WEBVTT 00:00:00.000 --> 00:00:01.992 align:middle line:90% [SQUEAKING] 00:00:01.992 --> 00:00:03.984 align:middle line:90% [RUSTLING] 00:00:03.984 --> 00:00:07.470 align:middle line:90% [CLICKING] 00:00:07.470 --> 00:00:12.970 align:middle line:90% 00:00:12.970 --> 00:00:14.930 align:middle line:84% RICHARD DE NEUFVILLE: What is simulation? 00:00:14.930 --> 00:00:19.600 align:middle line:84% Simulation is a way of replicating the outcomes 00:00:19.600 --> 00:00:21.880 align:middle line:90% of an uncertain process. 00:00:21.880 --> 00:00:25.100 align:middle line:84% That is, we know the forecast may be wrong, 00:00:25.100 --> 00:00:27.160 align:middle line:84% we're looking at different possibilities, 00:00:27.160 --> 00:00:30.910 align:middle line:84% and if this were to happen in the terms of the forecast 00:00:30.910 --> 00:00:33.430 align:middle line:84% in, say, the demand for cars or the amount 00:00:33.430 --> 00:00:37.030 align:middle line:84% of liquid fuels consumed, here's the way we would deal-- 00:00:37.030 --> 00:00:39.520 align:middle line:84% here's the way the system would deal with it, here's 00:00:39.520 --> 00:00:42.610 align:middle line:84% the value of the system or the sales or the losses 00:00:42.610 --> 00:00:43.810 align:middle line:90% or whatever. 00:00:43.810 --> 00:00:47.350 align:middle line:84% It's often called Monte Carlo simulation. 00:00:47.350 --> 00:00:49.330 align:middle line:90% Why is this? 00:00:49.330 --> 00:00:55.090 align:middle line:84% Because Monte Carlo is a European gambling hub. 00:00:55.090 --> 00:01:03.220 align:middle line:84% It's the Macau or the Las Vegas of the-- actually of the 1900s 00:01:03.220 --> 00:01:04.540 align:middle line:90% and 1920s. 00:01:04.540 --> 00:01:08.650 align:middle line:84% It was the place at that time, and therefore, it 00:01:08.650 --> 00:01:10.930 align:middle line:90% is designed to show about-- 00:01:10.930 --> 00:01:13.780 align:middle line:84% it's associated with luck and chance and uncertainty. 00:01:13.780 --> 00:01:19.150 align:middle line:84% That's why it's called Monte Carlo simulation as opposed 00:01:19.150 --> 00:01:23.480 align:middle line:90% to a deterministic simulation. 00:01:23.480 --> 00:01:27.830 align:middle line:84% So we don't use the Monte Carlo very often because it just 00:01:27.830 --> 00:01:31.970 align:middle line:84% adds to the language, it makes the words longer. 00:01:31.970 --> 00:01:37.100 align:middle line:84% But it's the uncertain aspect that we're talking about. 00:01:37.100 --> 00:01:38.410 align:middle line:90% So what it does. 00:01:38.410 --> 00:01:42.880 align:middle line:84% It provides us with a sample that 00:01:42.880 --> 00:01:46.930 align:middle line:84% is a distribution of what might happen if-- 00:01:46.930 --> 00:01:49.220 align:middle line:90% under some uncertainty. 00:01:49.220 --> 00:01:53.320 align:middle line:84% So think of it like you're flipping a coin 00:01:53.320 --> 00:01:57.040 align:middle line:84% and it comes up heads, heads, tails, tails, heads, heads, 00:01:57.040 --> 00:01:59.050 align:middle line:90% tails, heads, tails, whatever. 00:01:59.050 --> 00:02:01.600 align:middle line:84% That it will give you a distribution 00:02:01.600 --> 00:02:06.310 align:middle line:84% of the frequency of the possible outcomes that will occur. 00:02:06.310 --> 00:02:14.400 align:middle line:84% And the way it operates depends upon the underlying uncertainty 00:02:14.400 --> 00:02:15.540 align:middle line:90% with a process. 