Examination of Witnesses (Questions 19740
- 19753)
19740. So far as the identification of the plus
5dB uncertainty, how have you arrived at that?
(Mr Thornely-Taylor) This
was done specifically for Crossrail. At the end of the previous
Crossrail project they sponsored an exercise where we went out
and predicted first, measured second, so there was not any shoehorning,
as Dr Hunt calls it. The ground level noise in the basement of
Christies' auction rooms in Kings Street between Green Park and
Westminster, that showed that in dBA terms—it is what we
usually use, NC is quite unusual—there was an error of 1.8dB,
a standard error in the distribution of different levels from
different trains. That was a case where we could not measure the
actual rail roughness, we had to use the standard assumption that
we use when we do not have rail measurements. The Greenwich validation
exercise, which Dr Hunt has talked a lot about, showed that on
one tunnel there was 0.3 under-predictions and on the other tunnel
there was 3dB over-prediction, so we have taken those two results
and generally decided to apply 5dB uncertainty for all Crossrail
predictions, and we are applying them to each octave band in this
case. What you have seen from Dr Hunt's evidence is that you do
get uncertainties which do cause departures from the measured
but what mainly is happening is there is a frequency error. Peaks
are occurring in slightly different places in the spectrum and
it does mean, if you look at that particular spot (indicating)
there can be a big difference between the measured and the predicted
because the peak has gone to the left or the right. Because we
are either dealing with dBA or these curves it does not matter
a lot. If that peak went to there (indicating) or that peak went
to there, there is a small change in the overall level, but it
is nothing like as great as the difference you would find if you
looked only at that one frequency which is basically what Dr Hunt
was talking about. When one of my pairs of charts was put up showing
how the Greenwich validation exercise, the comparison measure
of a model there, is evaluated in terms of normal NC terms, I
say normal but normally they are octave bands, shifts in peaks
just disappear and the maximum error is something like 1.3dB which
is extremely good as modelling goes.
19741. I think the suggestion was that the uncertainty
associated with Findwave was to be criticised because there has
been no inter-model comparison, what do you say about the lack
of inter-model comparison?
(Mr Thornely-Taylor) There
has been in Denmark. I have been talking about the Copenhagen
Metro. We did a series of predictions there, and as a check, a
well known competitor of my firm in Denmark was asked to do their
version of it, quite a different process using something called
finite element so there is a different mathematical approach,
but it is a plain numerical model and they have got very similar
predictions. I will, when I have the time, see how it compares
with Pipe-in-Pipe. Unfortunately, proceedings here have been keeping
me busy and I have not had time to do it.
19742. Let us deal with one aspect more of Findwave.
Dr Hunt explained that he was not sure how certain non-linearities
were taken into account in Findwave, or indeed whether they were
taken into account, and he referred to pad specification and the
specification of sleeper pads. What do you say about that?
(Mr Thornely-Taylor) The
first thing to make clear is that Dr Hunt was saying that non-linearity
does occur or should be taken into account. One of the questions
put to him by Mr Newberry suggested that it was a shortcoming
of Findwave, as that could not cope with non-linearity. What was
suggested was models must be able to cope with linearity and non-linearity,
and it does, but it is actually one of the areas where there is
the least uncertainty because non-linearity occurs in two places,
principally, as he said: one is in the resilient pads that go
under the rails or the bearings that go under the floating track
slab. Of all the things in a track work specification, one of
the things that is most closely specified is the performance of
resilient elements, and contractors are required to put elements
through the most phenomenal series of tests to discover exactly
how they behave under the loads that they will be subjected to.
That is where the non-linearity occurs—if you squash a rubber
bearing it gets stiffer. That is done to a very high degree of
accuracy, and the model can reflect that very high degree of accuracy.
So that is not an area of uncertainty of any significance. Dr
Hunt also mentioned, quite rightly, that the soil can be non-linear,
and that is a very interesting topic. In one of the appendices
of the specialist technical reports that are in the corner of
the room there, the blue reports, there is a long dissertation
by me reviewing the literature on the mechanisms in the soil that
cause damping. In particular, I took a lot of interest in a PhD
thesis by a Norwegian friend of mine, Christian Madshus who has
studied non-linearity in the seabed from the point of view of
the legs of North Sea oil platforms, which sway about quite severely,
and it is important to understand exactly what he soil does that
they are supported from. That all leads to the conclusion that
at the tiny displacements that are involved in ground-borne noise,
tiny factions of huge movements that North Sea oil platforms make,
the non-linearity disappears to virtually nothing, and there is
effectively no contribution of this hysteretic effect, as it is
known, on the damping of soil where we are concerned. So Dr Hunt
is right in principle but it has no consequence in the case of
an underground railway making groundborne noise.
