《SAE-TPS-812000-01-5594.pdf》由会员分享,可在线阅读,更多相关《SAE-TPS-812000-01-5594.pdf(16页珍藏版)》请在三一文库上搜索。
1、For permission to copy or republish, contact the American Institute of Aeronautics and Astronautics or SAE International. SAE International 400 Commonwealth Drive Warrendale, PA 15096-0001 U.S.A. American Institute of Aeronautics and Astronautics 370 LEnfant Promenade, S.W. Washington, D.C. 20024 20
2、00-01-5594 The Three Facets of Risk Ted W. Yellman Boeing Commercial Airplane Group 2000 World Aviation Conference October 10-12, 2000 San Diego, CA Published by the American Institute of Aeronautics and Astronautics (AIAA) at 1801 Alexander Bell Drive, Suite 500, Reston, VA 22091 U.S.A., and the So
3、ciety of Automotive Engineers (SAE) at 400 Commonwealth Drive, Warrendale, PA 15096 U.S.A. Produced in the U.S.A. Non-U.S. purchasers are responsible for payment of any taxes required by their governments. Reproduction of copies beyond that permitted by Sections 107 and 108 of the U.S. Copyright Law
4、 without the permission of the copyright owner is unlawful. The appearance of the ISSN code at the bottom of this page indicates SAEs and AIAAs consent that copies of the paper may be made for personal or internal use of specific clients, on condition that the copier pay the per-copy fee through the
5、 Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923. This consent does not extend to other kinds of copying such as copying for general distribution, advertising or promotional purposes, creating new collective works, or for resale. Permission requests for these kinds of copying
6、 should be addressed to AIAA Aeroplus Access, 4th Floor, 85 John Street, New York, NY 10038 or to the SAE Publications Group, 400 Commonwealth Drive, Warrendale, PA 15096. Users should reference the title of this conference when reporting copying to the Copyright Clearance Center. ISSN #0148-7191 Co
7、pyright 2000 by The Boeing Company. Published by American Institute of Aeronautics and Astronau- tics, Inc. and SAE International with permission. All AIAA papers are abstracted and indexed in International Aerospace Abstracts and Aerospace Database. All SAE papers, standards and selected books are
8、abstracted and indexed in the Global Mobility Data- base. Copies of this paper may be purchased from: AIAAs document delivery service Aeroplus Dispatch 1722 Gilbreth Road Burlingame, California 94010-1305 Phone: (800) 662-2376 or (415) 259-6011 Fax: (415) 259-6047 or from: SAExpress Global Document
9、Service c/o SAE Customer Sales and Satisfaction 400 Commonwealth Drive Warrendale, PA 15096 Phone: (724) 776-4970 Fax: (724) 776-0790 SAE routinely stocks printed papers for a period of three years following date of publication. Quantity reprint rates are available. No part of this publication may b
10、e reproduced in any form, in an electronic retrieval system or otherwise, without the prior written permission of the publishers. Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE or AIAA. The author is solely responsible for the content of the
11、 paper. A process is available by which discussions will be printed with the paper if it is published in SAE Transactions. 2000-01-5594 The Three Facets of Risk Ted W. Yellman Boeing Commercial Airplane Group Copyright 2000 by The Boeing Company. Published by SAE International, and the American Inst
12、itute of Aeronautics and Astronautics, Inc. with permission. ABSTRACT Awareness of risk is probably greater today than it has ever been, and the assessment and management of riskwhether or not we think of it as riskhas become common in the lives of many of us. The first part of this paper notes that
13、 risk is never about predictable losses, or about either predictable or unpredictable gains; nor is risk a probability. Rather, risk is either a condition of, or a measure of, exposure to misfortunemore concretely, exposure to unpredictable losses. However, as a measure, risk is not one-dimensionali
14、t has three distinct aspects or “facets“ related to the anticipated values of unpredictable losses. The three facets are Expected Loss, Variability of Loss Values, and Uncertainty about the Accuracy of Mental Models intended to predict losses. This paper illustrates these three facets of risk by the
15、 use of three simple card-game “thought experiments“. An assessment of risk which does not consider all three of these facets is at least incomplete, and can be misleading. The second part of the paper presents an example of how the three-facets- of-risk concept applies specifically to a safety asse
16、ssment of a mission of an engineered system. The author believes that each risk of interest should be characterized explicitly by a well-defined Risk Eventan event defined by an event assertion which clearly specifies the causes and/or loss values intended to be included within the scope of a partic
17、ular risk. It is pointed out that the widely-used formula “Risk = Probability X Severity“ does not even recognize the second and third facets of risk. Nor does the formula make clear how to address the many practical situations in which a variety of possible causes (Hazardous Events, Outcome Events,
