Skip to main content

Lyceum 2021 | Вместе в завтрашний день

Знание условий геологической среды имеет решающее значение для успешного геотехнического анализа и проектирования.

Все же геомеханические свойства по своей природе изменчивы, и их трудно узнать, что приводит к неопределенности. Как же тогда мы сможем правильно понять геотехнический риск? На этом занятии представлено средство для характеристики риска как функции взаимосвязи между неопределенностями — представлены разделы для набора инструментов, доступных специалистам в области геонаук и инженерной геологии для устранения неопределенности в оценке рисков, и предложена структура, которая сопоставляет инструменты с характером риска в целях улучшения результатов этой оценки.



Рэй Йост (Ray Yost)
Главный инженер‑геомеханик, Advisian

Директор по геотехническому анализу, GeoStudio — Seequent


30 минут

Смотреть больше материалов Lyceum

Lyceum 2021

Узнайте больше о решении Seequent для горнодобывающей промышленности

Узнать больше

Расшифровка видеозаписи

[00:00:00.079]<br />
(upbeat music)<!— wpml:html_fragment </p> —>
<p>[00:00:10.170]<br />
<encoded_tag_open />v -<encoded_tag_closed />Hello and welcome to this presentation<encoded_tag_open />/v<encoded_tag_closed /></p>
<p>[00:00:11.950]<br />
on understanding geotechnical risk.</p>
<p>[00:00:14.480]<br />
My name is Chris Kelin,</p>
<p>[00:00:15.650]<br />
I’m the director of Geotechnical Analysis</p>
<p>[00:00:17.763]<br />
for the GeoStudio business unit here at Seequent.</p>
<p>[00:00:20.800]<br />
And I have the pleasure of introducing our speaker,</p>
<p>[00:00:23.380]<br />
Dr. Ray Yost.</p>
<p>[00:00:24.929]<br />
Ray has nearly 20 years of experience</p>
<p>[00:00:27.200]<br />
working in the fields of geology, hydrogeology,</p>
<p>[00:00:29.590]<br />
and geotechnical engineering</p>
<p>[00:00:31.720]<br />
for the civil and mining sectors.</p>
<p>[00:00:33.970]<br />
His career has included tenures</p>
<p>[00:00:35.920]<br />
at Oregon Department of Transportation,</p>
<p>[00:00:38.630]<br />
Rio Tinto minerals, Teck Resources</p>
<p>[00:00:41.770]<br />
and more recently, as a principal geotechnical engineer</p>
<p>[00:00:44.844]<br />
at Advisian.</p>
<p>[00:00:46.410]<br />
In this role,</p>
<p>[00:00:47.710]<br />
Ray serves as a subject matter expert</p>
<p>[00:00:49.870]<br />
for a wide range of engineering applications,</p>
<p>[00:00:52.570]<br />
including underground mining,</p>
<p>[00:00:54.430]<br />
surface mine design, tailing storage facilities,</p>
<p>[00:00:58.010]<br />
geo-hazard management, and much more.</p>
<p>[00:01:01.160]<br />
Today Ray will talk to us</p>
<p>[00:01:03.670]<br />
about a framework for understanding risk</p>
<p>[00:01:06.660]<br />
in geotechnical engineering.</p>
<p>[00:01:08.670]<br />
Ray, over to you.</p>
<p>[00:01:11.372]<br />
<encoded_tag_open />v -<encoded_tag_closed />Thank you, Chris.<encoded_tag_open />/v<encoded_tag_closed /></p>
<p>[00:01:14.070]<br />
So my talk is about understanding geotechnical risk</p>
<p>[00:01:16.607]<br />
and the corresponding uncertainty we often face.</p>
<p>[00:01:19.250]<br />
It’s a structure for understanding uncertainty.</p>
<p>[00:01:21.867]<br />
The next slide, please.</p>
<p>[00:01:24.920]<br />
It starts us with this idea</p>
<p>[00:01:26.930]<br />
that small data sets and the corresponding uncertainty</p>
<p>[00:01:30.160]<br />
that comes with them</p>
<p>[00:01:30.993]<br />
are a common circumstance in geological engineering.</p>
<p>[00:01:34.160]<br />
And by small data sets,</p>
<p>[00:01:35.540]<br />
I mean, either actual the small number of values</p>
<p>[00:01:38.210]<br />
that we might have,</p>
<p>[00:01:39.630]<br />
or small in a sense of a sample to volume ratio.</p>
<p>[00:01:42.360]<br />
We’re trying to characterize a very large volume of ground</p>
<p>[00:01:44.598]<br />
with a very small number of data points.</p>
<p>[00:01:48.390]<br />
And it creates two problems these small data sets</p>
<p>[00:01:50.960]<br />
in understanding risk.</p>
<p>[00:01:52.850]<br />
The first is pretty immediate.</p>
<p>[00:01:54.460]<br />
I mean, we have an analysis to do,</p>
<p>[00:01:55.790]<br />
we only have a few data points to choose from,</p>
<p>[00:01:58.580]<br />
and we have to pick an appropriate point</p>
<p>[00:02:00.700]<br />
that we think represents the ground conditions</p>
<p>[00:02:03.060]<br />
or wherever else we’re trying to characterize.</p>
<p>[00:02:05.770]<br />
The second problem is less immediate,</p>
<p>[00:02:08.730]<br />
but ultimately it’s a lot more important.</p>
<p>[00:02:11.840]<br />
And it’s the focus of this talk really.</p>
<p>[00:02:14.370]<br />
Because in selecting this value,</p>
<p>[00:02:18.270]<br />
what we’re doing is we’re making some assumptions</p>
<p>[00:02:20.180]<br />
