Lancaster Engineering Catastrophes

Causes and Effects of Major Accidents
3. Auflage 2005
ISBN: 978-1-84569-081-6
Verlag: Elsevier Science & Techn.
Format: EPUB
Kopierschutz: 6 - ePub Watermark

E-Book, Englisch, 288 Seiten

There is much to be gained from the study of catastrophes. Likewise the records of accidents in industry and transport are of great importance, not only by indicating trends in the incidence of loss or casualties, but also as a measure of human behaviour. The third edition of this well received book places emphasis on the human factor, with the first two chapters providing a method of analysing the records of accident and all-cause mortality rates to show their relationship with levels of economic development and growth rates, and to make suggestions as to the way in which such processes may be linked. Case histories are given throughout the book. These are designed to show how human frailty, or the unexpected weakness of materials, or a combination of both, can lead to dire and tragic circumstances. - Understand why disasters occur and how they could have been avoided - Emphasis is placed on the human factor in catastrophes - Case studies illustrate the factors that can lead to dire and tragic circumstances

Autoren/Hrsg.

Lancaster, J. F.

Leseproben

Chapter 1

Analysing casualty records

How big is a catastrophe?

It is natural to associate the word 'catastrophe' with some large-scale event such as the collison of two passenger aircraft, or the destruction by fire of a major off-shore oil platform like Piper Alpha. In the case of fatal accidents, however, it is not so simple. Where does one draw the line? Must there be one hundred deaths, or 50 or 20? There is no good answer to this question. Indeed, the premature accidental death of a single person is a tragedy for family and friends, and may have dire financial consequences. Therefore much of this book will be concerned with fatal accidents regardless of their scale.

A great deal of data concerning accidents has accumulated during the nineteenth and twentieth centuries. Early contributors to this collection in Britain were the Census Office (now the Office for National Statistics), which first produced figures for mortality in 1841, Her Majesty's Inspector of Factories, and the Railway Inspectorate. The introduction of the motor vehicle at the beginning of the twentieth century resulted in a sharp rise in casualities on roads. In Europe and North America government departments provided records of these. In Britain, the Office for National Statistics gives figures for accident mortality, including those that are due to road traffic, from 1901. Lloyd's Register of Shipping publishes an annual statistical summary for losses from the world's commercial fleet,1 the first copy of which was issued in 1891. The Boeing Company has performed a similar service for jet passenger aircraft since 1964.2 Although accidents are responsible for only a small proportion of deaths, they are a matter of universal concern, and if public interest fades from time to time, it is soon reawakened by some newsworthy disaster.

In making an analytical study of such material, the first step is to establish units. Time is the most straightforward; one year is almost universally taken as the unit. Seasonal variations in the number of accidents are not uncommon and the use of annual data eliminates this variable.

The quantitative measure for accidents must surely be the rate, that is, the annual number of deaths divided by the number of people employed in the activity concerned; in short, annual deaths/population. The mortality rate in a particular country conforms to this model, whether mortality is considered as that due to all causes or as that due to accidents. Records may not provide a direct measure of the mortality rate; for example, in the case of shipping, Lloyd's casualty reports give the percentage loss of ships from the world fleet. However, this figure could reasonably be taken as a measure of the proportional loss of ships' crews. Numbers of people killed in train accidents may be related to the number of journeys, this latter figure being a measure of the population of rail passengers.

Whatever the merits or demerits of such measures, it is a fact that during the twentieth century mortality and loss rates fell in the majority of cases and, moreover, the pattern of their decline was similar. It is therefore possible to analyse the historical records of accident rates in general, regardless of the activity to which they relate.

In analysing the accident record it is necessary to make use of various mathematical techniques. Details are given in Appendix 1. Only the results will be given in the main text.

Perspectives

Before proceeding with an examination of the historical data, it will be instructive to look at some figures for accidents as a whole. For the year 2000 in England and Wales the total number of accidental deaths (excluding suicide and homicide) was 11149, this being 2.08% of the number of deaths from all causes. The document from which these figures are taken3 allocates deaths in one of four categories, three of which are causes, whilst one (transport) is an activity. The allocation is shown as a bar chart in Fig. 1.1, from which it will be seen that falling (from a ladder, for example) is the commonest cause of accidental death. This category covers all activities other than transport, including industry, sport and domestic life, as does the category 'fire'. Fatalities from both these causes decreased throughout the twentieth century.