00:02:15.540 --> 00:02:19.560 align:middle line:84% In the coin toss, it is the probability 00:02:19.560 --> 00:02:22.950 align:middle line:84% that the coin comes up heads or tails, which we normally 00:02:22.950 --> 00:02:27.330 align:middle line:84% call 50%, but maybe there is a particular bias 00:02:27.330 --> 00:02:29.580 align:middle line:84% in the coin that you're flipping because of a neck 00:02:29.580 --> 00:02:32.450 align:middle line:84% or whatever, it might not be that. 00:02:32.450 --> 00:02:35.880 align:middle line:84% Now Monte Carlo simulation as it's 00:02:35.880 --> 00:02:41.610 align:middle line:84% practiced can work with any distribution, any distribution. 00:02:41.610 --> 00:02:46.590 align:middle line:84% It's no need to go for some analytically intractable 00:02:46.590 --> 00:02:49.590 align:middle line:84% distribution such as the so-called normal distribution 00:02:49.590 --> 00:02:54.630 align:middle line:84% or j-distribution or uniform distribution 00:02:54.630 --> 00:02:57.840 align:middle line:90% or binomial distribution. 00:02:57.840 --> 00:02:59.940 align:middle line:90% It can be regular. 00:02:59.940 --> 00:03:01.920 align:middle line:84% That is, one of these normal distributions. 00:03:01.920 --> 00:03:04.560 align:middle line:84% Or irregular, what you observe, which 00:03:04.560 --> 00:03:08.250 align:middle line:84% might be the distribution of rain or monsoons in India 00:03:08.250 --> 00:03:13.740 align:middle line:84% which have a different intensity and timing and so on and there 00:03:13.740 --> 00:03:17.190 align:middle line:84% are reasons why they might not be nice and-- 00:03:17.190 --> 00:03:19.470 align:middle line:90% nice, simple distributions. 00:03:19.470 --> 00:03:21.120 align:middle line:90% And it can be continuous or not. 00:03:21.120 --> 00:03:22.080 align:middle line:90% It could be bimodal. 00:03:22.080 --> 00:03:25.170 align:middle line:84% It could be mostly about one outcome 00:03:25.170 --> 00:03:29.580 align:middle line:84% or mostly about another outcome, sort of a two-state device 00:03:29.580 --> 00:03:30.910 align:middle line:90% with uncertainties. 00:03:30.910 --> 00:03:36.390 align:middle line:84% So the simulation works for any distribution that's available 00:03:36.390 --> 00:03:41.260 align:middle line:84% and it can be worked for lots of things being uncertain. 00:03:41.260 --> 00:03:45.210 align:middle line:84% So if you're thinking about running a company 00:03:45.210 --> 00:03:47.130 align:middle line:84% and you are thinking about your income, 00:03:47.130 --> 00:03:49.950 align:middle line:84% there might be distributions of the demand for it 00:03:49.950 --> 00:03:51.270 align:middle line:90% in terms of how many. 00:03:51.270 --> 00:03:54.450 align:middle line:84% There might be a distribution in terms of the prices for it. 00:03:54.450 --> 00:03:56.980 align:middle line:84% There might be distributions in terms of the productivity 00:03:56.980 --> 00:03:58.720 align:middle line:90% of your production process. 00:03:58.720 --> 00:04:03.330 align:middle line:84% So it can calculate it for two, three, four, more 00:04:03.330 --> 00:04:05.250 align:middle line:90% possible distributions. 00:04:05.250 --> 00:04:09.180 align:middle line:84% Now we tend to keep those numbers simple to focus 00:04:09.180 --> 00:04:12.600 align:middle line:84% on the main aspects of it, but it's perfectly possible 00:04:12.600 --> 00:04:14.550 align:middle line:90% to have a large number. 