FÖRORD 
Samhället kräver enligt Plan- och bygglagen (PBL) att byggnader ska ha ”betryggande stadga, bärförmåga och beständighet”. Utifrån dessa krav har Boverket utarbetat föreskrifter och råd för byggnader. Olika konstruktionstyper behandlas, exempelvis betong- och stålkonstruktioner samt geokonstruktioner, som avser byggnadsdelar i jord. Konstruktionsreglerna utgår från användning av sannolikhetsbaserade dimensioneringsmetoder. För undermarksanläggningar finns ännu ingen sådan tolkning till konstruktionsregler utan byggherren har att svara för att PBL följs. Vägverket ställer i egenskap av byggherre kravet att Boverkets regler i princip ska tillämpas även för bergtunnlar. EU har också tagit fram ett förslag till norm via Eurocode, som kommer att gälla även för bergkonstruktioner. 

Det råder i konstruktionshänseende väsentliga skillnader mellan konstruktioner där materialet kan föreskrivas och sådana där materialet är givet men delvis okänt. För den sistnämnda typen är tillämpningen av sannolikhetsbaserade metoder betydligt svårare. Forskning om hur metoderna skall kunna användas för konstruktioner i jord har pågått sedan slutet av 70-talet. För tillämpning i berg har bara enstaka arbeten utförts. Dimensionering baseras därför i hög grad på erfarenhetsbaserade, deterministiska metoder, bland annat med hjälp av klassificeringssystem i kombination med numeriska modellberäkningar. 

För att skapa en grund för vidareutveckling av sannolikhetsbaserad dimensioneringsteknik har en metodikstudie genomförts inom ramen för SveBeFos forskningsprogram FP 2000. Rapporten redovisar hur dimensioneringsprocessen i sin helhet kan bedrivas och hur man i olika stadier av processen måste fatta beslut under osäkerhet under det att man successivt vinner allt bättre information, därav uttrycket ”informationsbaserad design”. Metodiken har i delar tillämpats vid några större bergprojekt under senare år, vilket kort kommenteras i rapportens slutkapitel. Avsikten är att denna förstudie ska följas av fortsatt arbete för att utveckla och göra metodiken mera känd, där ett viktig moment är att demonstrera tillämpningar i genomförda och planerade byggprojekt. 

Riskbedömningar och sannolikhetsbaserade betraktelsesätt i samband med berg- och tunnelbyggande är ett område som uppmärksammas internationellt, vilket är ett av motiven för att publicera denna rapport på engelska. Det är samtidigt viktigt att föra ut informationen till dem som arbetar här i landet med kvalificerade projekterings- och bygguppdrag, vilket föreslås ske i seminarieform. 

