Вы находитесь на странице: 1из 10

Assessment of structural integrity of wooden poles.

Ian A. Craigheada, Steve Thackeryb, Martin Redsta11', Matthew R. Thomasb


a
Univ. of Strathclyde, Glasgow, Scotland Gi 1XJ
b
BT Research Laboratories, Martlesham Heath, Ipswich, 1P5 7RE, UK.

ABSTRACT
Despite recent advances in the development of new materials, wood continues to be used globally for the support of
overhead cable networks used by telecommunications and electrical utility companies. As a natural material, wood is
subject to decay and will eventually fail, causing disruption to services and danger to public and company personnel.

Enternal decay, due to basidomycetes fttngi or attack by termites, can progress rapidly and is often difficult to detect by
casual inspection. The traditional method oftesting poles for decay involves hitting them with a hammer and listening to the
sound that results. However, evidence suggests that a large number of poles are replaced unnecessarily and a significant
number of poles continue to fail unexpectedly in service. Therefore, a more accurate method of assessing the structural
:Lntegrlty ofwooden poles is required.

Over the last 25 years there have been a number of attempts at improving decay detection. Techniques such as ultrasound,
drilling, X rays etc have been developed but have generally failed to improve upon the practicality and accuracy of the
traditional testing method.

The paper describes the use of signal processing techniques to analyse the acoustic response of the pole and thereby
determine the presence of decay. Development of a prototype meter is described and the results of initial tests on several
hundred poles are presented.

Keywords: Wood decay, wood poles, impact response, acoustic response, condition monitoring.

1.INTRODUCTION
Wooden poles are widely used by telecommunications and electricity utility companies for the support of overhead cables.
They have been found to be a relatively cheap and reliable method for supporting an overhead cable network system. Poles
are typically 150 - 400 mm in diameter and can range in height up to 10 m or more. Traditionally they are placed vertically
in holes, which are drilled into the ground to provide a cantilever type of support.

The tree most commonly used for poles in the UK is the Scots Pine (Pinus Sylvestris) but other species are to be found in
significant numbers. Wood is a natural, biological material and like all such materials is subject to decay. Decay in wood is
caused by certain fungi, which use the wood as a food source. The outer surface of a pole, although usually treated with
creosote, can be attacked by the fungus ascomytes. Such 'soft rots" are relatively easily detected by visual inspection and
probing and are generally slow to progress to ultimate failure of the pole. However, the internal parts of the pole can be
attacked by basidiomycetes fungi, which enter the wood via checks in the pole, which occur as part of the drying out
process. This decay can advance rapidly inside the pole with no external evidence visible to a cursory inspection.
Subsequent failure of the pole can occur unexpectedly once the decay has weakened its structural strength sufficiently. The
decay is found most often at or near the ground level where moisture conditions are most conducive to fungal growth.
Unfortunately this is also the location ofgreatest bending moment on the pole, which magnifies the effect ofthe decay.

Replacement of a pole can cost as much as £1000 and they have an average life expectancy of45 years. The useful life of a
pole is affected by many factors, which are largely beyond the control of the utility operators. Variability in the quality of
the wood, preservative treatment, site conditions, loading etc. can all lead to unexpected failure of the pole. This can result
in serious consequential losses in terms of public/personnel safety and system integrity. To avoid such failures, most utility
companies employ "pole testers" whose job it is to inspect poles periodically to decide if they are fit to continue supporting
the network until the next inspection is to be undertaken. The frequency of inspections is typically 5 -7 years. A number of

In Nondestructive Evaluation of Aging Materials and Composites IV, George Y. Baaklini, Carol A. Lebowitz,
Eric S. Boltz, Editors, Proceedings of SPIE Vol. 3993 (2000) • 0277-786X/OO/$1 5.00 131
utility companies employ a basic acoustic response testing method to determine the presence of internal decay. The tester
strikes the pole with a hammer and listens to the noise that is made. Subjective assessment is then made by the tester as to
the severity of any decay present. Surprisingly, this method is reasonably reliable but understandably testers tend to err on
the side of caution. It is generally recognised that a large number of poles are classified as 'defective" when the level of
decay is insufficient to warrant their replacement. Such unnecessary replacements are costing utility companies £ms per
year. Over the last 25 years considerable efforts have been made to improve upon existing testing methods and develop a
system, which is reliable, portable and inexpensive.

