0
Research Papers: Gas Turbines: Turbomachinery

Robust Fault Detection to Determine Compressor Surge Point Via Dynamic Neural Network-Based Subspace Identification Technique

[+] Author and Article Information
Sayyid Mahdi Alavinia

Department of Electrical and
Robotic Engineering,
Shahrood University of Technology,
Shahrood 3619995161, Iran
e-mail: zalloi@shahroodut.ac.ir

Mohammad Ali Sadrnia

Department of Electrical and
Robotic Engineering,
Shahrood University of Technology,
Shahrood 3619995161, Iran
e-mail: masadrnia@shahroodut.ac.ir

Mohammad Javad Khosrowjerdi

Department of Electrical Engineering,
Sahand University of Technology,
Tabriz 3619995161, Iran
e-mail: khosrowjerdi@sut.ac.ir

Mohammad Mehdi Fateh

Department of Electrical and
Robotic Engineering,
Shahrood University of Technology,
Shahrood 3619995161, Iran
e-mail: mmfateh@shahroodut.ac.ir

1Corresponding author.

Contributed by the Turbomachinery Committee of ASME for publication in the JOURNAL OF ENGINEERING FOR GAS TURBINES AND POWER. Manuscript received November 27, 2013; final manuscript received January 26, 2014; published online February 28, 2014. Editor: David Wisler.

J. Eng. Gas Turbines Power 136(8), 082602 (Feb 28, 2014) (8 pages) Paper No: GTP-13-1440; doi: 10.1115/1.4026610 History: Received November 27, 2013; Revised January 26, 2014

In this paper, a dynamic neural network (DNN) based on robust identification scheme is presented to determine compressor surge point accurately using sensor fault detection (FD). The main innovation of this paper is to present different and complementary technique for surge suppressing studies within sensor FD. The proposed method aims to utilize the embedded analytical redundancies for sensor FD, even in the presence of uncertainty in the compressor and sensor noise. The robust dynamic neural network is developed to learn the input–output map of the compressor for residual generation and the required data is obtained from the compressor Moore–Greitzer simulated model. Generally, the main drawback of DNN method is the lack of systematic law for selecting of initial Hurwitz matrix. Therefore, the subspace identification method is proposed for selecting this matrix. A number of simulation studies are carried out to demonstrate the advantages, capabilities, and performance of our proposed FD scheme and a worthwhile direction for future research is also presented.

FIGURES IN THIS ARTICLE
<>
Copyright © 2014 by ASME
Your Session has timed out. Please sign back in to continue.

References

Figures

Grahic Jump Location
Fig. 1

Fault detection based on analytical redundancy

Grahic Jump Location
Fig. 2

Compressor performance curve

Grahic Jump Location
Fig. 3

Dynamic neural network schematic

Grahic Jump Location
Fig. 4

DNN based on robust property schematic

Grahic Jump Location
Fig. 5

The residual generation schematic via RDNN

Grahic Jump Location
Fig. 6

The fitness of actual and identified signals

Grahic Jump Location
Fig. 7

Mean squared error (MSE)

Grahic Jump Location
Fig. 8

DNN identifier and compressor outputs

Grahic Jump Location
Fig. 9

RDNN identifier and compressor outputs in the healthy scenario

Grahic Jump Location
Fig. 10

RDNN identifier and compressor outputs in the faulty scenario

Grahic Jump Location
Fig. 11

Residual signal in the RDNN identifier

Grahic Jump Location
Fig. 12

Residual signal and threshold

Grahic Jump Location
Fig. 13

y, y∧ and residual outputs in sensor cut off scenario

Tables

Errata

Discussions

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In