Compressor Anomaly Prediction via Permutation Entropy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Gas turbine systems experience anomalies like stall, surge, and instability in the compressor due to wear and tear, leading to decreased efficiency and costly maintenance, as existing prediction methods are inadequate for timely recognition and prevention of these issues.

Innovation Solution

A system utilizing pressure sensors to generate high-speed time-series signals between compressor blade tips and the casing, which are processed to determine permutation entropy patterns, allowing for prediction and categorization of anomalies, enabling proactive corrective actions to minimize or avoid these issues.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional monitoring methods are used for compressor anomalies, then the system structure remains simple, but the anomaly detection precision and timeliness are insufficient

Engineering Contradiction:
Improveanomaly detection precisionVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces traditional mechanical monitoring methods with signal processing and permutation entropy analysis. Pressure signals are transformed into permutation entropy values that quantify signal complexity, enabling precise anomaly detection without complex mechanical sensors or invasive measurements

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

Solution Approach 2:

The patent transforms pressure signals into permutation entropy parameters to detect anomalies. By changing the parameter from raw pressure values to permutation entropy (which measures signal complexity), the system achieves sensitive anomaly detection while maintaining relatively simple implementation

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If permutation entropy analysis is implemented for anomaly prediction, then the anomaly detection precision improves, but the computational complexity increases

Engineering Contradiction:
Improveanomaly prediction accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies permutation entropy analysis selectively to pressure signals from critical compressor stages rather than processing all sensor data comprehensively. This partial application achieves effective anomaly detection while limiting computational complexity to necessary calculations only

Inventive Principle:
Principle #16Partial or excessive action

3Reliability

If continuous monitoring of compressor pressure signals is performed, then the reliability of anomaly prediction improves, but the energy consumption increases

Engineering Contradiction:
Improveanomaly prediction reliabilityVSAvoidenergy consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent implements continuous monitoring of pressure signals through the compressor system, maintaining reliable anomaly detection capability. The permutation entropy calculation processes signals continuously to provide ongoing reliability assessment while energy consumption is managed through efficient signal processing algorithms

Inventive Principle:
Principle #20Continuity of useful action

4Loss of information

If multiple pattern categories are analyzed for anomaly classification, then the information completeness improves, but the difficulty of detecting and measuring increases

Engineering Contradiction:
Improveanomaly information completenessVSAvoidpattern classification difficulty
Core Design Contradiction:
Loss of informationVSDifficulty of detecting and measuring

Solution Approach 1:

The patent segments anomaly detection into multiple pattern categories (normal operation, developing anomaly, critical anomaly). Each category has distinct permutation entropy characteristics, allowing comprehensive information capture while simplifying classification through clear threshold-based differentiation

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3382208B1Systems and methods for compressor anomaly prediction
Publication Date: 2023.07.26 GENERAL ELECTRIC TECH GMBH
  • EP3382208B1 patent drawingFigure 1
  • EP3382208B1 patent drawingFigure 2
  • EP3382208B1 patent drawingFigure 3

AI summary

A controller comprising a non-transitory computer-readable storage medium (60) storing one or more processor-executable instructions wherein the one or more instructions, when executed by a processor (58) of the controller (56), cause acts to be performed including receiving (152) signals (100) representative of pressure between respective compressor blade (80) tips and a casing (25) of a compressor (24) at one or more stages (82), generating (154) multiple patterns based on a permutation entropy window (254) and the signals, identifying (156) multiple pattern categories (404, 454) in the multiple patterns, determining (158) a permutation entropy based on the multiple patterns and the multiple pattern categories (404, 454), predicting (160) an anomaly (180, 194) in the compressor (24) based on the permutation entropy, comparing (162) the multiple pattern categories (404, 454) to determined permutations (110) of pattern categories (404, 454) when an anomaly (180, 194) is present in the compressor (24), predicting (164) a category of the anomaly (180, 194) based on the comparison of the multiple pattern categories (404, 454) to the determined permutation (110) of pattern categories (404, 454) and determining and executing (166) a corrective action.