2. PREVIOUS TESTING METHODS


A wide range of technologies have been used in an attempt to produce a pole testing system that is more reliable than the
"wheeltapper' approach. The main methods, which are described in the literature or have been developed commercially, are
outlined.

2.1 Ultrasonics

Ultrasound is a vibration which is at a frequency which is higher than the audible range (> 20 kHz). Usually pulses of
ultrasound are directed into the pole in a radial direction. The time for the pulse to travel across the pole is measured. This
procedure is usually repeated at a number of locations around the circumference. The waves of ultrasound travel at speeds
of approximately 2000 m/s in sound wood and slow down when passing through or around decayed sections. The time of
flight gives an indication of the presence of decayed material. The method was pioneered in the 198Ots12 and developed
commercially3. It has been used successfully on building timbers but has had mixed success on utility poles. The cost of
the instrument and the time required to survey a complete pole tend to discourage more widespread use.

2.2 Medical techniques

The location of decay within a pole can be likened to identifying fractures or abnormalities within the human body. It is
therefore not surprising to find that techniques, long established in medical investigations, such as X rays, NMR etc. have
been applied to locating decay in poles. Little published information could be found on the use of these techniques,
presumably because developing a cost effective, safe and portable system would be difficult to implement.

2.3 Electrical impedance

As decay progresses in wood, the fungus produces ions, which change the electrical impedance properties of the wood.
Initial trials of a tomographic sensor based on this approach have been described5 but the accuracy of the system was
insufficient for it to be ofpracticai use. It is believed that further work on this system is ongoing.

2.4 CO2 detection

As well as the production of ions, the decay process results in the production of CO2. This can be detected by standard
chemical analysis equipment and this approach has been studied and evaluated by a number of companies. However, a
reliable and cost effective system is yet to be developed.

2.5 Invasive technologies

Probes forced into the pole have been used in a number of guises over the years. It is believed that there is a direct
correlation between the wood's resistance to penetration and its strength. Italian researchers have provided some scientific
basis for this approach6'7 and used the method successfi.illy on timbers in historical buildings. The basis of the method is to
measure the number of calibrated hammer blows per centimetre of penetration of a specially designed probe. An alternative
method uses an instrumented drill8, which provides a graph of penetration resistance against drill progress. Both methods
suffer from the drawback that a number of tests will be needed to survey a complete pole and each penetration provides a
new potential site for the initiation of decay.

132
2.6 Proof loading.

Applying a test load to a pole, which is in excess of the design load, is probably the most direct means of assessing the
fitness for purpose of a pole. This approach has been used successfully for testing lampposts9 and, in theory, could be
adapted for wooden poles. However there are two major drawbacks :- 1) Many wooden poles are situated at some distance
from roads so vehicles could not always be used to apply the necessary forces required, 2) failure of a pole in testing would
probably result in disruption to the network. For these reasons, proof loading has not been considered seriously for routine
pole testing work.

2.7 Vibration

The decay of part of a pole will result in changes of the Young's Modulus and density in the region of the decay. These
changes have been considered likely to affect the vibrational characteristics of the pole and if these could be measured an
estimate of pole condition could be obtained. Excitation of the pole is usually either by a single blow from an impact
hammer or a repetitive, cyclical force. One of the earliest methods (The Resotest'°) relied on a square wave force input to
the pole approximately im from the ground and measurement of the response at intervals going down to ground level.
Initial tests reported'° seemed to be encouraging but little further information could be found (26 years later) so it would
appear that the system failed to live up to its early promise.

There are a number of difficulties associated with applying vibrational techniques to poles. The bending natural frequencies
start at approximately 2 Hz and occur every 10-20 Hz higher up the frequency range. The precise values vary depending on
diameter, number of wires that are supported, presence of stays, additional masses supported on the pole etc. Internal decay
will only affect the bending strength and hence bending frequencies slightly until the point of collapse so the likelihood of
detecting any change due to decay is extremely small. Despite these disadvantages, the vibration approach has been used
successfully on railway sleepers" where structural parameters between sleepers are much more regulated than for poles.
The manufacturers of the sleeper tester have tried applying their method to poles but abandoned their research due to the
problems previously identified.

