PC BASED MONITORING AND FAULT PREDICTION FOR SMALL HYDROELECTRIC PLANTS
D S Henderson, K Lothian, J Priest
Napier University
INTRODUCTION
impact of the maintenance operation on product quality,
production costs, and more importantly, on bottom line
profit.
The traditional approach to metering of small scale
generation systems is to have a hard-wired metering
panel situated in the control room of the applicable
powerstation. In the case of small or micro hydroelectric
powerstations this might typically be restricted to the
measurement of current, voltage, power, kWh, hours-run
and frequency. Financial constraints often mean that
such plant run unattended and hence the recording and
monitoring of the performance of the plant relies on,
sometimes infrequent, visits by appropriate personnel.
The LabView package has recently appeared on the
market and this enables the creation of Virtual
Instruments which can display, on the screen of a PC,
measured quantities such as are required at a typical
small hydroelectric powerstation. This in itself is an
exciting prospect as it means that, subject to suitable
scaling, the metering of any generating scheme can be
performed by a standard hardware/software package.
Increasing maintenance costs, lost production due to
machine failure, the reluctance to tie up capital in backup machinery and spares, and the safety of personnel are
all factors which are encouraging maintenance managers
to look at modern machinery condition monitoring
techniques to provide a scientific basis for planned
maintenance.
During the 1996197 session, two final year Honours
projects were running at Napier University which related
to the use of affordable PC based systems, the first
performing a metering function, and the second
performing a fault prediction function. This paper
reviews maintenance strategies and condition
monitoring techniques and then describes the
development and outcomes of the two projects.
MAINTENANCE STRATEGIES
The use of a PC based metering scheme gives rise to
numerous added benefits; the measured data can be
stored in the memory of the PC to allow subsequent
downloading and analysis; the data can be analysed by
other software to detect and predict fault conditions, i.e.
to perform a condition monitoring function; and the data
can be interrogated remotely by a suitable
communications link to enable analysis and monitoring,
effectively from anywhere in the world.
Developing the condition monitoring aspect further, it is
noted that maintenance costs are a major part of the total
operating costs of all manufacturing and production
plants. Depending on the specific industry, maintenance
costs can represent between 15% to 40% of the costs of
goods produced.
Recent surveys of
maintenance management
effectiveness indicate that one third of all maintenance
costs is wasted as the result of unnecessary or
improperly carried out maintenance. The dominant
reason for this ineffective management is the lack of
factual data that quantifies the actual need for repair or
maintenance of plant machinery, equipment and
systems. Maintenance scheduling has been and, in many
instances, is predicated on statistical trend data or on the
actual failure of plant equipment. Until recently, middle
and corporate-level management have ignored the
28
Maintenance Management
There are three maintenance management strategies
available to the maintenance engineer. These are
Breakdown or Run-to-Failure, Preventative and
Predictive. In order to understand the role of modern
condition monitoring it is necessary to first look at the
background of maintenance management.
The objective of maintenance management is to achieve
a level of maintenance within financial constraints and
to a prescribed standard of safety. The objectives are
representative of the required balance between the
maintenance costs, the degree of plant availability and
the standard of performance. Until recently, a large
proportion of companies have approached their
maintenance programme with a breakdown or
preventative strategy.
Breakdown
(Run-to-failure)
Maintenance
Management. The logic of breakdown maintenance is
simple and straightforward: when a machine breaks
down, fix it. This if it aint broke, dont fix it method
of maintaining plant machinery has been a major part of
plant maintenance since the first manufacturing plant
Power Station Maintenance: Profitability Through Reliability, 30 March - 1 April 1998.
Conference Publication No. 452, 0 IEE, 1998
was built, and
using run-to-fi
money on mail
operate. This
technique that
before any ma
no maintenan
the most
management.
the surface, sounds reasonable. A plant
ure management does not spend any
nance until a machine or system fails to
:ype of management is a reactive
raits for machine or equipment failure
tenance action is taken. It is in truth a
? approach to management. It is also
pensive method of maintenance
Few plants
philosophy. In
preventative t
adjustments, e
major expense
management a
overtime labot
production ava
,e a true breakdown management
most all instances, plants perform basic
<s, such as lubrication and machine
n in a run-to-failure environment. The
issociated with this type of maintenance
: high spare parts inventory cost, high
:osts, high machine downtime, and low
ibility.
Preventative
many definitio
all managemei
words, maintei
hours of oper,
(MTTF) stat
preventative
programmes
lubrication anc
preventative n
lubrication, ac
critical mack
denominator f
programmes is
:aintenance Management. There are
of preventative maintenance, however,
programmes are time-driven. In other
nce tasks are based on elapsed time or
on known as the mean-time-to-failure
ic. The actual implementation of
iaintenance varies greatly. Some
: extremely limited and consist of
linor adjustments. More comprehensive
ntenance programmes schedule repairs,
stments, and machine rebuilds for all
ery in the plant. The common
all of these preventative maintenance
le scheduling guideline, time.