18、 and Tragic Events) of losses, and/or a variety of possible resulting Loss Values, are of interest. Furthermore, “Severity“ in the formula is often interpreted to mean the greatest possible Loss Value which can be caused by a Hazardous Eventand that interpretation can yield conclusions which are eit
19、her significantly optimistic or pessimist. INTRODUCTION Risk has always been with us. Whenever people engage in any action or enterprise from which they hope to gain something, risk is almost always one of the prices they pay. How they manage risk may determine not only their welfare, but their very
20、 survival, and this reality is probably more keenly recognized these days than ever before. Terms like risk, risk assessment, and risk management have become pervasive in the governmental, industrial, financial, public health, and scientific-and-engineering communities. In this paper, I will explain
21、 what I think risk meansor should meanparticularly when it is safety-related risk and it applies to missions of engineered systems. In the first part of the paper, I will state what risk in general is, and isnt, all about. I will argue that risk as a measure of possible future losses is not just a s
22、ingle measure, but rather that it is a multi-dimensional measure composed of three facetsall of which must be considered to completely characterize risk. In the second part of the paper, I will show how these three facets of risk apply to a heuristic mission safety-risk assessment of an engineered s
23、ystem. The second part of the paper will also reach well beyond the concept of the three facets of risk. I will try to show that a good risk assessment requires that each risk of interest be characterized by a well-defined set of causal events and various logical relationships between themas represe
24、nted by a single, well-defined, often composite, and sometimes quite complex, Risk Event. Finally, I will try to show why I think that some current and widely-used system-safety risk-assessment definitions and protocols miss the mark. THE HALLMARKS OF RISK “Crossing the street at this busy intersect
25、ion is a risk.“ “What is the risk of selling 100 shares of Microsoft short?“ “When I sold him the policy, I thought he was a good risk.“ “What is the risk that the flight to Detroit will crash?“ “The blood-clotting risk factor was low.“ “He risked his fortune on one last throw of the dice.“ The word
26、 “risk“, can be used as a noun, a verb, or an adjective, and it is tossed out casually. What does “risk“ really mean? Virtually every human action or enterprise will result in both gains and losses. How well we can predict those gains and losses will vary from situation to situation. Risk is only ab
27、out the losses from such enterprises, never about the gains. You will never talk about the “risk“ of inheriting money 1because, while how much you will inherit may be unpredictable, the event of interest (inheriting money) will not cause a loss. Furthermore, risk is only about those losses which are
28、 to a significant degree unpredictable. If you have been getting lousy fuel mileage on an old automobile, low fuel mileage on your next cross-country trip in that vehicle is not a riskit is a loss, all right, but it is highly predictable so it is not a risk. On the other hand, you are more likely to
29、 consider low fuel mileage on a sports-utility vehicle you just purchased from a used-car dealer to be a risk because it both results in a loss and lacks predictability. There is, as far as I have been able to determine, no word comparable to “risk” to specifically denote unpredictable gains. “Retur
30、ns“, if they come from a Treasury bond, for example, wouldnt usually be considered unpredictable. The lack of a word to describe unpredictable gains seems to indicate that people are usually much more concerned about unpredictable losses than about unpredictable gains, and others who have studied ri
31、sk 7 agree that is the case. Risk in general can be defined as either a condition of, or a measure of the amount of, exposure to misfortune that is, exposure to unpredictable losses. That definition seems to be broad enough to work irrespective of the particular kind of risk being described. In the
32、remainder of this paper, I will be using both the measure and the condition sense of the definition. Which sense is intended can be inferred from the context. However, it would be a mistake to assume that the “measure” sense of risk is only one-dimensional. Risk, as I will show, is better viewed as
33、a combination of three unpredictable- loss-related measures: Expected Loss, Variability of Losses, and Uncertainty about the Accuracy of Mental Models intended to predict losses. I will illustrate these three facets of Risk using a thought experiment consisting of three imaginary card games. GAME 1