It can be seen that considerable research effort'" has been put in worldwide over the last 30 years in an effort to improve
the reliability of wooden poles used in supporting cable networks. Because of the limited success of previous, alternative
studies it was decided to look more closely at the standard acoustic method which has been relied upon for decades and see
ifit could be improved by using modem technology.

3. THE IMPACT RESPONSE METHOD


Striking any object or structure with a hammer causes it to vibrate and this vibration may cause the surrounding air to
vibrate thus producing the sound ofthe impact that is heard. It was considered that it was this sound that pole testers use to
identify decay in poles. The structure, which is struck, usually vibrates at one or more of its natural frequencies. Usually,
only frequencies which lie in the range 0 I 500 Hz can be excited by a hammer blow'2. Verbal description by pole testers
suggested that the characteristic pitch or frequency ofthe impact fell when a decayed pole was stmck compared to a "good"
pole. This implies that one or more ofthe natural frequencies reduce due to the presence of decay.

To investigate this supposition, the acoustic signal and vibration response were recorded from a number of telegraph poles,
which were specifically used for training pole testers at one ofB.T's Training Centres. This work, reported in 13 confirmed
that lower frequencies were observed within the acoustic spectrum when decayed poles were impacted by a hammer. The
work resulted in an algorithm. which was implemented on a lap-top computer. The acoustic signal was picked up with a
Bruel & Kjaer l/2inch microphone and the signal digitised and fed into the computer using a Pico-Log AID board. The
equipment arrangement can be seen in Figure 1.

The laptop-based instrument was used to assess the condition of a number of poles and the predictions compared to the
assessment of experienced pole testers. The output from the pc was a display of the acoustic spectrum, a frequency value
based on the algorithm and a rating from 0 — 10 indicating the "severity" of the decay. Figure 2 shows a typical
measurement screen from the system.

133
Figure 1 Laptop based prototype instrument.

Smoothed Acoustic Spectrum Decay Meter

600C
500
A
— Sene?
4000 f
fl L t
— Sies2 6 Sene
Sers 4
: \,fJ 1 U 2 I
-—.-•
0
U)N.
U) — N r O'- •' .
N. Ut
U) O
Frequency Kz} Range 0-10 4Bad->Good
Actual

Frequency Parameter = 1 2141

Figure 2 Output display from laptop based instrument

The results from the system tended to agree with the assessment provided by traditional methods but there were a few
significant discrepancies. These were due to dominant frequency peaks, which were sometimes evident throughout the
spectrum and were random in nature. These peaks, when present, tended to result in the algorithm "mis-diagnosing" the
pole condition. The shortcomings of the lap-top system were considered to be due to using the algorithm on a single impact
sound whereas the original algorithm was based on an averaged spectrum taken from 16 impacts. it was considered that the
averaging process tended to remove the dominance of "random" peaks and result in a reduced mis-diagnosis rate.

To overcome the problems it was decided to collect sounds from a much larger selection of poles and investigate a number
of different algorithms to identify the best for use with a single impact.

134
4. DATA COLLECTION & ANALYSIS

A number of operational poles were identified at various locations throughout the U.K. Sound recordings were taken of
hammer impacts applied to various positions on the pole. The positions included were generally within 2m of the ground
and measurements were taken at positions where decay was suspected and also on sound sections of the pole. The presence
of decay was confirmed afterwards by drilling the pole. The data was digitised at a rate of 1 1025 samples per second
enabling frequencies of up to 5. 5 kHz to be assessed. The various impact sounds were stored in the form of .WAV files for
:ftirther processing. In total, 570 impacts were recorded.

.A software analysis system was developed, based around the FFT process, to enable various algorithms to be applied to the
data and evaluated. Figure 3 shows typical output from the system. Factors such as the weighting, band pass filtering,
number of data samples, smoothing etc. could be changed and their effect readily evaluated. In order to assess the accuracy
ofthe algorithms proposed, 100 test samples were identified. 50 were from poles, which were known to have decay present

Bang-A lyser

Z:
-30000
0 io 4'o 3 - . .
100

Time (ms)

C 500 1000 1500 2000

Frequency Spectrum (Hz)

Audio Sampling Rate; 11025 samples per second.


