1 Scope and objectives

Transcription

1 1 Scope and objectives 1.1 Introduction This chapter of the Report defines the scope. This includes what kinds of risk are covered, who the intended readers are and what, in broad terms, the Report seeks to achieve. 1.2 Types of risk This Report covers all risks which fall to structural engineers to manage. Primarily, this means risks to the health or safety of workers and the public, and serious environmental damage, where there are both legal and moral obligations to keep risks low. Significant cost and programme risk arising from structural engineering decisions is also in the scope. Business risk in engineering projects is not unique to structural engineering and is covered elsewhere 1.1, 1.2, so is not included. Communication and co-ordination is an essential part of risk management. The structural engineer is in a good position to take a lead role on suitable projects and should develop the competences required to do this. The scope is not limited to design of new works; risks will occur throughout the life cycle of a building or structure; from concept through detailed design, construction, commissioning, operation, maintenance, dismantling, demolition and ultimately to disposal. 1.3 Intended readers The objective of this Report is to help structural engineers in all parts of the industry to identify risks and deal with them. Structural engineers includes not only designers and consultants but also engineers engaged in site work, existing structures and demolition. An understanding of the whole picture is important for engineers in all sectors and will assist exchange of useful information to reduce risk. The Report is relevant to all structural engineers, from students and graduates learning how to manage risks to experienced engineers and technicians involved with any scale of project from domestic improvements to unusual or high risk projects. The Report is mainly based on UK experience but the principles of good practice are applicable worldwide. Local laws, cultures and economic circumstances may require a different approach in detail. Although the recommendations deal with structural engineering, examples of risk issues from related industries are used to illustrate common themes and to show the potential to apply learning from any source. This Report has been compiled by a group of practising engineers and defines the Institution of Structural Engineers view of good practice in risk management. As legal requirements vary between countries the Report does not, in general, define how to comply with the law. The exceptions to this are the chapter on legal requirements and where reference to laws is used to illustrate a principle. So far as the authors are aware, none of the guidance conflicts with any legal requirements. The Report is divided into chapters but important concepts are illustrated throughout. Anyone concerned with risk in structural engineering, at any stage in the life of a structure, is encouraged to read the whole Report. 1.4 Relevance to clients While structural engineers, as designers or constructors, should have the experience and competence to understand and manage risks, clients also have a significant role. They are responsible for providing sufficient information, time and resources to ensure that risks can be properly managed, and they often have essential knowledge on the use and lifetime management of the structure. The engineer should ensure that the client understands these responsibilities. Clients should take care as to whom they allocate risk, as some risks are best managed by the client. Engineers are encouraged to point out to clients the value of sound risk management, underpinned by a culture of safe design, and to support them in achieving this. 1.5 Summary of objectives The Report promotes a proportionate and open approach to risk assessment and management as a process for preventing harm, not for producing documents. It aims to put more meaning behind the words risk in structural engineering by describing the types of risks that may affect structural engineering work and by providing a decision making framework for risk, taking account of legal principles, available guidance and research. As set out by the UK s Engineering Council 1.3, good risk management requires engineers to make judgements, avoid risk aversion and hence facilitate innovation. The Report describes the legal background, in the UK and elsewhere, and provides examples of risks that were managed badly and that resulted in serious accidents. Tools, techniques and selected references to assist engineers are described, although once the principles are understood, managing risk is mostly a matter of attitude, culture and common sense. The Report cannot say everything about risk in structural engineering, but it allows the reader to make further studies. The Institution of Structural Engineers Risk in structural engineering 1

2 3 Principles of risk management 3.1 Introduction This chapter discusses structural engineering risk in general terms; how and why it may arise and some of the principles that influence risk management. Specific guidance on how to apply risk management tools is given in Chapter The importance of risk management In most countries, the accident rate in construction is well above the average for industry in general. Even without accidents, many common construction practices can affect workers health. Unforeseen events can cause projects to run seriously late or over budget. Frequently, the lack of foresight occurs during the structural design phase or in the detailed planning for construction. Every risk is a learning opportunity. Only a small percentage of structures fail in harmful ways but minor mishaps are more common. The evidence from individual tragedies and near misses is crucial to education in assessing risks and producing safer designs, particularly for low probability, high consequence events. All parties have an obligation to assess and minimise risks that might cause personal harm and, in some jurisdictions, specific roles and records are required. Many engineers will also perceive a profound self interest in minimising risks that might damage their professional lives. There is also good economic justification for taking the trouble to manage risks properly. Failures and accidents can be very expensive. The cost of structural failure invariably exceeds the costs of preventing the incident by a significant margin. In addition to the human cost, a single fatality can result in millions of pounds in direct and indirect costs, such as stopping work, internal inquiries and contract penalties, with the possibility of prosecution. The true cost of accidents and ill-health resulting from work is often underestimated. Research in the UK has reported that various employers estimated their costs at to 2 million per year 3.1. Large organisations tend to have a better understanding, due to the number of accidents; in a small company, accidents are infrequent and the costs are not appreciated. The cost to employers of workplace injuries, work-related ill-health and accidental damage events in the UK has been estimated to be between 3.5 billion and 7.3 billion a year 3.2. The economic cost and disruption to society from infrastructure failures such as loss of a bridge is almost incalculable. 