Hybrid analytical/neural network model of variable displacement pump dynamics

J M McNamara, K A Edge, N D Vaughan

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

The response of a variable displacement pump to a control signal can be an important factor in the overall system response. It is therefore necessary in simulating such systems to have an accurate model of the dynamic behavior of the pump and its interaction with the system. This paper presents a comprehensive model of a typical variable displacement swash-plate piston pump and its servo controller. Because of the complex and non-linear form of the pumping dynamics previous researchers have used over-simplified models, particularly in respect of swash-plate forces during the pumping cycle. This paper describes a hybrid approach which combines physical models and a neural network to simulate swash-plate dynamics. Training data for the neural network is obtained through a detailed simulation of pumping dynamics. This allows the creation of a black-box model of swash-plate moments as a function of delivery pressure and swash-plate position and velocity. Training data and the hybrid model of the pump system have been verified experimentally. The final pump system model takes into account the interactions between pump, its controller and the delivery line
Original languageEnglish
Pages (from-to)71-76
Number of pages6
JournalAmerican Society of Mechanical Engineers, The Fluid Power and Systems Technology Division (Publication) FPST
Volume4
Publication statusPublished - 1997

Fingerprint

Pumps
Neural networks
Reciprocating pumps
Controllers

Cite this

@article{e26ce54a52e245b4b147f07cc4c0a28a,
title = "Hybrid analytical/neural network model of variable displacement pump dynamics",
abstract = "The response of a variable displacement pump to a control signal can be an important factor in the overall system response. It is therefore necessary in simulating such systems to have an accurate model of the dynamic behavior of the pump and its interaction with the system. This paper presents a comprehensive model of a typical variable displacement swash-plate piston pump and its servo controller. Because of the complex and non-linear form of the pumping dynamics previous researchers have used over-simplified models, particularly in respect of swash-plate forces during the pumping cycle. This paper describes a hybrid approach which combines physical models and a neural network to simulate swash-plate dynamics. Training data for the neural network is obtained through a detailed simulation of pumping dynamics. This allows the creation of a black-box model of swash-plate moments as a function of delivery pressure and swash-plate position and velocity. Training data and the hybrid model of the pump system have been verified experimentally. The final pump system model takes into account the interactions between pump, its controller and the delivery line",
author = "McNamara, {J M} and Edge, {K A} and Vaughan, {N D}",
year = "1997",
language = "English",
volume = "4",
pages = "71--76",
journal = "American Society of Mechanical Engineers, The Fluid Power and Systems Technology Division (Publication) FPST",

}

TY - JOUR

T1 - Hybrid analytical/neural network model of variable displacement pump dynamics

AU - McNamara, J M

AU - Edge, K A

AU - Vaughan, N D

PY - 1997

Y1 - 1997

N2 - The response of a variable displacement pump to a control signal can be an important factor in the overall system response. It is therefore necessary in simulating such systems to have an accurate model of the dynamic behavior of the pump and its interaction with the system. This paper presents a comprehensive model of a typical variable displacement swash-plate piston pump and its servo controller. Because of the complex and non-linear form of the pumping dynamics previous researchers have used over-simplified models, particularly in respect of swash-plate forces during the pumping cycle. This paper describes a hybrid approach which combines physical models and a neural network to simulate swash-plate dynamics. Training data for the neural network is obtained through a detailed simulation of pumping dynamics. This allows the creation of a black-box model of swash-plate moments as a function of delivery pressure and swash-plate position and velocity. Training data and the hybrid model of the pump system have been verified experimentally. The final pump system model takes into account the interactions between pump, its controller and the delivery line

AB - The response of a variable displacement pump to a control signal can be an important factor in the overall system response. It is therefore necessary in simulating such systems to have an accurate model of the dynamic behavior of the pump and its interaction with the system. This paper presents a comprehensive model of a typical variable displacement swash-plate piston pump and its servo controller. Because of the complex and non-linear form of the pumping dynamics previous researchers have used over-simplified models, particularly in respect of swash-plate forces during the pumping cycle. This paper describes a hybrid approach which combines physical models and a neural network to simulate swash-plate dynamics. Training data for the neural network is obtained through a detailed simulation of pumping dynamics. This allows the creation of a black-box model of swash-plate moments as a function of delivery pressure and swash-plate position and velocity. Training data and the hybrid model of the pump system have been verified experimentally. The final pump system model takes into account the interactions between pump, its controller and the delivery line

M3 - Article

VL - 4

SP - 71

EP - 76

JO - American Society of Mechanical Engineers, The Fluid Power and Systems Technology Division (Publication) FPST

JF - American Society of Mechanical Engineers, The Fluid Power and Systems Technology Division (Publication) FPST

ER -