an operating state signal when exceeding a predetermined limit, allowing for precise estimation of system status and maintenance planning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If automated monitoring systems (SCADA) are used to monitor production facilities, then production downtime can be minimized and maintenance can be planned, but detection reliability and precision for predictive error detection are insufficient

Engineering Contradiction:
Improvedetection reliabilityVSAvoidmonitoring system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent replaces traditional mechanical monitoring approaches with signal processing and machine learning techniques. Specifically, it transforms operating parameters from time domain to frequency domain using Fourier transformation, then applies machine learning models to classify operating states, substituting complex mechanical monitoring systems with computational methods that achieve higher detection reliability.

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

Solution Approach 2:

The patent changes the representation parameters of operating data by transforming time-domain signals into frequency-domain spectra. This parameter transformation enables better differentiation between normal and fault states, improving detection reliability while the automated classification reduces the need for complex manual monitoring configurations.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If traditional monitoring methods are used, then system operation can be maintained, but precision for predictive error detection is insufficient leading to increased downtimes

Engineering Contradiction:
Improveerror detection precisionVSAvoidproduction downtime
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent implements preliminary classification of operating states using machine learning models that analyze frequency-domain characteristics before actual faults occur. By training the system to recognize patterns indicative of developing faults, maintenance can be performed proactively, improving detection precision and preventing unplanned downtime through advance warning of potential failures.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent creates spectral copies of operating parameters in the frequency domain that reveal hidden patterns not visible in time-domain data. These spectral representations serve as enhanced copies of the original signals, providing more precise information for detecting early signs of faults and enabling more accurate predictive maintenance timing.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP3734384B1Method and device for monitoring an operational status of a system
Publication Date: 2023.12.20 ETM PROFESSIONAL CONTROL GMBH
  • EP3734384B1 patent drawingFigure 1a~2

AI summary

A method for monitoring the operating state of a plant, comprising the following steps: a) acquiring an operating parameter in a reference operation of the plant as a reference parameter (10), b) acquiring the operating parameter in a normal operation of the plant as a normal parameter (11), c) acquiring the operating parameter in a fault operation of the plant as a fault parameter (12), d) generating and training a plant model (20) based on the principle of machine learning with the reference parameter (10), the normal parameter (11) and the fault parameter (12), e) acquiring the operating parameter in an ongoing operation of the plant as a current parameter (13), f) generating an operating state estimator using the plant model (20) and the current parameter, wherein an operating state signal (30) is output when the operating state estimator exceeds a predetermined limit value.