Rankine cycle analysis using a neural network

Kaufui Wong, Kang Yao

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The Rankine cycle plays a very important role in most steam power plants; much of this analysis needs to be done, especially by undergraduate thermodynamics students. Therefore, it is very necessary to obtain the Rankine cycle efficiencies quickly and accurately. In the present work, we have done this by using a neural network - a branch of artificial intelligence. A wide range of the Rankine cycle efficiencies were calculated using the classic efficiency formula with data excerpted from a Mollier Chart and, a neural network was set up to learn from these known facts. Then, a test data group was chosen to check the predictions made by the trained neural network and, it was found that all the predictions by the trained network matched with the theoretical analysis. The training time was not long and, the prediction accuracy could be improved by adjusting the network operational parameters. After the neural network was trained, tedious calculations and table searching become unnecessary because trained network acts as an expert; if a student inputs the temperature and pressures, the network predicts the efficiency. The whole system is user friendly and easy to understand. Besides, the present work shows that the neural network approach is not only suitable for analysis of simple ideal Rankine cycles, but also could be modified for practical Rankine cycles (i.e. reheat, etc.). In the classroom, the trained network might also be used to find the optimum Rankine cycle operational parameters under constraints of pressure and temperature ranges.

Original languageEnglish
Title of host publicationAmerican Society of Mechanical Engineers, Advanced Energy Systems Division (Publication) AES
Place of PublicationNew York, NY, United States
PublisherPubl by ASME
Pages199-204
Number of pages6
Volume27
ISBN (Print)0791810739
StatePublished - Dec 1 1992
EventWinter Annual Meeting of the American Society of Mechanical Engineers - Anaheim, CA, USA
Duration: Nov 8 1992Nov 13 1992

Other

OtherWinter Annual Meeting of the American Society of Mechanical Engineers
CityAnaheim, CA, USA
Period11/8/9211/13/92

Fingerprint

Rankine cycle
Neural networks
Students
Steam power plants
Artificial intelligence
Thermodynamics
Temperature

ASJC Scopus subject areas

  • Energy Engineering and Power Technology
  • Mechanical Engineering

Cite this

Wong, K., & Yao, K. (1992). Rankine cycle analysis using a neural network. In American Society of Mechanical Engineers, Advanced Energy Systems Division (Publication) AES (Vol. 27, pp. 199-204). New York, NY, United States: Publ by ASME.

Rankine cycle analysis using a neural network. / Wong, Kaufui; Yao, Kang.

American Society of Mechanical Engineers, Advanced Energy Systems Division (Publication) AES. Vol. 27 New York, NY, United States : Publ by ASME, 1992. p. 199-204.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Wong, K & Yao, K 1992, Rankine cycle analysis using a neural network. in American Society of Mechanical Engineers, Advanced Energy Systems Division (Publication) AES. vol. 27, Publ by ASME, New York, NY, United States, pp. 199-204, Winter Annual Meeting of the American Society of Mechanical Engineers, Anaheim, CA, USA, 11/8/92.
Wong K, Yao K. Rankine cycle analysis using a neural network. In American Society of Mechanical Engineers, Advanced Energy Systems Division (Publication) AES. Vol. 27. New York, NY, United States: Publ by ASME. 1992. p. 199-204
Wong, Kaufui ; Yao, Kang. / Rankine cycle analysis using a neural network. American Society of Mechanical Engineers, Advanced Energy Systems Division (Publication) AES. Vol. 27 New York, NY, United States : Publ by ASME, 1992. pp. 199-204
@inproceedings{52b737f747c64ce0944ecbf2f455dd3b,
title = "Rankine cycle analysis using a neural network",
abstract = "The Rankine cycle plays a very important role in most steam power plants; much of this analysis needs to be done, especially by undergraduate thermodynamics students. Therefore, it is very necessary to obtain the Rankine cycle efficiencies quickly and accurately. In the present work, we have done this by using a neural network - a branch of artificial intelligence. A wide range of the Rankine cycle efficiencies were calculated using the classic efficiency formula with data excerpted from a Mollier Chart and, a neural network was set up to learn from these known facts. Then, a test data group was chosen to check the predictions made by the trained neural network and, it was found that all the predictions by the trained network matched with the theoretical analysis. The training time was not long and, the prediction accuracy could be improved by adjusting the network operational parameters. After the neural network was trained, tedious calculations and table searching become unnecessary because trained network acts as an expert; if a student inputs the temperature and pressures, the network predicts the efficiency. The whole system is user friendly and easy to understand. Besides, the present work shows that the neural network approach is not only suitable for analysis of simple ideal Rankine cycles, but also could be modified for practical Rankine cycles (i.e. reheat, etc.). In the classroom, the trained network might also be used to find the optimum Rankine cycle operational parameters under constraints of pressure and temperature ranges.",
author = "Kaufui Wong and Kang Yao",
year = "1992",
month = "12",
day = "1",
language = "English",
isbn = "0791810739",
volume = "27",
pages = "199--204",
booktitle = "American Society of Mechanical Engineers, Advanced Energy Systems Division (Publication) AES",
publisher = "Publ by ASME",