about that range of data.</p>
<p>[00:02:22.430]<br />
And that goes into our analysis,</p>
<p>[00:02:23.800]<br />
and that goes into our risk quantification.</p>
<p>[00:02:26.130]<br />
And ultimately that goes to our resource allocation</p>
<p>[00:02:29.180]<br />
that we have for mitigating that risk.</p>
<p>[00:02:32.640]<br />
And we have this now line</p>
<p>[00:02:34.970]<br />
between the inherent uncertainty that we’re dealing with</p>
<p>[00:02:37.910]<br />
from these small datasets,</p>
<p>[00:02:39.800]<br />
all the way to the end,</p>
<p>[00:02:40.720]<br />
where we’re actually allocating resources</p>
<p>[00:02:42.610]<br />
to mitigating that problem.</p>
<p>[00:02:46.230]<br />
So it’s really important to understand</p>
<p>[00:02:49.100]<br />
how we think about uncertainty</p>
<p>[00:02:51.430]<br />
so that when we get to this resource allocation,</p>
<p>[00:02:53.770]<br />
we’re actually applying optimal levels</p>
<p>[00:02:55.680]<br />
of mitigation to a problem.</p>
<p>[00:02:58.870]<br />
Next slide.</p>
<p>[00:03:02.760]<br />
So one of the things</p>
<p>[00:03:04.040]<br />
its not that when we say uncertainty,</p>
<p>[00:03:05.970]<br />
it’s not just this big black box of unknowns, this void.</p>
<p>[00:03:11.390]<br />
One of the advantages we have in geomechanics</p>
<p>[00:03:13.680]<br />
is that a lot of our data sets,</p>
<p>[00:03:15.120]<br />
or rather, the types of data and information we use</p>
<p>[00:03:18.260]<br />
are fairly quantitative.</p>
<p>[00:03:20.220]<br />
And so, because of that,</p>
<p>[00:03:21.960]<br />
we can develop this relationship</p>
<p>[00:03:23.300]<br />
between the little things that we know</p>
<p>[00:03:25.660]<br />
that the small data set that we have,</p>
<p>[00:03:27.810]<br />
and this larger uncertainty</p>
<p>[00:03:29.520]<br />
about what the possible range could be.</p>
<p>[00:03:32.280]<br />
We have this idea that variation</p>
<p>[00:03:34.270]<br />
is the range of what we know, whatever that range is.</p>
<p>[00:03:38.560]<br />
And the uncertainty is what we don’t know.</p>
<p>[00:03:41.820]<br />
And given that it’s quantitative often</p>
<p>[00:03:44.030]<br />
we can have an open door or closed end to that uncertainty.</p>
<p>[00:03:48.370]<br />
A lot of times the minimum value is often zero.</p>
<p>[00:03:51.320]<br />
The other end, it can be open in certain circumstances,</p>
<p>[00:03:55.180]<br />
Q values, compressive strength, things like that.</p>
<p>[00:03:58.780]<br />
But at a certain point, it doesn’t matter anymore.</p>
<p>[00:04:00.930]<br />
Once you get past a certain data point</p>
<p>[00:04:03.303]<br />
or restraint value or whatever,</p>
<p>[00:04:05.500]<br />
it doesn’t matter if it’s 350 MPa or 325 MPa,</p>
<p>[00:04:09.980]<br />
it’s strong enough, basically.</p>
<p>[00:04:12.550]<br />
So when we start to overlay these two,</p>
<p>[00:04:15.340]<br />
we see that this is a useful building block now</p>
<p>[00:04:17.760]<br />
for understanding risk,</p>
<p>[00:04:19.960]<br />
because we have this chunk of certainty or knowns</p>
<p>[00:04:23.310]<br />
in the middle,</p>
<p>[00:04:24.143]<br />
and then a chunk of a uncertainty around the sides.</p>
<p>[00:04:28.050]<br />
Go to the next slide, please.</p>
<p>[00:04:31.400]<br />
It’s a simple diagram,</p>
<p>[00:04:32.604]<br />
and it’s going to be the basis for what I’m talking about</p>
<p>[00:04:35.160]<br />
with respect to risk.</p>
<p>[00:04:36.470]<br />
But I’ve started off right away</p>
<p>[00:04:38.040]<br />
with this very idealistic version of what this looks like.</p>
<p>[00:04:41.020]<br />
I’ve got this range of variation that we know in the middle</p>
<p>[00:04:45.905]<br />
bounded by this equal ranges of uncertainty on either end.</p>
<p>[00:04:50.660]<br />
Chances are going to be a lot better actually</p>
<p>[00:04:52.930]<br />
that there’s an asymmetry involved.</p>
<p>[00:04:55.170]<br />
Either there’s going to be a lot more uncertainty</p>
<p>[00:04:57.450]<br />
on one end or the other.</p>
<p>[00:04:59.740]<br />
And this is going to, again,</p>
<p>[00:05:01.060]<br />
influence how we think about risk</p>
<p>[00:05:02.720]<br />
as a function of uncertainty.</p>
<p>[00:05:05.330]<br />
Now I’ve talked about this being quantitative information.</p>
<p>[00:05:09.580]<br />
So it’s easy to think about this</p>
<p>[00:05:11.850]<br />
in terms of a number line and zero at the left side</p>
<p>[00:05:15.490]<br />
and whatever the maximum value is at the right side.</p>
<p>[00:05:19.600]<br />
And that’s okay to think about it that way.</p>
<p>[00:05:22.280]<br />
Since we’re talking about risk though,</p>
<p>[00:05:24.250]<br />
and sometimes low values can be lower risk</p>
<p>[00:05:26.370]<br />
or high values can be lower risk.</p>