34、Suppose that you have known me for years, and that I am a solid, honest and completely trustworthy person. (Remember, this is only an example.) The one weakness I have is a taste for gambling. One day, I propose to you that we play the following card game, which I will subsequently call “Game 1“. Yo
35、u will supply a standard deck of 52 cards (no joker), and inspect and shuffle the deck before each game. The game will consist of you simply drawing a single card from the deck. If you draw the ace of hearts, you will have to pay me $52.00. There should be one ace of hearts in the deck, randomly loc
36、ated. So the probability that you will pick it, and thus lose $52.00, should be (1/52). If we were to play this game over and over forever, what is the average per-game amount you should anticipate losing? In general, that amount can be calculated as the sum of the products of each possible Loss Val
37、ue times the Probability of that particular Loss Value occurring. For Game 1, that calculation would look like this: Per-game Long-Term Average Loss Value = ($0.00)(51/52) + ($52.00)(1/52) = $1.00 Mathematicians call such an anticipated long-term average value an “Expected Value“ or an “Expectation“
38、. An Expected Value can be defined with its positive sense being either a gain or a loss, and, as I have previously observed, risk is only about losses. An Expected Loss is an anticipated per-experiment long-term-average loss, determined from a mental-model experiment intended to represent some real
39、-life experiment. The distinction between real-life experiments and mental-model experiments is not crucial for this card game, but it becomes more important, as we shall see, when mental models are not such perfect predictors of real-life phenomena. An Experiment is either a real or imagined sequen
40、ce of physical phenomena which results from some particular well-defined stimuli, and it occurs within some well-defined spatial and temporal envelope (spatial boundaries, and start and finish times, which are either fixed or rule-based variable). A Mental Model is an imagined version of real life.
41、Real Life refers to the actual physical phenomena which occur around, and sometimes to, usas opposed to our representations of such phenomena in our mental models. Now you would be kind of goofy to play a game from which you anticipate sustaining net losses averaging $1.00 per game. So to induce you
42、 to play, I will pay you $1.25 for every game we playoffsetting your expected loss of $1.00 per game, and in addition giving you a handsome average profit of 25 cents per game. This offsetting arrangement, being a pure gain, has nothing to do with risk. I introduce it only to keep the proposed game
43、from being even more irrational than it already is. In real life, risks must also have offsetting gains predictable and/or unpredictable in varying degreesif they are to be accepted rationally. Figure 1 shows the probability mass function for Game 1. It shows a probability mass function because the
44、values of your possible losses from Game 1 are discretezero and 52 dollars. If continuous loss values were being modeledthat is, an infinite number of possible-loss valuesI would have used the more familiar probability density function, an example of which is the familiar “bell curve“. The Figure 1
45、probability mass function shows that there are two possible loss values0 and 52 dollarsand that the probabilities of those losses are (51/52) and (1/52) respectively. Also shown on the figure is your per-game Expected Loss of $1.00. I will defer further comment on this game until after I describe a
46、second game. GAME 2 Again, you and I will play this game, and you will supply the deck and inspect and shuffle it before each game. However, instead of drawing one card, in this game you will draw four cards in a row without replacing any of them in the deck. Furthermore, in this game, you will lose
47、 (only) if you draw an ace of hearts, an ace of spades, an ace of diamonds, and an ace of clubsand draw them in that order. And there is one final difference. If you lose you must pay me not $52.00, but $6,497,400.00. Otherwise you will pay me nothing. So what is the per-game probability you will be
48、 hit with a $6,497,400.00 loss if you play Game 2? That probability calculation looks like this: P = (1/52)(1/51)(1/50)(1/49) = (1/6,497,400) The average per-game amount you can anticipate losing is, again, the sum of the products of each possible-loss value times its probability: Per-game Long-Term
49、 Average Loss Value = ($0.00)(6,497,399/6,497,400) + ($6,497,400)(1/6,497,400) = $1.00 Well, heres a surprise! The Expected Loss for Game 2 is exactly the same as the Expected Loss for Game 1 that is, $1.00! (Actually, this wasnt a complete coincidence.) So again, I will pay you $1.25 at the beginning of each gameto offset your Expected Loss of $1.00 per game and also give you an average profit of 25 cents per game. Figure 2 shows the probability mass function for Game 2. Your two possible loss values are $0.00 and $6,497,400.00, and their probabilities ar
链接地址:https://www.31doc.com/p-3793946.html