00:04:14.550 --> 00:04:19.920 align:middle line:84% I have a former student now who is working on a spreadsheet 00:04:19.920 --> 00:04:26.010 align:middle line:84% that, as I understand it, has 20,000 entries on it-- 00:04:26.010 --> 00:04:27.150 align:middle line:90% lines to it. 00:04:27.150 --> 00:04:32.830 align:middle line:84% To me that seems absurd, but the company is billing by the hour, 00:04:32.830 --> 00:04:37.960 align:middle line:84% so maybe it's a good thing to have a lot of work to do. 00:04:37.960 --> 00:04:40.770 align:middle line:84% But in any case, the point is it can 00:04:40.770 --> 00:04:42.730 align:middle line:84% work with almost any distribution, 00:04:42.730 --> 00:04:46.710 align:middle line:84% it's not limited by some mathematical formula 00:04:46.710 --> 00:04:50.430 align:middle line:84% or limited to a particular approach. 00:04:50.430 --> 00:04:54.420 align:middle line:84% And I say that because there are other forms 00:04:54.420 --> 00:04:58.290 align:middle line:84% of dealing with uncertainties where really it 00:04:58.290 --> 00:05:00.960 align:middle line:84% assumes that you have a particular distribution 00:05:00.960 --> 00:05:03.280 align:middle line:84% or particular aspects of distribution. 00:05:03.280 --> 00:05:06.600 align:middle line:84% So for example, those of you who might be in supply chain 00:05:06.600 --> 00:05:08.610 align:middle line:84% or have otherwise dealt with it, you 00:05:08.610 --> 00:05:13.080 align:middle line:84% might be using dynamic programming. 00:05:13.080 --> 00:05:15.870 align:middle line:84% Dynamic program is a very powerful way 00:05:15.870 --> 00:05:18.670 align:middle line:84% of thinking about what might happen in the future, 00:05:18.670 --> 00:05:22.590 align:middle line:84% but it has very strict assumptions about it 00:05:22.590 --> 00:05:25.950 align:middle line:84% which, in fact, don't apply to the kind of problems 00:05:25.950 --> 00:05:28.920 align:middle line:90% that we're dealing with here. 00:05:28.920 --> 00:05:32.130 align:middle line:84% To give it a name, it assumes passive independence. 00:05:32.130 --> 00:05:33.630 align:middle line:84% That is, there is nothing management 00:05:33.630 --> 00:05:37.410 align:middle line:84% can do to change the situation, which is, of course, exactly 00:05:37.410 --> 00:05:38.610 align:middle line:90% where we differ. 00:05:38.610 --> 00:05:41.640 align:middle line:90% 00:05:41.640 --> 00:05:48.440 align:middle line:84% So simulation has not been part of standard engineering 00:05:48.440 --> 00:05:50.240 align:middle line:90% education. 00:05:50.240 --> 00:05:51.440 align:middle line:90% Why is that? 00:05:51.440 --> 00:05:53.450 align:middle line:84% It's because it implies if you're 00:05:53.450 --> 00:05:57.740 align:middle line:84% going to do it in a way that really makes a lot of sense, 00:05:57.740 --> 00:06:00.090 align:middle line:84% you want to have lots of different samples. 00:06:00.090 --> 00:06:02.810 align:middle line:84% If you're looking at different things happening, 00:06:02.810 --> 00:06:05.580 align:middle line:84% you want to have lots of flips of the coin, so to speak. 00:06:05.580 --> 00:06:10.580 align:middle line:84% You might want to have 1,000 repetitions. 00:06:10.580 --> 00:06:13.550 align:middle line:84% Well, modern computers can do it. 00:06:13.550 --> 00:06:15.470 align:middle line:90% 20 years ago they couldn't. 00:06:15.470 --> 00:06:17.450 align:middle line:90% So simulation has not been-- 00:06:17.450 --> 00:06:19.400 align:middle line:84% Monte Carlo simulation has not been 00:06:19.400 --> 00:06:23.060 align:middle line:90% part of engineering practice. 00:06:23.060 --> 00:06:24.530 align:middle line:90% So it's now possible. 