All preventatii
degrade withii
classification.
mode of opera
directly affect
The mean-timc
same, for exill
and one hand11
using MTBF s
an unnecessaq
programmes assume that machines will
a time frame typical of its particular
le problem with this approach is that the
In and system or plant-specific variables
le normal operating life of machinery.
letween-failures (MTBF) will not be the
)le, for a pump that is handling water
;abrasive slurries. The normal result of
tistics to schedule maintenance is either
%air or a catastrophic failure.
Predictive
maintenance,
definitions. I
electrical swit
problems, to
machinery in a
to prevent
supposition o
monitoring of
efficiency, a1
condition of
required to I
repairs.
4aintenance.
Like
preventative
xedictive maintenance has many
im monitoring infrared images of
gear and motors to detect developing
nonitoring the vibration of rotating
attempt to detect incipient problems and
catastrophic failure. The common
xedictive maintenance is that regular
;actual mechanical condition, operating
other indicators of the operating
ocess systems will provide the data
sure the maximum interval between
In predictive maintenance programmes, the specific
failure mode can be identified before failure. Therefore,
the correct repair parts, tools, and labour skills can be
available to correct the machine problem before
catastrophic failure occurs. Perhaps, the most important
difference between reactive and predictive maintenance
programmes is the ability to schedule the repair when it
will have the least impact on production. The production
time lost as a result of reactive maintenance is
substantial and can rarely be regained. Most plants,
during peak production periods, operate twenty four
hours per day, therefore lost production time cannot be
recovered.
Condition monitoring (CM) is essential for predictive
maintenance schemes, which can make dramatic savings
in maintenance and operating costs, increase safety and
have environmental benefits. CM requires the
monitoring of various parameters and this is usually
achieved through the utilisation of sensing devices.
A PC BASED
SYSTEM
GENERATOR MONITORING
The first of the two projects concerned the development
of a PC based generator monitoring system. The system
was designed around a specific micro hydroelectric
generation station, Ashfield Mill, which is located two
miles North of Dunblane on the Allan Water, Henderson
and Maclean (1). The weir is located at an old cloth
dyeing mill site and creates a net head of approximately
6m. Hydro development, providing mechanical power,
began at this site in the 19th century, however the
original powerhouse was abandoned in the 1970s. The
present site owner and developer has installed two
turbine-generator sets to capture the energy available in
the water.
The existing metering system of the generator concerned
comprises
hardwired
instruments.
Continuous
monitoring of the generator is impossible at present as
no form of data storage exists. Whilst a control console
has been implemented it is anticipated that a completely
automatic monitoring system could also be incorporated,
on a trial basis; allowing measurement and data storage
at regular, pre-specified intervals. Using the present
arrangement, operator attendance would be required for
data storage - this will not be necessary using LabView
and an appropriate database system.
LabView is a program development application, much
like various commercial C or Basic development
systems. However LabView is different from those
applications in one important respect; other
programming systems use text based languages to create
lines of code, while LabView uses a graphical
programming language, G, to create programs in block
diagram form.
29
Tests were done on a laboratory based motor-generator
set and the displayed values on the VIs were verified
with meters and other instrumentation. The VIs gave
on-line readings which were comparable with the
metering. The measured data was then sent to an Excel
spreadsheet for longer term storage, and if required,
further analysis. This was achieved, but with limited file
storage space in the prototype version.
LabView programs are called Virtual Instruments (VIS),
because their appearance and operation imitate actual
instruments. However, they are analogous to functions
from conventional language programs. VIS have both an
interactive user interface and a source code equivalent,
and accept parameters from higher level VIS. These
features are discussed below:
The interactive user interface of a VI is called the
front panel, because it simulates the panel of a
physical instrument. The front panel can contain
push buttons, graphs, meters, and other controls and
indicators. Data is entered via a mouse or a
keyboard; the results are then viewed on the
computer screen through the indicators i.e. the
meters and graphs.
CONDITION MONITORING
PREDICTION PROGRAM
AND
FAULT
A work-placement experience with a company
specialising in small hydroelectric plant identified scope
for a final year Honours project at Napier University
which would address the condition monitoring of small
hydro powerstations in light of modern maintenance
philosophy. The purpose of this project was to develop a
computer programme for small hydro-electric plant
capable of evaluating various live data and producing
a comparison between this collected data and expected,
pre-determined values in order to detect and predict
certain faults occurring within the plant.
The VI receives instructions from a Block Diagram constructed in G coding. The block diagram is a
pictorial solution to a programming problem, with
the various icons and connections being made with
wires. The block diagram is the source code for
the VI; it sends instructions and information to the
front panel.