Samples Presented to FFT: 1024 samples from sample 70 to 1093.
Fourier Weighting Window: Cos2.
Frequency Weighting: None.
Founer Low Frequency Cutoff: Disabled.
Fourier High Frequency Cutoff: Disabled.
Fourier Smoothing: Enabled. 3 point averaging.
Fourier Vertical Scale: Relative (i.e. values adjusted so highest peak is always at 1.0).

Key Frequencies: ow cutoff 377Hz, balance 528Hz, high cutoff = 538Hz, bandwidth = 161Hz.
Figure 3 Typical results from the sound analysis

135
and 50 were from poles that were considered to be in "good" condition. The algorithms were each applied to the test suite of
impact sounds and their performance was evaluated. The output of each algorithm was a characteristic frequency and it was
found that "good" poles tended to produce a higher characteristic frequency than poles that had decay in the vicinity of the
impact site.

It was then necessary to identify a threshold or cut-off frequency that would discriminate between "good" poles and those
with decay present. Unless the system was lOO% accurate there would be the likelihood that some decayed poles would be
misdiagnosed as being "good" and some "good" poles being incorrectly condemned as being decayed.

In order to quantify the accuracy, two criteria were identified:-

4.1 Minimum error setting.

A cut-off frequency was selected so that the number of good poles which were condemned was equal to the number of
decayed poles which were mis-diagnosed to be "good". The number of cases mis-diagnosed should approach zero as the
algorithm accuracy improves.

4.2 Maximum caution setting.

A second, higher cut-off frequency was identified to ensure that all decayed poles were identified. The number of "good"
poles condemned then became a reflection of the accuracy of the system. A perfect system would have both cut-off
frequencies the same, resulting in zero mis-diagnoses and lOO% decayed pole detection. The performance ofthe algorithms

Pole Decay Algorithm Performance


I - Mm. Error 2 - Max. Caution.

aa ab ac ad ae af ag au ai aj ak al am an ao ap aq ar as at au av aw

a Series I Series 2

Figure 4 Algorithm performance. Better algorithms have a high Mm. Error value and a low Max. Caution value.

tried was found to be disappointing and considered unlikely to lead to a reliable instrument. Many of the initial algorithms
produced more mis-diagnoses than correct assessments. Further inspection of the recorded sounds in the time domain and
frequency domain resulted in the proposal of additional algorithms, which produced better results. The performance of these
algorithms on the test suite of pole impacts can be seen in Figure 4. The most successful algorithm was "am" with a success
rate of 84% in terms of minimum error and 26% of "good" poles condemned on the maximum caution evaluation.

Figure 5 shows the distribution of frequency parameters for algorithm "am". The minimum error threshold frequency is at
950 and the maximum caution threshold at 1070.

136
Jgonthm PerFormance

4O
120

100

60 •1•
.-:
——-• •• -w
• L__
S

—s-- S.I
••
• •.

.
.I I
____________
•LcaYedRie1
SQxxiRIes

20
,
0 S •• • .
0 500 1000 1500 2000 2500 3000
Frequency Parameter

Figure 5 Best algorithm performance (am).

5.HARDWARE DYELOPMENT.

The accuracy of the algorithm was considered to be sufficient to warrant design and production of 5prototype meters based
upon the software developed. The meter will be a battery powered, handheld system similar to that shown in Figure 6. The
microphone will take in the sound from a hammer/pole impact and the signal will trigger the digitisation process. The sound
will be recorded for approximately 90 msec and this signal will then be the basis for the proposed algorithm. Calculation of
the characteristic frequency for the sound and comparison with the threshold frequency will enable display of the frequency
and determination of a "decay parameter" probably on a scale of 0 — 10. A reading of 10 should indicate a "very good" pole
or one where there is a high degree of probability that no decay is present whereas a reading of 0 or 1 would indicate severe
decay with some certainty. Values in between these extremes are likely to provide a less certain diagnosis but it is expected
that it will be more reliable and consistent than the present subjective,aural techniques. At the present time, the prototypes
are being manufactured and it is expected that field trials will commence in the near ftiture.