<p>[00:05:28.968]<br />
It’s best not to think about it necessarily as numbers</p>
<p>[00:05:32.210]<br />
just as relative better or worse</p>
<p>[00:05:35.090]<br />
in terms of where this certainty lies.</p>
<p>[00:05:40.950]<br />
There’s also a possibility where it could be gapped.</p>
<p>[00:05:43.690]<br />
We could have some sort of chunk of what we know</p>
<p>[00:05:46.300]<br />
and the uncertainty, another chunk of what we know again,</p>
<p>[00:05:49.830]<br />
and then uncertainty on either side of that.</p>
<p>[00:05:52.683]<br />
For the purposes of this talk</p>
<p>[00:05:54.590]<br />
and just to simplify matters a little bit,</p>
<p>[00:05:56.640]<br />
I’m going treat this as basically a bi-modal variation.</p>
<p>[00:06:00.500]<br />
And we just have the same sort of circumstance.</p>
<p>[00:06:03.870]<br />
There’s a range of certainty that we know about or we know,</p>
<p>[00:06:07.470]<br />
and then a range of uncertainty on either side of it.</p>
<p>[00:06:10.240]<br />
Next slide please.</p>
<p>[00:06:13.910]<br />
So now we want to talk about upside or downside asymmetry.</p>
<p>[00:06:17.210]<br />
So I’ve talked about this idea</p>
<p>[00:06:18.590]<br />
that we can have significantly</p>
<p>[00:06:21.450]<br />
more uncertainty on one side or the other</p>
<p>[00:06:23.640]<br />
of our range of what we know.</p>
<p>[00:06:26.850]<br />
And to do that,</p>
<p>[00:06:27.683]<br />
we want to think about this critical value,</p>
<p>[00:06:29.820]<br />
this concept of a critical value.</p>
<p>[00:06:31.470]<br />
This is the value at which</p>
<p>[00:06:34.550]<br />
if you have an input value,</p>
<p>[00:06:36.270]<br />
you’re going to get an output value.</p>
<p>[00:06:38.150]<br />
And if you put in a lower input value,</p>
<p>[00:06:39.840]<br />
you will get a worse answer.</p>
<p>[00:06:40.950]<br />
Or anything to the left of that will be worse.</p>
<p>[00:06:44.960]<br />
Anything to the right is better.</p>
<p>[00:06:46.370]<br />
So this is the value.</p>
<p>[00:06:48.030]<br />
I mean, probably the easiest way to think about it</p>
<p>[00:06:49.690]<br />
is, say a stability analysis,</p>
<p>[00:06:52.310]<br />
and you need a certain compressive strength</p>
<p>[00:06:53.840]<br />
to produce a factor of safety.</p>
<p>[00:06:56.750]<br />
So if you have a less compressive strength</p>
<p>[00:06:58.180]<br />
or a lower compressive strength,</p>
<p>[00:06:59.730]<br />
you’ll get a worse factor of safety,</p>
<p>[00:07:01.461]<br />
and a higher compressive strength</p>
<p>[00:07:03.530]<br />
is a better factor of safety.</p>
<p>[00:07:05.810]<br />
So it’s this critical value.</p>
<p>[00:07:06.957]<br />
And now we can start to see where does our variation lie</p>
<p>[00:07:10.260]<br />
and versus where does our uncertainty lie</p>
<p>[00:07:12.160]<br />
relative to this critical value?</p>
<p>[00:07:15.111]<br />
So we can have asymmetric downside risk,</p>
<p>[00:07:18.090]<br />
we’re basically what we don’t know makes the problem worse,</p>
<p>[00:07:22.360]<br />
or asymmetric upside risk,</p>
<p>[00:07:24.070]<br />
what we don’t know makes the problem better</p>
<p>[00:07:26.030]<br />
relative to this critical value.</p>
<p>[00:07:29.290]<br />
Next slide, please.</p>
<p>[00:07:32.830]<br />
Now we want to talk about magnitude of uncertainty.</p>
<p>[00:07:35.750]<br />
How big is this range?</p>
<p>[00:07:38.170]<br />
We can have of course,</p>
<p>[00:07:39.480]<br />
significant downside uncertainty</p>
<p>[00:07:41.820]<br />
in the case that I’ve shown.</p>
<p>[00:07:43.490]<br />
There’s a lot of uncertainty below this critical value,</p>
<p>[00:07:46.676]<br />
or we can have minor downside uncertainty.</p>
<p>[00:07:49.500]<br />
There’s just a little bit.</p>
<p>[00:07:51.340]<br />
Again, if we think about</p>
<p>[00:07:52.921]<br />
a lot of different geomechanical data,</p>
<p>[00:07:55.900]<br />
the minimum value is zero.</p>
<p>[00:07:57.660]<br />
So if the far lowest known point is slightly more than zero,</p>
<p>[00:08:02.210]<br />
yeah, there’s some uncertainty,</p>
<p>[00:08:03.250]<br />
maybe there’s a value that would fit into that range,</p>
<p>[00:08:05.660]<br />
but it’s a pretty small range</p>
<p>[00:08:07.970]<br />
between zero and whatever our minimum value is.</p>
<p>[00:08:11.020]<br />
So again, for the purposes of this talk</p>
<p>[00:08:13.420]<br />
and to keep things simple,</p>
<p>[00:08:15.070]<br />
if we have minor downside uncertainty,</p>
<p>[00:08:17.020]<br />
I’m not even going to think about that as uncertainty,</p>
<p>[00:08:19.270]<br />
it’s just treated with extending your variation</p>
<p>[00:08:23.230]<br />
a little bit more.</p>
<p>[00:08:24.730]<br />
Really the purpose of this</p>
<p>[00:08:25.860]<br />
and talking about risk and uncertainty</p>