3.3 How big is the risk? The combination of the type of harm, its severity and its probability defines the risk. To decide where effort at risk minimisation may be best applied, it is necessary to understand how these factors affect the size of each risk, so that they can be given consideration in proportion to their seriousness. Neither structural engineering nor risk management is an exact science. While the probability of a particular load, such as wind, can be estimated, it is generally difficult to quantify structural risk in numerical terms. Although many techniques exist to attempt this, some of which are discussed in Chapter 6, their main use is to make comparisons between risks. Assessment of residual risk for different options can help to select the best approach. Most risks can be minimised by following good practice and using engineering judgement which, in turn, requires the kind of background information this Report provides. Even within the profession, there will be different attitudes to risk; engineers are required to produce safe structures against a background of uncertainty in loading, using materials that have variable properties, all supported off uncertain ground conditions. To understand a risk it is necessary, among other things, to have a realistic or bounding estimate of the probability of the initiating event. Perceptions of risk vary, and are not always an accurate pointer to areas for risk reduction. Some risks may have high probability but limited consequences, while others may have low probability with serious consequences. If there are risks with high probability and high consequence they are unlikely to be tolerable; the hazard should be eliminated or the project would not be viable. Engineers often work with unknowns, and have developed methods to manage this. While the actual strength of a given section of concrete is unknown, as is the highest wind speed next winter, statistical methods and factors of safety enable engineers to have appropriate confidence in their designs. Wind speed and concrete strength may be defined as known unknowns, because the events they relate to have been identified. There are also unknown unknowns, where even the possibility has not occurred to the engineers involved. If the designers of the World Trade Center had anticipated that terrorists might deliberately crash a large fully fuelled aircraft into their structure, they might have been able to estimate the range of possible impact parameters, or known unknowns. As it was, the event was outside anyone s imagination at the time of design 3.3. Completely unknown events are sometimes known as black swan events 3.4. Until the first black swans were discovered in Australia, ornithologists never considered the possibility that swans could be anything other than white. A further example is the failure of the Nishinomiya Harbour Bridge during the Kobe Earthquake (1995) 3.5. The made ground slumped towards the harbour, taking the bridge foundation with it and causing the side span to lose its bearing (Figure 3.1); it appears that this failure mode was not anticipated by the designer. Robustness and sensitivity studies are the best approach to managing such unknown risks. 6 The Institution of Structural Engineers Risk in structural engineering

3 Principles of risk management 3.4 Figure 3.1 Nishinomiya Harbour Bridge 3.4 Competence Competence is crucial to risk management; decisions should only be made by people who have appropriate skills, knowledge and experience. Sometimes, this requires involvement of more than one person. A group of experienced people, using a brainstorming approach, will often identify risks and ways to reduce them which would not occur to an individual working alone. Ensuring adequate competence and resources is a fundamental part of procurement for any contract, be it for design, construction, maintenance or demolition. Passing risk along the supply chain may appear to save money, but it will not do so if the risk becomes the responsibility of somebody who is unable to manage it. 3.5 Communication Management of any risk requires good communication, co-ordination and co-operation. Many environmental and human disasters are caused not by deliberate omission but by oversight and lack of communication. Many accidents occur because of changes made by somebody who didn t understand why it was done the way it was. In many countries there is now a legal requirement for co-ordination of both design and construction work. In the European Union, this stems from the Temporary or Mobile Construction Sites Directive 3.6, implemented in the UK as the Construction (Design and Management) Regulations 3.7. This has resulted in an unfortunate trend for co-ordination of health and safety to be seen as a separate responsibility from co-ordination of the design itself. Structural engineers should, in order to do their job, be competent in relevant health and safety matters, which in many cases means they should be the best people to co-ordinate the design to minimise risk. Communication is not only necessary from engineer to engineer, but between engineers and the public. It is very easy to find that public opinion is opposed to a development because the risks (both from building it and from not building it) have not been fully explained. There was controversy in 2012 when a number of Italian earthquake scientists were convicted for apparently giving inappropriate reassurance that a severe earthquake in L Aquila 3.8 was not imminent. The error was said to be not in their scientific predictions, but in the way these were expressed to the public. 3.6 What is an acceptable level of risk? The acceptability of the risk partly depends on the type of harm. In the case of harm to people, then the tolerability of that risk should be lower. If the consequence is only commercial, such as delay or lack of performance, the client and design team are entitled to consider how much they are prepared to spend to control the risk. In some cases there are legal or regulatory requirements for particular types of risk, but the difficulty of quantifying risk results in many laws being targeted at processes and behaviour rather than directly at risk. Consequence is an important part of risk. Structural failure in one situation may have relatively low consequences; in another situation a similar amount of structural damage could result in much greater harm to people, financial cost or disruption to society. As an example, consider a bridge leading to a farm; the potential consequences of collapse into a river, onto a main road or onto a high speed railway could be very different. The Institution of Structural Engineers Risk in structural engineering 7