1.1 Causes of accidental death in England and Wales for the year 2000.3 Inset figures are percentages of the total.

Transport accounted for nearly one-third of accidental deaths at the beginning of the twenty-first century. Of these, 92% are the result of road traffic accidents. Rail and other forms of transport cause relatively few casualties. Since 1900 the increase in range and speed of both air and road transport has been dramatic, and in both cases it has been necessary not only to learn how to operate the individual vehicle, but also to develop systems whereby the risk of collision is minimised. The success of this learning process is evidenced by the reduction in loss and fatality rates as discussed below.

The trend curve for casualty rates: the normal case

Figure 1.2 plots the annual percentage loss in numbers of ships from the world fleet for the period 1891 to 1999. For clarity, data points are shown at ten-year intervals. The solid line in Fig. 1.2 is the trend curve for the set of shipping loss data points. It may be determined using the following procedure:

1 Take the natural logarithms (logarithms to base e) of the loss rates.

2 Plot the resulting figures against the corresponding year dates.

3 Determine the equation of the line that gives the best fit to these data pairs. This equation will take the form

r=bt+constant

  [1.1]

4 Take the antilogarithm of this equation to obtain the equation of the trend curve

=aebt

  [1.2]

1.2 The percentage of the total number of ships in the world’s commercial fleet that was lost annually between the years 1890 and 2000.1

where a is another constant, t represents time and b is the slope of the ln r versus time line. In the case of the shipping losses portrayed in Fig. 1.2, a = 7.99 and b = - 0.026, time being measured in anno domini (1891, 1910 etc.). Details of the procedure are given in Appendix 1 but a scientific calculator can be used to obtain a and b directly from the raw data.

The form of this curve is exponential. It is a basic characteristic of an exponential curve that the proportional gradient is constant. In the present case

/rdr/dt=b

  [1.3]

Thus, b is the proportional gradient of the loss rate trend curve. The dimension of b is 1/(time) and since in this book the unit of time is 1 year, this dimension is 1/(year). The value of b is therefore independent of the units (annual percentage loss, annual fatality rate per million population, etc.). 'Proportional gradient' is an awkward designation and its meaning is not immediately obvious. In later sections of this book the term 'decrement' is used as an alternative. Decrement is the opposite of increment and means the fact or process of decreasing. Here it will be quantified to mean the annual proportional or percentage reduction in the casualty rate. Thus for shipping losses the proportional gradient of - 0.026 may be represented as an annual decrement of 2.6%. This is an approximation in two respects. First, it implies an annual stepwise change, whereas for the trend curve, and the reality which it represents, the change is continuous. Secondly there is a slight numerical inaccuracy; a proportional gradient of - 0.026 causes an annual decrement of 2.566491%. The decrement represents therefore a stepwise model that is accurate enough for yearly time intervals and which is easier to understand than the continuous model. A regression analysis using the exponential model is, however, the best way to organise accident and economic growth rate data and obtain values for the decrement, as is shown in Appendix 1.

Collective skills

It is easy to envisage a growth process because this is a matter of common everyday experience. Trees grow, cities spread, traffic increases and national income rises. There is no such obvious model for the progressive diminution of accident rates that was observed here. However, a possible way out of this difficulty is to suppose that as time goes on, the people concerned become progressively more adept in the avoidance of accidents. In other words, the fall in accident rates is a reflection of an increase in skill on the part of the operators in the industry or mode of transport concerned. It seems reasonable in this connection to speak of collective skill since large numbers of individuals are involved and in many instances they must operate a system, for example drivers of vehicles must observe a set of traffic rules. Thus, we may consider that for any particular activity the collective skill increases exponentially with time. The higher the level of collective skill, the lower the loss rate. And it would be expected that the average annual proportional change in skill would be numerically similar to the...