<p>[00:08:27.560]<br />
is talking about circumstances</p>
<p>[00:08:29.250]<br />
where we have significant</p>
<p>[00:08:30.930]<br />
either downside or upside uncertainty,</p>
<p>[00:08:33.470]<br />
where we have a lot of unknowns on one side or the other</p>
<p>[00:08:36.850]<br />
of that critical value.</p>
<p>[00:08:38.810]<br />
Next slide, please.</p>
<p>[00:08:42.110]<br />
Now, of course,</p>
<p>[00:08:42.943]<br />
there’s a sensitivity too that we have to consider.</p>
<p>[00:08:45.320]<br />
This is how sensitive is the output value</p>
<p>[00:08:48.650]<br />
to a change in the input value.</p>
<p>[00:08:51.260]<br />
We can have circumstances</p>
<p>[00:08:52.910]<br />
where our output is insensitive and reasonably linear</p>
<p>[00:08:57.339]<br />
as we make these changes and gradual changes</p>
<p>[00:09:00.380]<br />
in putting in a higher or lower values</p>
<p>[00:09:07.360]<br />
relative to this critical value,</p>
<p>[00:09:09.220]<br />
we don’t see much change or an answer,</p>
<p>[00:09:11.630]<br />
or we can have very sensitive</p>
<p>[00:09:13.410]<br />
and potentially non-linear answers relative to inputs.</p>
<p>[00:09:17.680]<br />
We can start to see</p>
<p>[00:09:19.980]<br />
that we either get a significant change</p>
<p>[00:09:23.222]<br />
in the slope of that output,</p>
<p>[00:09:25.830]<br />
or we have just a very significant sensitivity</p>
<p>[00:09:30.410]<br />
at the end of the day.</p>
<p>[00:09:31.730]<br />
So either one is a cause for concern in this case.</p>
<p>[00:09:36.740]<br />
Next slide, please.</p>
<p>[00:09:40.280]<br />
And then of course, risk.</p>
<p>[00:09:41.927]<br />
We have all of the different things around the probability</p>
<p>[00:09:44.750]<br />
and the range of inputs,</p>
<p>[00:09:46.200]<br />
and essentially what that value is going to look like.</p>
<p>[00:09:48.560]<br />
And the other half of risk is the of course, consequences.</p>
<p>[00:09:51.810]<br />
And our consequences can like sensitivity</p>
<p>[00:09:54.450]<br />
be low to moderate.</p>
<p>[00:09:56.460]<br />
As we’re changing that input value,</p>
<p>[00:10:00.020]<br />
we don’t really see a change in the consequence that much.</p>
<p>[00:10:04.210]<br />
So say again factor of safety in a stability analysis</p>
<p>[00:10:07.850]<br />
is our example.</p>
<p>[00:10:12.200]<br />
And we’re reducing that input material strength,</p>
<p>[00:10:15.440]<br />
and we’re getting a failure.</p>
<p>[00:10:16.780]<br />
But the size of the failure</p>
<p>[00:10:18.090]<br />
is not really changing,</p>
<p>[00:10:19.670]<br />
the run-out distance isn’t changing.</p>
<p>[00:10:21.780]<br />
We’re not really seeing huge differences</p>
<p>[00:10:24.640]<br />
in the consequences of that,</p>
<p>[00:10:26.900]<br />
even though the factor of safety might be dropping,</p>
<p>[00:10:28.780]<br />
it’s not really having an effect</p>
<p>[00:10:30.290]<br />
on what the impacts of that would be.</p>
<p>[00:10:33.213]<br />
So we can have this lower,</p>
<p>[00:10:35.330]<br />
and again, linear consequences</p>
<p>[00:10:37.500]<br />
as we go down this potential range of uncertainty,</p>
<p>[00:10:42.396]<br />
or we can have very high consequences,</p>
<p>[00:10:44.620]<br />
and even non-linear consequences again.</p>
<p>[00:10:48.060]<br />
Now one note on the consequences,</p>
<p>[00:10:50.150]<br />
we have both downside consequences.</p>
<p>[00:10:53.230]<br />
These are often going to take</p>
<p>[00:10:54.230]<br />
the form of unmitigated liability.</p>
<p>[00:10:56.560]<br />
And why I say liability instead of risk,</p>
<p>[00:10:58.740]<br />
is that it’s sort of the next piece.</p>
<p>[00:11:02.740]<br />
Things could be worse than we assume,</p>
<p>[00:11:05.410]<br />
higher risks, and then these risks</p>
<p>[00:11:07.330]<br />
haven’t been attenuated or mitigated</p>
<p>[00:11:09.040]<br />
because we aren’t aware of them,</p>
<p>[00:11:11.610]<br />
and that’s going to create a liability.</p>
<p>[00:11:13.290]<br />
So that’s the downside consequences,</p>
<p>[00:11:15.490]<br />
this unmitigated liability.</p>
<p>[00:11:17.830]<br />
And the upside consequences are going to be more</p>
<p>[00:11:19.950]<br />
in the form of opportunity costs.</p>
<p>[00:11:22.630]<br />
Essentially we could have had a leaner, meaner construction</p>
<p>[00:11:27.490]<br />
of whatever sort.</p>
<p>[00:11:28.323]<br />
We didn’t have to have a to have slope angle that was that shallow.</p>
<p>[00:11:32.660]<br />
We didn’t have to have an embankment that was that big.</p>
<p>[00:11:35.290]<br />
We dedicated resources to something</p>
<p>[00:11:37.030]<br />
that we didn’t need to necessarily</p>
<p>[00:11:38.880]<br />
to achieve our desired outcome in terms of safety.</p>
<p>[00:11:42.900]<br />