VIS are hierarchical and modular; they can be used
as top level programs, or as subprograms within
other programs or subprograms. A VI within another
VI is called a sub-VI. A sub-VI must have an icon
and connector pane - these work like a graphical
parameter list so that other VIS can pass data to it as
a sub-VI, (2)
This project involved the design of a PC based condition
monitoring programme to enable fault detection and
fault prediction in small hydro plant. The monitoring
programme makes use of data collected by the
supervisory control and data acquisition system
(SCADA). The SCADA system consists of a Toshiba
PROSEC T2 Programmable Logic Controller (PLC)
supplied by data collected by individual sensors. The
data is interfaced between the PLC and the PC. It was
decided that the most appropriate way of obtaining the
necessary data whilst trying to minimise maintenance
and support costs was an on-line monitoring system. The
data collected can be used by engineers to make
operating decisions on the plant . The information will
also form a basis for anticipating equipment repair
requirements and producing maintenance schedules.
Although the programme was designed with portability
in mind, it was necessary to use a typical hydro-electric
scheme as a model. The scheme selected was Balgonie
Castle near Glenrothes, Fife.
The use of LabView was investigated to allow the
measurement of the parameters specified by the site
owner. These are: 3 x Line Current, Line Voltage,
Power, Frequency, Speed, Hours-run and kWhs. Whilst
the design is for the smaller of the two turbine-generator
sets at the site, it is anticipated that the package will be
capable of monitoring both machines at their different
rated outputs; and be adaptable for any similar
installation.
The interface between the PC and the power system
comprises a PC mounted Data Acquisition Card
(LabPC+) and an interface board for connection
between the PC and the plant mounted transducers. VIs
were developed initially to display the r.m.s. values of
line voltage and the 3 line currents. VIs for frequency
and running speed were developed from the line voltage
signal. Hours-run was established from PC based timers,
however a current snag is that these unfortunately reset
on shutdown of the programme. A real power VI was
developed which was based on the product of the
voltage and current values, also taking into account the
phase angle between the two. The latter was calculated
by determining the relative angular displacement
between the maximum values of their respective
waveforms. kwh was thus determined by integrating the
power over the running time.
The monitoring programme makes use of two
measurement systems to detect a change in the operating
condition of plant equipment;
Static measurement - uses static alarm limits which are
pre-selected thresholds that are a constant within the
programme. If the measured parameters exceed the preset limit, an alarm is displayed.
Dynamic measurement - uses dynamic limits and
monitors the rate of change in the measured parameters.
It detects minor deviations in the rate that a component
is degrading and anticipates when an alarm will be
reached.
30
d parameter, the system allows the user
dynamic limits. This option greatly
as the programme can be set up for
schemes. The option
between samples for
The programme also detected when the measured
parameter had deviated from the previous reading
beyond a specified limit, and a dynamic limit alarm was
raised and logged. For demonstration purposes, samples
are suspended when the alarm page is showing.
However, in practice the alarms page would be updated
whilst on show and as stated, can be used as a front end
to the programme.
The programme was initially written in the computer
language Borland C and then upgraded using C++. The
parameters included within the prototype program were;
Head water level
0
Tail water level
0
Turbine bearing temperature
0
Generator bearing temperature
0
Stator winding temperature.
CONCLUSIONS
The laboratory tested VIS proved to operate very
satisfactorily. It now remains to perform on-site testing
of the system at Ashfield Mill, and it is hoped that this
will have been achieved by the time of this presentation.
Further investigations will be done on the data storage
aspect, as well as remote interrogation of the data via a
modem.
The ultimate intmtion is to monitor all of the parameters
shown in Figure 1.
PROG
F/
AMME PARAMETERS
Simulated testing of the Condition Monitoring and Fault
Prediction program showed the potential for this system
when applied in practice. Although the programme
fulfils its monitoring requirements, it is still very much a
prototype and clearly some work remains in respect of
interfacing the program to the designated site, however
no insurmountable problems are foreseen.
J
The two projects described in this paper ran
independently from one another, however there is
undoubtedly scope for integrating the two, using the data
from the metering system as input to the fault prediction
algorithm, and hence increasing the benefit from their
combined implementation on a single PC.
Figure 1 - Parameters intended to be monitored.
In-situ testing was unfortunately not achieved within the
project timescale, however, data was obtained from a
caretakers daily maintenance report and the expected
parameter rangcs were built in to the programme for
simulation purposes. A random number generator
produces this data within the required ranges and the
programme analyses this simulated data as though it was
read from the ports. When any static limits were
exceeded , the programme was able to detect and
identify them. An alarm was raised and logged on an
alarms page, the page updated as each error occurred
showing the nur iber, type, value and time.
REFERENCES
1. Henderson, D.S., Maclean, A. 1997. Application of
an Innovative Electronic Load Governor at Ashfield
Mill - A Case Study. Hidroenergia 97.
I
2. LabVIEW for Windows Tutorial Book. 1994
National Instruments. Chapter 1.
ACKNOWLEDGEMENTS
The authors of the paper would like to thank Napier
University for the opportunity to write and present this
paper.
31