137
Start Arm (÷)

Sleep (.)

Armed

Please Wad

12345...
I
Score 04 82%

865Hz

Figure 6 Schematic diagram of pole decay meter.

6.DJSCUSSION

The data collection was undertaken by BT personnel and the recording system worked very well The impact sounds were
readily converted to WAy files and subsequent data handling was straightforward The signal analysis software was
extremely user friendly and allowed efficient evaluation of the algorithms under consideration. The system underwent
constant improvement during the research as minor shortcomings were identified and remedied. At the present time, the
system has been automated to such an extent that the 100 test sounds can be analysed in a matter of a few minutes

The initial results from the 570 pole sounds were based on the original algorithmU, which relied on smoothing the averaged
spectrum and identifying a lower threshold frequency The results, on the whole, tended to follow earlier results but there
were a significant number of anomalies, which were usually due to individual peaks appearing at higher (>600 Hz) or lower
(K 150 Hz) frequencies causing radical changes to the threshold value obtained It was considered that this was due to
considering a single impact in the current tests whereas the original work was based on a spectrum averaged from 16
impacts on the same section It was felt that the averaging process probably reduced the influence of these peaks, which
were of a somewhat random nature, in a practical implementation of the algorithm, it is unlikely that a tester would want to
impact the pole a number of times to enable averaging to be undertaken A meter based on a single Impact would therefore
be preferable

138
In an attempt to eradicate these anomalies associated with single impacts, a number of different algorithms were tried. The
algorithms were based upon a range of different frequency characteristics of the signal. However, their success rate
continued to be disappointing.

A number of alternative algorithms were identified after considering the time histories and spectra of the 100 test samples
more closely. The accuracy of these algorithms improved with the better ones achieving typical values of 80% for the
minimum error setting and 25% for the maximum caution setting. The best algorithm was identified and it's performance is
shown in Figure 5. Initially a value of 84% was obtained for the minimum error value and 26% of good poles were
condemned in the maximum caution setting. However, it can be seen (Figure 5) that there is a cluster of circular dots
towards the lower left corner of the graph, which are separated, from the other circular dots for the poles without decay. By
moving the cut-off frequency from 950 to 1050 almost all decayed poles are identified (98%) with no increase in the
number ofgood poles condemned.

Furthermore, the cluster of circular dots which gave a relatively low frequency are all from tests conducted at Harrogate on
American Southern Pine poles or European redwood poles rather than Scots Pine. In addition it was raining during testing of
these poles and the wood surface was reported to be wet. It would appear that either the wood type and/or the wetness of the
wood has resulted in a lower frequency for these good poles resulting in the erroneous diagnosis of decay present. It is felt
that both of these factors could be taken into account, after further research, resulting in a much-improved performance of
the meter. If this cluster of circular dots are discarded, the algorithm gives a virtually, clean separation of circular (good)
and diamond (diamond) points. The success rate for minimum error would become 98% and the maximum caution criterion
would result in 3% ofgood poles being condemned.

It could be argued that the decayed samples are from poles with significant decay present (at least sufficient decay to be
detected by current methods) and that it is unlikely that the algorithm will perform as well on poles with small levels of
decay present (the borderline cases). This is probably true but it should be remembered that the test result diagnoses are
based on a single impact. A tester currently will usually impact the pole a number oftimes to get a general impression of the
decay. This could be done with the proposed meter and figures averaged to improve diagnosis reliability. It should also be
remembered that if there is only slight decay present it is likely that the pole is not in imminent danger of collapse and the
chances are that it will survive until the next inspection, so reduced reliability in these cases may be acceptable.

The most successftil algorithm was therefore considered to be most appropriate for implementation in the prototype meters.
Because of the unresolved matter with the cluster of poor results it was decided to make the cut off frequency adjustable to
allow further investigation and refinement to be undertaken. The output from the meter was chosen to be a scale from 0 —
10. Scores of 4 and under should indicate decay and scores of 5and above a good pole. The further the reading is from the
mid range the more reliable the diagnosis is likely to be.