Next slide, please.</p>
<p>[00:11:47.000]<br />
So given this construct with sensitivity,</p>
<p>[00:11:50.230]<br />
greater or lesser than,</p>
<p>[00:11:52.210]<br />
the asymmetry in the outcome upside or downside,</p>
<p>[00:11:57.370]<br />
and then the consequences either higher or lower.</p>
<p>[00:12:00.310]<br />
We have this box of possibilities</p>
<p>[00:12:02.040]<br />
in terms of these risks scenarios now,</p>
<p>[00:12:04.410]<br />
and uncertainty scenarios</p>
<p>[00:12:06.790]<br />
that we’ve got eight different circumstances</p>
<p>[00:12:09.510]<br />
that we can look at</p>
<p>[00:12:11.020]<br />
in terms of all of these different ways</p>
<p>[00:12:12.760]<br />
we can think about risk as a function of uncertainty.</p>
<p>[00:12:17.530]<br />
Next slide, please.</p>
<p>[00:12:22.180]<br />
So we’ve talked about now</p>
<p>[00:12:23.655]<br />
the first two pieces of that flow diagram</p>
<p>[00:12:28.070]<br />
that I showed in the earlier slide</p>
<p>[00:12:30.220]<br />
with uncertainty and assumptions.</p>
<p>[00:12:32.630]<br />
That’s how we start to think about risk.</p>
<p>[00:12:34.740]<br />
Now we talk about the analysis and the risk mitigation,</p>
<p>[00:12:37.670]<br />
and this is through the tools that we use.</p>
<p>[00:12:40.521]<br />
These are all these different tools</p>
<p>[00:12:42.250]<br />
that are available to us as geotechnical engineers</p>
<p>[00:12:44.670]<br />
to address this uncertainty.</p>
<p>[00:12:47.050]<br />
How do we think about uncertainty?</p>
<p>[00:12:50.980]<br />
I won’t say that this is the definitive way</p>
<p>[00:12:53.280]<br />
to think about these tools,</p>
<p>[00:12:55.620]<br />
but I will say that there is a continuum of sorts</p>
<p>[00:12:58.410]<br />
between all these different tools that we have.</p>
<p>[00:13:00.390]<br />
And this slide isn’t meant to capture all the tools</p>
<p>[00:13:02.830]<br />
that are available to us as geotechnical engineers.</p>
<p>[00:13:05.670]<br />
But to talk about them in terms of these broad categories,</p>
<p>[00:13:10.310]<br />
where we have tools</p>
<p>[00:13:11.520]<br />
that are based in inductive reasoning and inference,</p>
<p>[00:13:15.600]<br />
these are things like the first picture</p>
<p>[00:13:17.980]<br />
where we have something about A that we don’t know.</p>
<p>[00:13:21.110]<br />
This could be again, a material strength.</p>
<p>[00:13:23.850]<br />
We know a lot about A, or something about A,</p>
<p>[00:13:27.880]<br />
we tend to know a lot more about B</p>
<p>[00:13:29.670]<br />
including that material strength</p>
<p>[00:13:31.070]<br />
this target thing we want to know.</p>
<p>[00:13:32.510]<br />
And A and B share enough characteristics</p>
<p>[00:13:35.630]<br />
that we can assume that whatever material strength B has,</p>
<p>[00:13:38.560]<br />
A has the same strength or similar strength.</p>
<p>[00:13:42.320]<br />
We have tools that fall into this</p>
<p>[00:13:43.730]<br />
sort of proportional relationship.</p>
<p>[00:13:46.036]<br />
A is somehow relative to B,</p>
<p>[00:13:48.180]<br />
think about, we have a few compressive strength samples</p>
<p>[00:13:51.810]<br />
where we’ve incurred the higher cost</p>
<p>[00:13:53.060]<br />
to do the compressive strength testing</p>
<p>[00:13:55.380]<br />
and a whole lot of point load samples.</p>
<p>[00:13:58.250]<br />
And there’s a proportionality between those two.</p>
<p>[00:14:00.270]<br />
So we can look at the range</p>
<p>[00:14:02.250]<br />
of compressive strength variation</p>
<p>[00:14:05.800]<br />
as a function of the range of point load strength variation.</p>
<p>[00:14:11.160]<br />
There’s a lot more, of course,</p>
<p>[00:14:12.320]<br />
again, these are not meant to be exhaustive lists</p>
<p>[00:14:15.350]<br />
of all these different tools,</p>
<p>[00:14:16.440]<br />
just to get the sense of what an inference</p>
<p>[00:14:20.220]<br />
or inductive reasoning type of tool looks like.</p>
<p>[00:14:23.720]<br />
We have parametric tools, or these are basically</p>
<p>[00:14:26.880]<br />
this one of the kitchen-sink approach.</p>
<p>[00:14:29.310]<br />
We’re throwing a lot of different things</p>
<p>[00:14:31.460]<br />
in sampling from them into a bin.</p>
<p>[00:14:34.840]<br />
This can be a lot of different types of variables.</p>
<p>[00:14:37.330]<br />
And we’re trying to come up</p>
<p>[00:14:38.470]<br />
with some sort of parametric analysis</p>
<p>[00:14:40.980]<br />
based on Monte Carlo, Latin Hypercube sampling, whatever</p>
<p>[00:14:44.560]<br />
that produces this range of outcomes.</p>
<p>[00:14:46.900]<br />
And we can start to look at that range of outcomes</p>
<p>[00:14:48.970]<br />
and make some conclusion from that.</p>
<p>[00:14:52.730]<br />
There’s a lot of things around say subjective probability</p>