}

TY - GEN

T1 - Rankine cycle analysis using a neural network

AU - Wong, Kaufui

AU - Yao, Kang

PY - 1992/12/1

Y1 - 1992/12/1

N2 - The Rankine cycle plays a very important role in most steam power plants; much of this analysis needs to be done, especially by undergraduate thermodynamics students. Therefore, it is very necessary to obtain the Rankine cycle efficiencies quickly and accurately. In the present work, we have done this by using a neural network - a branch of artificial intelligence. A wide range of the Rankine cycle efficiencies were calculated using the classic efficiency formula with data excerpted from a Mollier Chart and, a neural network was set up to learn from these known facts. Then, a test data group was chosen to check the predictions made by the trained neural network and, it was found that all the predictions by the trained network matched with the theoretical analysis. The training time was not long and, the prediction accuracy could be improved by adjusting the network operational parameters. After the neural network was trained, tedious calculations and table searching become unnecessary because trained network acts as an expert; if a student inputs the temperature and pressures, the network predicts the efficiency. The whole system is user friendly and easy to understand. Besides, the present work shows that the neural network approach is not only suitable for analysis of simple ideal Rankine cycles, but also could be modified for practical Rankine cycles (i.e. reheat, etc.). In the classroom, the trained network might also be used to find the optimum Rankine cycle operational parameters under constraints of pressure and temperature ranges.

AB - The Rankine cycle plays a very important role in most steam power plants; much of this analysis needs to be done, especially by undergraduate thermodynamics students. Therefore, it is very necessary to obtain the Rankine cycle efficiencies quickly and accurately. In the present work, we have done this by using a neural network - a branch of artificial intelligence. A wide range of the Rankine cycle efficiencies were calculated using the classic efficiency formula with data excerpted from a Mollier Chart and, a neural network was set up to learn from these known facts. Then, a test data group was chosen to check the predictions made by the trained neural network and, it was found that all the predictions by the trained network matched with the theoretical analysis. The training time was not long and, the prediction accuracy could be improved by adjusting the network operational parameters. After the neural network was trained, tedious calculations and table searching become unnecessary because trained network acts as an expert; if a student inputs the temperature and pressures, the network predicts the efficiency. The whole system is user friendly and easy to understand. Besides, the present work shows that the neural network approach is not only suitable for analysis of simple ideal Rankine cycles, but also could be modified for practical Rankine cycles (i.e. reheat, etc.). In the classroom, the trained network might also be used to find the optimum Rankine cycle operational parameters under constraints of pressure and temperature ranges.

UR - http://www.scopus.com/inward/record.url?scp=0026987602&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0026987602&partnerID=8YFLogxK

M3 - Conference contribution

SN - 0791810739

VL - 27

SP - 199

EP - 204

BT - American Society of Mechanical Engineers, Advanced Energy Systems Division (Publication) AES

PB - Publ by ASME

CY - New York, NY, United States

ER -