Signal Propagation on a Railway Wireless Condition Monitoring System

7/26/2019 Signal Propagation on a Railway Wireless Condition Monitoring System

1/17

Essential Engineering Intelligence

IET Colloquium on Antennas, Wireless and Electromagnetics

27 May 2014, Ofcom, London, UK

Signal Propagation on a Railway

Wireless Condition MonitoringSystem

Ahmadreza Faghih,Costas Constantinou, Edward Stewart

School of Electronic, Electrical and Computer

Engineering, The University of Birmingham


2/17


Railway Research Centre

University of Birmingham

Acknowledgments: Thanks to Motorail Logistics, owners of LongMarston railway test track


3/17


Introduction

Railway condition monitoring:

Wheel Profile Monitoring, Bogie Performance Detector, Tread

Condition Detectors, Hot Axle Bearing, Acoustic bearing Defect

Detectors and

Reduce running and maintenance costs by using a train-wide

wireless sensor network (WSN) Need to characterise wireless channel reliability on moving

train, in different environments and geographic locations

Quantify reliability of railway-WSN


4/17


Time-Varying Channel

Multipath channel with significant time variability due to

Relative movement of linkages and

changing surrounding environment train speeding past embankments,trees, tunnels, station platforms, etc.

Because of train movement, vibration, travelling around bendsand proximity of scatterers in different environments, both

shadowing and multipath fading are highly variable Receiver signal strength, BER , packet loss and data rate are

stochastic functions of time and sensor locations

Propagation time is much lower than frame transmission time intarget railway systems MAC protocol operation is efficient


5/17


Channel Signal Statistics For WSN

A static train configuration in situated in an open area isamenable to electromagnetic modelling of the radio channel

between nodes Moving train through a variety of environments is too

complex to model, so we adopt a statistical signal description

approach Emphasis of this preliminary study is slow channel path loss

variability using a 2.4GHz COTS transceiver pair

We present a simple statistical characterisation of the timevarying channelon a real train in a test track environment


6/17


TRC104NI

DAQ

Excel

File

PIC

MCU

2.4GHz

transceiver

RSSI

ValueTRC104

2.4GHz

transceiver

SPI

Data Acquisition Method I

Pair of wireless Transceivers: TRC104 2.4-2.52GHz , 128

channel with 2.16 dBi monopole antennas

Sample and digitise RSSI at a sampling rate of 104samples/s

Transmitter operates in continuous mode

System calibration to convert RSSI to path loss done inanechoic chamber


7/17


Data Acquisition Method II

Test Route at Long Marston train test track

Divide route in different zones to evaluate channel for

straight track, bends, and/or objects beside track, open areas


8/17


Nodes Location


9/17


Nodes Location (cont.)

Checked for free channel using spectrum analyser and set

channel: 2.520GHz Ptx= 2 dBm

Receiver kept in one location

Transmitter location moved as shown in config110

~ 10 m


10/17


Statistical Analysis of Results

First order statistics presented only

Construct histogram/PDF (Probability Density Function) and

CDF (Cumulative Distribution Function) for received power

and hence path loss

Evaluate the median path loss and IQR for each configuration

during train movement and stop conditions in various

sections of track as well as around entire circuit

The number of significant energy transport paths are related

to separation and nodes location


11/17


Transmitter and receiver were installed on opposite sides of a

tanker wagon with NLOS, so bends do not have noticeable effect

on signal strength

Estimated that dominant paths go around as well as below wagon Buildings and parked trains beside track have little influence on

spread of signal values

First example of observed Prx

PDF

0

0.02

0.04

0.06

0.08

0.1

-95 -93 -91 -89 -87 -85 -83 -81 -79 -77 -75 -73 -71 -69 -67 -65 -63 -61 -59 -57 -55 -53 -51 -49 -47 -45 -43 -41

Frequency

Prx (dBm)

Config3


12/17


Transmitter installed on passenger wagon, receiver on tankerwagon and on opposite sides of train

The bottom of passenger wagon was lower compared to tanker

wagons so the path(s) under the wagon are weaker

There was a big building beside track on the transmitter side for

part of this railway track

Bends have a marked effect on signal strength giving rise to two

peaks in PDF

Second example of observed PrxPDF

0

0.01

0.02

0.03

0.04

0.05

-95 -93 -91 -89 -87 -85 -83 -81 -79 -77 -75 -73 -71 -69 -67 -65 -63 -61 -59 -57 -55 -53 -51 -49 -47 -45 -43 -41

Pr

obability

Prx (dBm)

Config2


13/17


CDF for static train

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

-100 -80 -60 -40 -20 0

Config4

Config5

Config1

Config6

Config2

Probability

Prx dBm

dTx-Rx

(m)

M

(Loss)

IQR

(Loss)

Config1 10.5 79 dB 2.5 dB

Config2 456.2

dB1 dB

Config3 4.75 55 dB 0.7 dB

Config4 1170.3

dB0.8 dB

Config5 19 77 dB 0.8 dB

Config6 11 73 dB 3.4 dB


14/17


CDF for moving train along whole track

dTx-Rx

(m)

M

(Loss)

IQR

(Loss)

Config1 10.584.2

dB

10.7

dB

Config2 4 70 dB13.2

dB

Config3 4.75

62.5

dB 7.5 dB

Config4 11 77 dB10.5

dB

Config5 1979.3

dB 8.9 dB

Config6 1178.5

dB

10.8

dB

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

-100 -80 -60 -40 -20 0

Config3

Config4

Config5

Config1

Config6

Config2

Prx dBm

Probability


15/17


Future work

Next steps are to:

Calculate BER from SNR statistics

Calculate PER from packet length and error correcting code Determine appropriate system thresholds (minimum acceptable PER

for sensing application)

Calculate WSN outage probabilities assuming transmit node power

control characteristics

Calculate signal transmission delay distributions

Compute second order statistics to establish fade durations

and level crossing rates

Translate to WSN outage duration statistics

Examine mitigation strategies and associated WSN protocols


16/17


Summary

Preliminary measurement results on real trains presented

Limited first order statistics considered only in this work

Distance dependence is not a reliable indicator for path loss Most paths are non-line of sight

Train movement results in 5-14 dB of increase in median pathloss

Train movement results in 7-13 dB increase in IQR of path loss

Scatterers in immediate vicinity to track (parked trains,building, etc.) as well as bends in track have a strong effect on

path loss variation and often observed to give rise to multi-modal path loss distributions


17/17


References

A. Goldsmith (2005) Wireless Communication, Cambridge University

H. Chuan, M.S. Alouini (2011) Order Statistics in WirelessCommunication, Cambridge University

E. Biglieri, S. Benedetto (1999) Principle of Digital Transmission withWireless Application, Prentice-Hall

W. Stallings (2007) Data And Computer Communications, 7thEdition,Prentice-Hall

N. Yaakob, I. Khalil and J. Hu (2010) Performance Analysis of OptimalPacket Size for Congestion Control in Wireless Sensor Networks, IEEE

K. Doddapaneni, E. Ever (2012) Path Loss Effect on Energy Consumptionin a WSN, International Conference on Modelling and Simulation

T. Stoyanova, F. Kerasiotis (2009) A Practical RF Propagation Model forWireless Network Sensors, International conference on SensorTechnologies and Applications

Signal Propagation on a Railway Wireless Condition Monitoring System

Documents