<p>[00:14:55.810]<br />
that might fall into this as well.</p>
<p>[00:14:59.220]<br />
A lot of different tools</p>
<p>[00:15:00.240]<br />
where we’re basically just looking at</p>
<p>[00:15:02.640]<br />
what the distributions are</p>
<p>[00:15:04.170]<br />
across a lot of different ranges of our variables.</p>
<p>[00:15:08.050]<br />
And then we have these direct</p>
<p>[00:15:09.130]<br />
or deductive reasoning types of tools</p>
<p>[00:15:11.977]<br />
where we’re just either looking at</p>
<p>[00:15:14.340]<br />
what information we have, this is the variation,</p>
<p>[00:15:17.180]<br />
or maybe we’re extrapolating from something.</p>
<p>[00:15:19.460]<br />
A lot of times frequency or recurrence interval</p>
<p>[00:15:23.320]<br />
might fall into these types of deductive reasoning tools.</p>
<p>[00:15:27.220]<br />
We have a bunch of data from a time history,</p>
<p>[00:15:29.455]<br />
and we’re going to extrapolate that out a little bit</p>
<p>[00:15:31.867]<br />
and pretty much this is what we can assume</p>
<p>[00:15:33.930]<br />
about the circumstance from the information we have.</p>
<p>[00:15:37.780]<br />
We could also assume some cases, just the minimum value.</p>
<p>[00:15:40.790]<br />
If we know that the range is bound in a certain way,</p>
<p>[00:15:45.940]<br />
it starts at zero, it goes to a hundred,</p>
<p>[00:15:48.120]<br />
maybe we pick one or the other</p>
<p>[00:15:49.340]<br />
as far as an upper or lower bound to what that would be.</p>
<p>[00:15:53.590]<br />
So these things as well,</p>
<p>[00:15:54.730]<br />
I’m going to argue, fall on a continuum.</p>
<p>[00:15:56.850]<br />
There’s not any necessarily hard lines between them but—</p>
<p>[00:16:01.940]<br />
Next slide, please.</p>
<p>[00:16:02.990]<br />
We will talk about the relative strengths</p>
<p>[00:16:06.140]<br />
and weaknesses of them.</p>
<p>[00:16:07.510]<br />
And again, not meant to be an exhaustive list</p>
<p>[00:16:09.780]<br />
just to illustrate that each of these has a place and a use</p>
<p>[00:16:14.480]<br />
in terms of addressing uncertainty.</p>
<p>[00:16:17.250]<br />
With inference and inductive reasoning,</p>
<p>[00:16:20.370]<br />
a lot of times we’re using a lot of our knowledge</p>
<p>[00:16:23.610]<br />
and understanding as geotechnical engineers</p>
<p>[00:16:25.590]<br />
to relate one thing to another,</p>
<p>[00:16:27.940]<br />
or use some other bit of data to modify</p>
<p>[00:16:33.330]<br />
or increase the precision of our estimate.</p>
<p>[00:16:35.342]<br />
We could say, if we don’t know a material strength,</p>
<p>[00:16:39.250]<br />
we can assume it’s zero, that’s pretty conservative.</p>
<p>[00:16:42.140]<br />
We don’t want to do that necessarily,</p>
<p>[00:16:43.990]<br />
so we’re using inference</p>
<p>[00:16:46.140]<br />
to increase the precision of that estimate.</p>
<p>[00:16:49.370]<br />
Of course, the weakness of this,</p>
<p>[00:16:50.620]<br />
it it’s really based on knowledge</p>
<p>[00:16:52.380]<br />
from practitioner to practitioner, that can vary.</p>
<p>[00:16:55.340]<br />
I might be really good at estimating material strength</p>
<p>[00:16:57.970]<br />
from all these other materials strengths that I’m aware of,</p>
<p>[00:17:00.640]<br />
the next person has maybe more of a limited expertise</p>
<p>[00:17:05.340]<br />
in that area.</p>
<p>[00:17:06.670]<br />
And you’re going to get very different answers</p>
<p>[00:17:08.190]<br />
from inference and inductive reasoning</p>
<p>[00:17:10.280]<br />
from practitioner to practitioner,</p>
<p>[00:17:12.110]<br />
probably the basis of a lot of arguments</p>
<p>[00:17:13.780]<br />
that we have as geological engineers.</p>
<p>[00:17:18.450]<br />
The direct or deductive reasoning,</p>
<p>[00:17:20.760]<br />
the strength there is that</p>
<p>[00:17:22.230]<br />
since you’re assuming either from what you know,</p>
<p>[00:17:26.250]<br />
or from more importantly from some end value of this,</p>
<p>[00:17:33.420]<br />
you sort of covered all the bases.</p>
<p>[00:17:35.330]<br />
You’re not going to be surprised</p>
<p>[00:17:37.140]<br />
by something that wasn’t captured in your assumption.</p>
<p>[00:17:40.130]<br />
The weakness of course,</p>
<p>[00:17:41.040]<br />
is that these can be fairly conservative estimates.</p>
<p>[00:17:44.730]<br />
With the parametric tools,</p>
<p>[00:17:48.030]<br />
the strength is that it’s actually</p>
<p>[00:17:50.260]<br />
kind of drawing on the strengths</p>
<p>[00:17:51.640]<br />
of both inference and inductive reasoning</p>
<p>[00:17:54.640]<br />
as well as direct and deductive reasoning.</p>
<p>[00:17:57.440]<br />
And so it’s pulling the best of each of those.</p>
<p>[00:18:00.770]<br />
The weakness is that</p>
<p>[00:18:02.230]<br />
this can require considerable time expertise.</p>