19743. Dr Hunt also said that the Committee
ought to adopt a plus or minus 10dB approach to any modelling,
and he explained that was based on his experience, and indeed
upon using Pipe-in-Pipe. What do you say about that contention
in relation to Findwave?
(Mr Thornely-Taylor) It
is not necessary. The work that I have been describing at some
length has shown that a 5dB uncertainty is adequate. All projects
compile a risk register and they ascertain what the risks are
of all sorts of things turning out different ways. The risk associated
with something going wrong with a 5dB uncertainty is well within
the normally accepted uncertainties in the design of any major
project. You could put 10dB on all predictions and you would have
a much more expensive railway, and it would probably be much quieter
than you need.
19744. What relevance are conclusions drawn
from operating Pipe-in-Pipe to its validation, when trying to
draw conclusions about uncertainties using Findwave?
(Mr Thornely-Taylor) The
uncertainties in Pipe-in-Pipe are not quite what they seem. They
are largely to do with this point I made about the peak cropping
up in a slightly different place. There is at any one frequency
a very large difference when a peak moves up or down in frequency.
One of the things Dr Hunt said and did—one of his celebrated
physical demonstrations—at the meeting (which I think probably
is the reason why he is here) when he told a group of people for
the Institute of Acoustics at the Royal Society about Pipe-in-Pipe,
was he showed that it is very sensitive to position on the ground,
and that a 5 metre movement in position gave a very large difference
in prediction. He marched across the stage saying if the tunnel
was here it would be 10dB different at one frequency to if the
tunnel was here. It was quite impressive but, of course, we are
always predicting groundborne noise in buildings and buildings
are always more than five metres in size. The excitation of the
foundations of the building means that what comes into the building
from different parts combines in the complete excitation of the
building as a whole. You do not see those effects in practice.
You do not. We can go at random to any location above any underground
railway and do vibration measurements. We find an effect with
distance as we move away from the railway but it does not go up
and down by 10dB, I can assure you. If necessary we can go out
and do it again just to prove it.
19745. Thank you. Is there anything else you
wanted to draw to the Committee's attention at this stage?
(Mr Thornely-Taylor) I think
we have covered everything.
Examined by The Committee
19746. Mr Hopkins: You have got lots
of experience in actually measuring after models have been decided
and predictions made. Your experience, you were saying, is of
experiments comparable to tests at plus or minus 3 decibels.
(Mr Thornely-Taylor) Yes.
19747. Have you found many cases where there
has been plus or minus 10 decibels?
(Mr Thornely-Taylor) There
have been some cases where I have predicted too high and, actually,
Portcullis House is a case in point. The actual vibration in columns
of the building immediately the railway opened, although the spectrum
was the right shape, was lower than predicted. If I was going
to be pedantic I would say it is plus five, maybe minus something
rather more than that, but plus costs money and minus does not,
so we tend to concentrate on the risk of predictions being too
low rather than the predictions being consistently too high and
being built to expensive railways, but exceedances are the things
to worry about and 5dB is, from an engineering point of view,
an acceptable margin to use.
19748. It is a question of scale, is it not?
(Mr Thornely-Taylor) Yes.
19749. Ten decibels over is a very big jump.
(Mr Thornely-Taylor) It
is a huge jump. If you were out by 10dB you would have got the
amount of energy wrong by ten-fold, and instead of there being
one watt per square metre coming through as vibration there would
be 10 watts coming through. You would have to make a heck of a
mistake to be 10dB out.
19750. Mr Newberry: Can I just get a
steer from you as to timing? I do not think I can finish by five
o'clock.
19751. Chairman: Then I think we will
rise because I cannot keep the stenographers beyond five. Are
you talking about with this witness or—
19752. Mr Newberry: Yes, I shall be more
than ten minutes with this witness.
19753. Chairman: Then we will rise and
return tomorrow at 10.00 am.
| 