The meters are being designed to be easy to use so that there is little scope for operator error. It is envisaged that the meter
could be manufactured relatively cheaply (several hundred pounds) so that each tester could carry one around in their
toolbag. As the meter relies on an acoustic signal, then any ambient, background noise may affect the accuracy of the
analysis. It is considered that the meter should be capable of reliable operation at any location where current, aural test
methods are used.

It is anticipated that 5 prototype units will soon be available for initial field trials to be undertaken. The meters will be hand
held, batteiy powered devices for ease of portability. The analysis should only take 10-1 5seconds and then a further impact
can be captured. Averaging around a circumferential line on the pole will be easy to undertake to improve reliability
compared to a diagnosis after a single impact. Use of these meters will provide valuable information on the performance and
reliability of the system and enable further research to improve the algorithm by taking into account factors such as wood
type, pole diameter, moisture content etc if applicable.

139
7. CONCLUSIONS
Over 500 pole/hammer impact sounds have been recorded and analysed in the time and frequency domains. Approximately
40 algorithms have been tested on this database to find the best at identifying differences in the sound between decayed and
non-decayed poles. The best algorithm resulted in correctly diagnosing 84% of decayed poles in a representative sample
whilst condemning 26% ofgood poles. 18% ofthe "good" poles were tested when the wood was wet and ifthese results are
discounted the algorithm resulted in an accuracy of98% with only 3% of"good" poles incorrectly diagnosed as decayed.

Prototype meters are currently being manufactured, based on the identified algorithm and ftirther testing and field trials are
planned.

ACKNOWLEDGEMENTS

The authors would like to express their gratitude to BT Research Labs for the support provided to enable the work to be
undertaken and agreement to publication of this work. Thanks are also expressed to BT Training Centre for allowing access
to their facilities.

REFERENCES
1. K. Arito and K. Kuratani, "Wooden Pole Tester for determining the strength of decayed
wooden poles," Japanese Telecommunications Review, 26, Pt 3, pl67-l'73, 1984.

2. Y. Tomikawa et al. "Nondestructive inspection ofa wooden pole using Ultrasonic


Computed Tomography," ,IEEE Trans on Ultrasonics, Ferroelectrics andFrequency Control, 4, p354-358,
July 1986

3. Fujikura Europe Ltd - Sales literature 1999. Brook Lane, Westbury, Wiltshire, BA13 4ES

4. W.W. Wilcox, "Detection ofEarly Stages ofWood Decay with Ultrasonic Pulse Velocity",
Forest Products Journ., 38, No 5, p 68-73,May 1988.

5. N. White and R. Waterfall "EIT for the Condition Monitoring ofWood Poles Carrying
Overhead Power Lines", Procs oflEE Coil on Advances in Electrical Tomography, Paper 13,June 1996.

6. B. Giuriani and A. Gubana, "A Penetration Test to Evaluate Wood Decay and its application
to the Loggia Monument", Material & Structures , 26, p8-l'4, 1993.

7. P. Ronca and A. Gubana "Mechanical Characterisation ofWooden Structures by means of


an In-situ Penetration Test", Construction & BuildingMaterials, 12, p233-243,1998.

8. Resistograph Technical Literature. Distributed by Fujikura Europe Ltd. Brook Lane, Westbury, Wiltshire, BA 13 4ES.

9. L. Hibbert "Throwing Light on a Hidden Menace' IMechE. ProfessionalEngineering, 11, No 3, p22, Feb 1998.

10. A.D. Shaw, "Sonic Vibration technique for Rot Detection in Wood Poles", IEEE Trans on
Instrumentation & Measurement, 23, No 3, p 240-244, Sept 1974.

ii. Beran Instruments Ltd, Hatchmore Industrial est., Torrington, Devon EX38 7HP - personal communication.

12. J.T. Broch, Mechanical Vibration & ShockMeanirements, Bruel & Kjaer,Denmark,l981.

13. l.A. Craighead, M. Redstall and M.R. Thomas "Detection of Decay in Wooden Poles" LCP Conference, Chorley,
Lancashire,UK, Sept. 1999

140