<p>[00:18:04.942]<br />
You would have to pull from a lot of different people.</p>
<p>[00:18:07.880]<br />
You’re going to have to deal</p>
<p>[00:18:08.860]<br />
with some of those issues around,</p>
<p>[00:18:10.450]<br />
again, both of the weaknesses of each method.</p>
<p>[00:18:13.760]<br />
The other one that it can cause</p>
<p>[00:18:16.490]<br />
is that you end up with this range of possible outcomes</p>
<p>[00:18:20.180]<br />
that’s going to vary</p>
<p>[00:18:21.900]<br />
from some extreme adverse outcome to extreme good outcome,</p>
<p>[00:18:27.224]<br />
and you’re going to have to make some decisions</p>
<p>[00:18:29.780]<br />
around which one’s going to be the appropriate outcome.</p>
<p>[00:18:32.610]<br />
How do you decide?</p>
<p>[00:18:33.500]<br />
Do you have a cluster of outcomes around a central value?</p>
<p>[00:18:37.490]<br />
And that’s a good thing.</p>
<p>[00:18:38.720]<br />
Or do you have these long tails</p>
<p>[00:18:40.790]<br />
that you have to make some decisions about?</p>
<p>[00:18:43.590]<br />
It can sort of solve some of the problems</p>
<p>[00:18:46.127]<br />
of inference or deductive reasoning on the front end,</p>
<p>[00:18:50.300]<br />
but cause more problems on the backend.</p>
<p>[00:18:53.230]<br />
So no one tool is perfect,</p>
<p>[00:18:55.990]<br />
but they all have their advantages and disadvantages.</p>
<p>[00:18:59.460]<br />
So next slide.</p>
<p>[00:19:02.580]<br />
So now we’ve compartmentalized</p>
<p>[00:19:05.200]<br />
all these different circumstances of uncertainty and risk,</p>
<p>[00:19:10.340]<br />
and now we have the different tools that we apply.</p>
<p>[00:19:12.930]<br />
And we’re going to start talking about</p>
<p>[00:19:15.100]<br />
how each of those tools fits</p>
<p>[00:19:17.030]<br />
each of these different circumstances.</p>
<p>[00:19:19.030]<br />
So we have that box of possibilities</p>
<p>[00:19:21.270]<br />
from the previous slide,</p>
<p>[00:19:22.190]<br />
and we split it,</p>
<p>[00:19:23.340]<br />
and we’re looking at the downside on the left side,</p>
<p>[00:19:26.140]<br />
and the upside on the right side.</p>
<p>[00:19:28.130]<br />
We can start to look at how these tools</p>
<p>[00:19:30.330]<br />
apply to these different circumstances</p>
<p>[00:19:33.140]<br />
as a function of sensitivity and consequence,</p>
<p>[00:19:36.490]<br />
and then upside or downside risk.</p>
<p>[00:19:39.310]<br />
So I’d like to talk about these for a little bit,</p>
<p>[00:19:41.950]<br />
and I’ll start with the downside risk</p>
<p>[00:19:44.160]<br />
in the lower left-hand corner.</p>
<p>[00:19:47.070]<br />
We have a situation where we have pretty low consequences,</p>
<p>[00:19:50.350]<br />
pretty low sensitivity, or in sensitivity,</p>
<p>[00:19:53.720]<br />
it’s downside risk, but essentially you’re not going to,</p>
<p>[00:19:57.380]<br />
because of this insensitivity and lower consequences,</p>
<p>[00:20:02.360]<br />
you can assume fairly extreme values</p>
<p>[00:20:05.530]<br />
without really any cost in terms of allocation of resources.</p>
<p>[00:20:09.220]<br />
So that’s a pretty good place for that tool to sit.</p>
<p>[00:20:12.650]<br />
In this middle band,</p>
<p>[00:20:14.087]<br />
we have either the higher consequences</p>
<p>[00:20:17.043]<br />
with the higher sensitivity,</p>
<p>[00:20:19.105]<br />
(coughing) excuse me.</p>
<p>[00:20:20.765]<br />
Some of these inductive tools are going to be more important</p>
<p>[00:20:24.680]<br />
because now either we have to think about consequences</p>
<p>[00:20:29.690]<br />
or we have to think about that sensitivity.</p>
<p>[00:20:31.330]<br />
We do want a little bit more precision</p>
<p>[00:20:33.394]<br />
in how we approach this.</p>
<p>[00:20:34.890]<br />
We want to be aware of implausible or extreme values</p>
<p>[00:20:40.320]<br />
and how those might affect our answer,</p>
<p>[00:20:42.065]<br />
but we don’t want to let them influence our answer too much</p>
<p>[00:20:46.490]<br />
because they could lead to such an extreme outcome</p>
<p>[00:20:49.250]<br />
in our risk assessment</p>
<p>[00:20:50.410]<br />
that we’re, again, misallocating resources.</p>
<p>[00:20:52.610]<br />
So we want to start using some of these inference tools</p>
<p>[00:20:55.140]<br />
to increase the precision of our input assumptions.</p>
<p>[00:20:59.400]<br />
And then finally, when we get to the upper right side</p>
<p>[00:21:02.180]<br />
on the upper right quadrant on the downside risk,</p>
<p>[00:21:06.698]<br />
it really speaks to,</p>
<p>[00:21:08.150]<br />
we have a lot of sensitivity, high consequences.</p>
<p>[00:21:10.870]<br />
We need to look at this parametric approach</p>
<p>[00:21:12.700]<br />
because we want to capture potentially some relationships</p>