00:06:24.530 --> 00:06:26.960 align:middle line:84% Secondly, the model basically can 00:06:26.960 --> 00:06:29.555 align:middle line:84% be very simple, a spreadsheet is very powerful, 00:06:29.555 --> 00:06:34.670 align:middle line:84% s and you can get it with an Excel add-ins that can do it. 00:06:34.670 --> 00:06:37.910 align:middle line:84% The garage case template which I've provided 00:06:37.910 --> 00:06:39.900 align:middle line:90% is an example of that. 00:06:39.900 --> 00:06:42.230 align:middle line:84% They are very slick versions available. 00:06:42.230 --> 00:06:44.840 align:middle line:90% @RISK is one of them. 00:06:44.840 --> 00:06:47.210 align:middle line:84% The case I'll be talking about later on 00:06:47.210 --> 00:06:51.230 align:middle line:84% has another one called Crystal Ball. 00:06:51.230 --> 00:06:54.050 align:middle line:84% So there are various forms of it. 00:06:54.050 --> 00:06:56.630 align:middle line:84% The slick versions often don't allow 00:06:56.630 --> 00:06:59.150 align:middle line:84% you to do the kind of flexibility analysis 00:06:59.150 --> 00:07:03.690 align:middle line:84% that we'll be talking about in the course, 00:07:03.690 --> 00:07:06.350 align:middle line:84% but just to give you-- there's a range of it. 00:07:06.350 --> 00:07:09.710 align:middle line:84% And there are, of course, other simulations that are possible. 00:07:09.710 --> 00:07:13.430 align:middle line:84% Systems dynamics models, queuing models, neither of which 00:07:13.430 --> 00:07:15.740 align:middle line:84% really apply to the kind of issues 00:07:15.740 --> 00:07:18.680 align:middle line:84% that we're dealing with in terms of system design 00:07:18.680 --> 00:07:22.400 align:middle line:84% and management, but I don't want to exclude them 00:07:22.400 --> 00:07:26.170 align:middle line:90% from the listing. 00:07:26.170 --> 00:07:29.540 align:middle line:84% So what do you need for it to do a simulation? 00:07:29.540 --> 00:07:32.680 align:middle line:84% You need distributions for the key parameters, 00:07:32.680 --> 00:07:37.990 align:middle line:84% which, in general, observed, estimated, assumed, or guessed. 00:07:37.990 --> 00:07:40.720 align:middle line:90% Let's think about some examples. 00:07:40.720 --> 00:07:41.980 align:middle line:90% Observe. 00:07:41.980 --> 00:07:44.050 align:middle line:90% For example, the river flow. 00:07:44.050 --> 00:07:48.670 align:middle line:84% If you're trying to build an electric power hydrodam-- 00:07:48.670 --> 00:07:53.470 align:middle line:84% hydropower dam for the Nile as they have done now in Ethiopia, 00:07:53.470 --> 00:07:56.950 align:middle line:84% you'd like to know something about the uncertainties 00:07:56.950 --> 00:07:59.080 align:middle line:84% about how much water is going to come down 00:07:59.080 --> 00:08:02.530 align:middle line:84% and how fast you could fill up the dam 00:08:02.530 --> 00:08:04.730 align:middle line:90% and how does that all work? 00:08:04.730 --> 00:08:07.820 align:middle line:84% So you can have observations of that 00:08:07.820 --> 00:08:11.310 align:middle line:84% for the past 50 to 100 years depending on where you are. 00:08:11.310 --> 00:08:14.540 align:middle line:84% So a river is an observable distribution. 00:08:14.540 --> 00:08:16.910 align:middle line:84% And it may have some peculiarities. 00:08:16.910 --> 00:08:21.260 align:middle line:84% It doesn't have to follow any kind of normal 00:08:21.260 --> 00:08:26.290 align:middle line:84% or Gompertz or other curve, it is what it is. 00:08:26.290 --> 00:08:27.460 align:middle line:90% It can be estimated. 