4 6 Statistical and probabilistic methods 6.1 Introduction This chapter provides an overview of the numerical background to risk management. It is mainly relevant to structural stability rather than hazards on construction sites. It is written for practising structural engineers, not risk specialists, and is not intended to be academically rigorous. A bibliography is provided for further study. experienced individuals. In advanced structural engineering, reliability can be assigned a numerical value 6.2. BS EN , the head Eurocode, offers a methodology for specialist calculation of an appropriate load factor taking account of views on the accuracy of analysis, confidence on material properties and so on. 6.3 Quantifying probability 6.2 Background Day to day structural engineering does not generally require direct application of statistical methods. Indirectly, however, statistics have a significant influence, being used extensively in assessing and describing material strengths and in describing the likelihood of random loadings such as wind. For designs that meet the requirements of a code of practice, the safety factors reduce the probability of failure to a low enough level to take account of the uncertainties in design, loading and material quality. An understanding of statistics and probabilities can be useful in decision making and in assessing the capability of existing buildings where code shortfalls have been identified. More widely, probabilistic approaches can provide useful insights for comparing risks, providing perspective to help decide whether a risk is tolerable, to decide which risk requires most attention or to support a decision that the risk has been reduced to an acceptable level. Since the magnitude of many loads on a structure is fundamentally uncertain, they can only rationally be defined in terms of probabilities. An example is the 1 in 50 year wind, defined as the wind speed with a 1 in 50 probability of being exceeded at least once during a year. This gives a 64% chance that it will be exceeded at least once in 50 years. Low probability events do happen and this wind speed could be exceeded more than once in the period or even in the first year. For structures where very high reliability is sought, a 1 in year event might be used 6.1. There is a debate about the use of probability as the sole way to deal with such low frequency natural events, compared to the resilience approach described in Section The use of numerical assessment can be deceptive, suggesting a degree of precision that rarely exists. A probability might be calculated as 1 in per year (often written as 10 4 per year), but due to the many assumptions which have to be made in the assessment it might in practice lie between 10 2 and Hence these methods are somewhat approximate and should not be allowed to dominate any risk assessment; the techniques are best used in conjunction with deterministic methods. Moreover, because this field is so specialised, the preparation of probabilistic numerical arguments is best left to There are various ways of quantifying probability. In everyday language the chances of winning a lottery might be considered as 1 in 14 million per ticket purchased, or the probability of dying from smoking as 1 in 200 per year. In engineering practice, rational methods of defining safety margin or probability of failure can be explored using numerical methods and, in some branches of engineering, such methods are used quite widely. If enough similar items are in use, the proportion that fail can be counted and used to calculate the probability of others failing in the future. Thus, for a complex electrical system, with known failure probabilities for the separate items within that system, it is mathematically possible to define the overall level of reliability or, looked at the other way around, the overall probability of failure. This can be compared with what might be considered an acceptable risk. This would also need an understanding of the way components interact, such as the possibility that one failure could trigger another, or that one event could cause multiple failures. To apply this process to structures, there would need to be some basis for counting actual failures. This is not really available, certainly for the whole structure. The reliability of electronic components can be calculated easily, as thousands are made to the same design. Few structures are identical and as explained elsewhere in this Report, many of the causes of failure are independent of the design code used or the precise details of the structure. One area where failure statistics are used to good effect is in preparedness for emergency response. In most parts of the world, buildings follow a common form, e.g. in the UK, the brick built three bedroom semi-detached house. In an area of high seismicity, there will be a historical understanding of the response of the local type of houses or bridges to earthquake loading of certain intensities and it is possible to make an estimate as to how many might be damaged in a particular earthquake and hence decide what preparedness could be undertaken. This approach becomes more effective as more data is available, such as through improved seismology and satellite based damage surveys. Calculated probabilities will only be correct if the data used to generate them is correct. For example, it may be assumed that concrete meets the specified characteristic strength; if, as often happens, it is over strength, the probability of failure may be lower (provided the amount of reinforcement is adequate). It is usual to calculate the probabilities using what are 26 The Institution of Structural Engineers Risk in structural engineering