<p>[00:21:15.970]<br />
between different, either within a variable</p>
<p>[00:21:19.410]<br />
or due to non-linear responses,</p>
<p>[00:21:22.330]<br />
or maybe some combination of variables</p>
<p>[00:21:24.315]<br />
that may not be intuitive.</p>
<p>[00:21:26.720]<br />
We really want to see what that full range of outcomes</p>
<p>[00:21:29.180]<br />
looks like.</p>
<p>[00:21:30.770]<br />
So for the upside,</p>
<p>[00:21:33.780]<br />
we have a similar set of different tools</p>
<p>[00:21:36.433]<br />
that are going to be applied to these different compartments,</p>
<p>[00:21:40.340]<br />
but a slight difference.</p>
<p>[00:21:41.602]<br />
If we start in the lower right-hand corner of this time,</p>
<p>[00:21:46.490]<br />
we have low sensitivity, low consequences,</p>
<p>[00:21:49.060]<br />
but because it’s more of a matter of opportunity costs,</p>
<p>[00:21:52.790]<br />
we want to use this parametric approach</p>
<p>[00:21:54.830]<br />
to understand those a little bit better.</p>
<p>[00:21:58.450]<br />
There’s some value in looking at those</p>
<p>[00:22:00.850]<br />
so that we’re understanding</p>
<p>[00:22:01.870]<br />
that we’re again, allocating resources appropriately.</p>
<p>[00:22:05.380]<br />
For these middle two boxes</p>
<p>[00:22:07.263]<br />
where we have the higher consequences, but less sensitivity,</p>
<p>[00:22:11.730]<br />
these again, these indirect and inductive reasoning methods</p>
<p>[00:22:14.950]<br />
are important to increase that precision around our answer.</p>
<p>[00:22:20.800]<br />
But once we get into the lower consequences</p>
<p>[00:22:23.620]<br />
with greater sensitivity,</p>
<p>[00:22:25.360]<br />
the direct and deductive approaches</p>
<p>[00:22:27.320]<br />
are more important to use.</p>
<p>[00:22:30.040]<br />
And of course, when we get up</p>
<p>[00:22:31.130]<br />
into the upper left quadrant there</p>
<p>[00:22:34.370]<br />
with greater sensitivity and higher consequences,</p>
<p>[00:22:37.760]<br />
we want to use those parametric approaches</p>
<p>[00:22:39.616]<br />
again, to understand</p>
<p>[00:22:41.520]<br />
if there’s some sort of non-intuitive outcome</p>
<p>[00:22:44.810]<br />
that we can experience,</p>
<p>[00:22:46.795]<br />
or to look at whether those outcomes are clustered</p>
<p>[00:22:50.400]<br />
around some sort of central value</p>
<p>[00:22:52.100]<br />
or have these longer tails</p>
<p>[00:22:53.610]<br />
that might be important to consider.</p>
<p>[00:22:55.650]<br />
Again, it speaks to</p>
<p>[00:22:57.600]<br />
how do we start to look at the tools</p>
<p>[00:23:00.410]<br />
versus the circumstance</p>
<p>[00:23:01.561]<br />
to properly mitigate risk</p>
<p>[00:23:05.930]<br />
or allocate resources to mitigate risk.</p>
<p>[00:23:08.730]<br />
Next slide, please.</p>
<p>[00:23:12.510]<br />
So where do we come to with all this?</p>
<p>[00:23:15.450]<br />
We can use this relationship between what we do know</p>
<p>[00:23:19.590]<br />
and the range of what we may not know or don’t know</p>
<p>[00:23:23.156]<br />
to think about how to characterize uncertainty</p>
<p>[00:23:26.930]<br />
relative to risk.</p>
<p>[00:23:29.520]<br />
And you can agree or disagree</p>
<p>[00:23:31.330]<br />
with any parts of this discussion</p>
<p>[00:23:34.810]<br />
or any parts of my presentation.</p>
<p>[00:23:38.160]<br />
What it does come down to</p>
<p>[00:23:39.410]<br />
again, is this fundamental idea</p>
<p>[00:23:42.290]<br />
that we can discuss this and go on and on</p>
<p>[00:23:45.540]<br />
and talk about our different approaches and whatnot.</p>
<p>[00:23:50.460]<br />
But we do have to think at the end of the day</p>
<p>[00:23:52.340]<br />
about that allocation of resources.</p>
<p>[00:23:54.800]<br />
And so the purpose of all this</p>
<p>[00:23:56.540]<br />
is just to highlight</p>
<p>[00:23:57.730]<br />
that there is this structure to uncertainty.</p>
<p>[00:24:01.081]<br />
There are impacts that the tools</p>
<p>[00:24:03.940]<br />
that we use as geological engineers have</p>
<p>[00:24:06.160]<br />
to how we think about that.</p>
<p>[00:24:07.750]<br />
And when we start to marry those two</p>
<p>[00:24:10.230]<br />
and look at the circumstances of uncertainty</p>
<p>[00:24:13.140]<br />
and the tools that we have for addressing it,</p>
<p>[00:24:16.790]<br />
we really want to make sure that that’s a good marriage</p>
<p>[00:24:19.900]<br />
in terms of producing this optimal allocation of resources</p>
<p>[00:24:24.970]<br />
at the end of the day.</p>
<p>[00:24:26.520]<br />
And that’s really the message of this entire talk.</p>
<p>[00:24:31.260]<br />
Next slide, please.</p>
<p>[00:24:33.870]<br />
Thank you for your time and attention.</p>
<wpml_invalid_tag original=»PHA+» />[00:24:35.258]<br />
(upbeat music)

Close Menu