00:08:27.460 --> 00:08:31.390 align:middle line:84% That is, if you're trying as in the mining case which I'll 00:08:31.390 --> 00:08:36.130 align:middle line:84% show later on today, you can consult the experts 00:08:36.130 --> 00:08:38.830 align:middle line:84% who have designed processing plants 00:08:38.830 --> 00:08:41.890 align:middle line:84% and who can have their best estimates of how 00:08:41.890 --> 00:08:45.380 align:middle line:84% this particular one is going to work out. 00:08:45.380 --> 00:08:48.640 align:middle line:84% And so you can have a technical model of it, 00:08:48.640 --> 00:08:53.410 align:middle line:84% and that estimates the performance of the situation. 00:08:53.410 --> 00:08:54.910 align:middle line:90% It can be assumed. 00:08:54.910 --> 00:08:59.980 align:middle line:84% That is, you may have forecast of prices of your product 00:08:59.980 --> 00:09:02.260 align:middle line:84% because you know something about the competition, 00:09:02.260 --> 00:09:06.070 align:middle line:84% somewhat about how they behave and so forth so that you 00:09:06.070 --> 00:09:11.320 align:middle line:84% make forecast ahead assuming various aspects of the market, 00:09:11.320 --> 00:09:16.340 align:middle line:84% the actions of your competitors, the demand for your product, 00:09:16.340 --> 00:09:17.830 align:middle line:90% and so forth. 00:09:17.830 --> 00:09:20.640 align:middle line:90% Or it simply may be guess. 00:09:20.640 --> 00:09:25.500 align:middle line:84% Your production may depend upon the possible environmental 00:09:25.500 --> 00:09:26.370 align:middle line:90% rules. 00:09:26.370 --> 00:09:29.790 align:middle line:84% If you're into renewable energy, the rules 00:09:29.790 --> 00:09:33.330 align:middle line:84% on taxation and environment and so forth 00:09:33.330 --> 00:09:36.960 align:middle line:84% may be affected by national legislative-- or local 00:09:36.960 --> 00:09:41.010 align:middle line:84% legislation, and we don't have observations 00:09:41.010 --> 00:09:43.650 align:middle line:84% of how they behave in those situations in the past, 00:09:43.650 --> 00:09:46.140 align:middle line:84% but we could go to people who understand 00:09:46.140 --> 00:09:48.810 align:middle line:84% the legislature or the process and they 00:09:48.810 --> 00:09:53.140 align:middle line:84% might tell you whether it's a 50% chance or a 10% chance. 00:09:53.140 --> 00:09:57.000 align:middle line:84% And it may be useful to you, but it's clearly guessed. 00:09:57.000 --> 00:09:59.310 align:middle line:90% So we do the best we can. 00:09:59.310 --> 00:10:05.170 align:middle line:84% And in defense of dealing with things which are guesses, 00:10:05.170 --> 00:10:07.690 align:middle line:84% I would say that guessing there's a chance 00:10:07.690 --> 00:10:10.150 align:middle line:84% is better than ignoring a chance, which 00:10:10.150 --> 00:10:14.050 align:middle line:84% is a form of guess also, but with a lot 00:10:14.050 --> 00:10:16.340 align:middle line:90% less information about it. 00:10:16.340 --> 00:10:20.500 align:middle line:84% So there's no claim that these distributions can 00:10:20.500 --> 00:10:23.840 align:middle line:84% be-- are perfect or scientific in that way, 00:10:23.840 --> 00:10:25.420 align:middle line:84% but we have to deal with the world 00:10:25.420 --> 00:10:31.840 align:middle line:84% as it is and recognize that their chances and uncertainties 00:10:31.840 --> 00:10:35.070 align:middle line:90% is in the right direction. 00:10:35.070 --> 00:10:40.000 align:middle line:90%