5 Statistical and probabilistic methods 6.4 known as conservative values, that is, the value is chosen cautiously, so that the final probability of failure is probably an overestimate. It is often argued that conservative values should be used where there is uncertainty, but this approach requires care. Risk management often involves compromises, comparing one option against another to see which has the lowest overall risk 6.4, but inconsistent levels of conservatism could overestimate one side of the balance, and so skew the judgement. 6.4 Safety factors To avoid the need for engineers to apply probability theory in routine design, structural design generally includes a safety factor or load factor (usually built up from partial factors) which ensures that the probability of overall failure is low enough. The value of the factor was traditionally subjective, based on collective historical experience, but there is now an aspiration to derive the factors statistically or by reliability theory. A structure with design factors lower than in the design code is not necessarily unsafe, but it should be expected to have a higher probability of failure than one that meets the code. Excluding gross error, structural failure would not normally occur unless the combined probability of adverse variation in applicable loading, loading configuration, material quality, workmanship, etc., all become coincidentally too high. Structures meeting the code requirements, therefore, have an acceptably low probability of failure. If very high reliability is sought, this can be effected by increasing the load factor, using reliability theory to obtain a specific increase. Conversely, if an increased probability of failure is acceptable, perhaps because a structure need only have limited functionality after the event, then the required load factor may be reduced. For example, a building under construction would be unoccupied and construction work would usually stop in very strong winds, so the design wind loading during construction may be based on a two year recurrence period rather than 50 years. Time at risk may be considered as a reason to reduce the factor of safety. This may be valid for transient risks or those that could occur during only a small fraction of the life of the structure and where it might be disproportionate to use the usual factor. When designing for a specific situation that only lasts a short time, such as the construction phase example in the previous paragraph, this may not be valid, and each case should be taken on its merits. For the people involved, construction is usually a full time activity, on one site after another, so increasing their risk would be unjustified. Confidence in material properties clearly affects the selected value of a load factor. Material testing is bound to show a scatter of results and this is managed by using such concepts as the 95% confidence level, which typically defines the characteristic strength. In some codes, the partial factors are varied explicitly to take account of known factors affecting the probabilities. For example, in UK masonry design codes (both BS and BS EN with UK National Annex) the partial factor for materials depends upon the workmanship and quality control of blocks and mortar used. This approach recognises the overall lower probability of failure if workmanship is controlled more closely. Partial factors are also varied for different load combinations (e.g. dead þ live þ wind) to reflect the lower overall probability of that combination of circumstances arising. It is important to be aware that some uncertainties, such as the accuracy of our models of structural behaviour, do not have a specific partial safety factor but are included within other factors. This means that even if the load is known exactly, a partial factor for load of 1.0 may be inappropriate, as the factor also covers other, unstated aspects which are still uncertain. 6.5 Low probability events For events with low probability but high consequences, the reliability provided by standard design codes may not be adequate. This becomes more relevant as the population of the world grows and the number of people living in areas vulnerable to infrequent but severe natural hazards such as floods and earthquakes increases. Coupled with the wider development of an engineered infrastructure, there is a growing belief among the public that the human consequences of such disasters are avoidable. Nevertheless, communities still have to cope with the aftermath of severe wind, severe temperatures and severe flooding even in countries with a well-established infrastructure such as the UK and USA. A practical question engineers have to address is to determine the likelihood of such events and then decide what resources can be afforded to defend against them. Techniques such as quantitative risk assessment (QRA) are available to predict the (numerical) magnitudes of rare natural events and to predict the likelihood of process plant failures. An early use of QRA 6.7 was in the study carried out in 1978 to assess the risk that the chemical plant on Canvey Island posed to London 6.8. Major accidents to such plants do happen, such as the failure at Flixborough 6.9 (1974) or at Buncefield 6.10 (2005). How can engineers decide which modes of failure are possible, what the probabilities of those failures are and hence the risk they pose in terms of their location relative to populated areas? These risks can be expressed in numerical terms. In areas where the population are at risk from flood, an event with a statistical probability of occurrence, one way of assessing the likelihood of damage, or the required height of a flood protection system, is to assess the risks using statistics and numerical values against a target acceptance value. In London, a quantitative flood risk assessment is required to review the effectiveness of the Thames Barrier The consequences of the floods in New Orleans following Hurricane Katrina 6.12, 6.13 (2005) illustrate the need. In the nuclear industry, when assessing the safety of facilities, a combination of deterministic (designing for specified events) and probabilistic assessment (the probability of those events not occurring) is used. Neither method is used exclusively, since both give insights into the overall safety of the plant. Probabilistic methods are often used to define the deterministic events. The accepted safe target in the UK is that the probability of significant harm to the public from radioactivity should have a probability of 10 7 per year or lower. In practice, events as rare as The Institution of Structural Engineers Risk in structural engineering 27