Tomas Franzén 


Summary 

The objective of this report is to discuss the design issues related to underground excavations with a risk design (or better expressed information based design) perspective. Design work for an underground project involves much more than structural engineering. The layouts, establishment of alignment, measures to get acceptable environmental impact are all part of the design work. Under some circumstances also construction method has to be addressed by the designer. 
The overall design process of an underground project is characterised by a chain of design decisions taken during different phases. They are all related to each other and the free flow of information from one decision to another is essential. The basis for the decisions is the objectives of the different phases. Other important issues are the outer requirements as the geology and topography, the owner’s functional and economical requirements and the prerequisite from the society. These issues are very important part in the process of identify and analyse the problem. 
Different solutions to the design problem may exist. However, a solution to a design problem is normally built up by different components in a more or less complex interaction. An essential tool for handling these components and interactions is to describe them as a system, which will enable a reliability analysis. Different hazards, uncertainties related to the outer requirements and limits in the detailed understanding of the rock mechanics are all very specific for underground project and have to be taken in account in the design work. The complexity and involved uncertainties characterise the design process. Characteristic for many design situations in rock engineering is this fact that the decision has to be taken under uncertainties, which implies that a good design management and quality assurance are also very essential. 
Every design decision can thus be described by the following seven steps: 
· Problem analysis and system identification. 
· System analysis. 
· Analysis of the uncertainties and probabilities connected to the parameters of the system. 
· Reliability analysis of the system. 
· Decision analysis based on estimated probabilities and consequences 
· The need of technical measures to improve the uncertainties and reduce the consequences like additional investigation and observation during construction. to be used in the decision analysis. 
· Analysis of important information to flow through the project.(Design management and quality assurance). 
The report has therefore been divided into different chapters each describing one of the steps presented above. 
Designing underground openings in rock 
In principle the designer has to establish that the bearing capacity is higher than the load factor to a certain degree. In this respect the design situation is similar to other design situation. However, the mechanical system is normally much more complicated. The mechanical system can in principle be described as an interaction between the rock mass and the installed support. It can also be described as in principle an unloading situation. The stress changes will give a typical deformation pattern with movements directed towards the opening. Thus, the basic problem to design the rock support measures cannot be analysed from the concept of constant load on the support since the load is deformation dependent. A more deeper analyses of the mechanical system is required. 
In principle, underground construction projects are unique as the conditions and demands vary from one project to another. The most crucial factor is that the rock as the building material cannot be prescribed. The process of designing constructions in rock can thus be characterised by the condition that the final design cannot be completed before the rock has been inspected and actual condition been determined after the opening has been excavated. From a designer’s perspective the preliminary design must be based on a relevant estimation of the actual conditions. The design must be able to distinguish between the uncertainties associated with limited information of the actual rock condition and the uncertainties related to the rock mechanics models for the structural design and the corresponding rock mass properties given an inspection of the actual conditions at a certain tunnel reach. 
The design of today can be described as to great extent a subjectively based design. In principle three different design-approaches can be recognised, different empirical based design methods, numerical calculations and the “Observational method “ or “Active design” methods. Large underground openings have been built or will be built based on these design approaches and to full satisfaction. However, in many cases the question if the design is over conservative has been raised. On the other hand several accidents have also occurred after the openings have been taken in service. Both these observations indicate that there are limitations of the knowledge and design approaches of today. Knowledge transfer and extrapolation are normally very difficult with empirically and subjectively based design approaches. According to our knowledge there exists no overall theory today where these different design approaches can be united and the reliability of the design can be expressed in adequate terms like probability of failure or safety index. 
The pronounced uncertainties involved with underground construction have implied that risk analysis have been a very interesting tool to obtain a better understanding of the related problems. Risk analysis have been used as a part in different proposed method to better estimate the time and cost consequences of a tunnel project or to take decision of construction method and suitable working procedure to be able to in a adequate way handle a difficult and dangerous situation. 
It is important to point out that hazards are not only related to technical matters like geology. Hazards can be found in all types of activities related to the design process. The organisation of the work or the contract for the design process may be built up in such way that they contain potential treats to a successful completion of the design work. Such matters or obstacles have also to be analysed when the problem has to be identified. 
Approaches to problem analysis and system identification 
The key to a successful problem solution starts out with a careful consideration on what really is the problem for the system we want to analyse. This contains both problem analysis and system identification. 
It turns out that many problems in design and construction of underground openings are to be regarded as very complex and concern the interaction between technology and people. This means that the wealth of problem solution techniques developed within ‘operational analysis’ is of high interest. It must be emphasised that the key to an adequate problem solution is to consider every problem as a decision problem. Behind every solution of a problem there has been a decision where different alternatives have been evaluated and weighed against each other based on the uncertainties related to the different alternatives. 
When entering into an underground design problem the engineer may first consider the task given being quite straightforward. However, what may appear to be a simple engineering issue of, say selecting proper dimensions of reinforcement may turn out to be a wider issue on selecting construction methods, installing a proper control system or the overall design of the excavation. In a wider perspective a simple task is usually a part of a much wider context. For proper problem identification the technical project and the people involved in the project should be assessed as an integrated system developing in time. There are well developed means of defining and analysing systems, which we recommend using. 
For any civil engineering project operational analysis provides insights in the overall issues and strategy for handling a project. Specifically, experiences gained are useful as regards: 
· Actors – who is affected (directly or indirectly) by the project? 
· System and system identification – what is the problem about and what lies outside the project? 
· Methods and models – how to analyse a problem? 
· Uncertainty, risk and the decision to find an optimal solution. 
· Communication. 
System analysis and identification 
System analysis is key to the problem identification. A system is an entity, which consists of different parts interacting though processes and event. The relations – the interactions – are at least as important as the individual parts. System identification can be made by different means. In simple cases reasoning and assessment of the key factors of the mechanical system to be analysed may be sufficient. However, in other cases the system may contain many different parts, whose interactions are not evident. In such cases more formal approaches for system identification could be used. An important example of formal methods for system identification is the interaction matrix approach. Fault trees and event trees are other important tools for system analysis. 
Uncertainties and probabilities 
Many design situations can be handled by risk based analyses. It implies that the uncertainties have to be expressed in terms of probability. This involves several considerations that have to be addressed. Examples of theses are the issues of the variations of a properties in the space and taken over a certain volume (mean value process). Handling and describing uncertainties is thus essential for proper decision analysis. However, even if the risk analysis require quantitative uncertainty estimates, there are many decision situations where rough uncertainty estimates suffice for making a decision. 
There are also less quantifiable uncertainty relating to future events (scenarios) and the conceptual understanding of the system. Handling these uncertainties is part of the System analysis described in previously. 
This report discusses uncertainties, probabilities and provides some tools for how to describe uncertainties. In general, part of the problem to be solved has to do with that the system is not fully known and that consequences of decisions are uncertain. 
System reliability 
The reliability of a system can be expressed as the likelihood that it will fulfil its given task or achieve its specific objective. There are various tools for exploring reliability. Of particular importance is whether the system is a parallel system, a series system or a combination thereof. 
In order to have a good basis for decisions, it is often necessary to calculate the reliability of different possible designs, construction methods etc. Exact solutions generally implies solving multiple integrals analytically and is thus seldom done. Usual calculation methods for the direct calculation of the probability of failure are numerical integration often by using simulation methods. In the case where a high degree of accuracy is not called for, one might use risk analysis methods to calculate the probability of the occurrence of an undesired event (i.e. the failure). In the construction industry, a proxy safety measure, the safety index b, has come into use. When the safety index is calculated according to certain principles, the probability of failure can be calculated from the safety index. In order to have a safety measure, which is more nuanced than the conventional safety factor and at the same time simple to use, the concept of partial coefficients has been introduced. The basic difference from the ordinary safety factor, different partial coefficients are applied to the different uncertain variables in the limit state expression, at least to the load effect and the resistance. 
Decisions and decision aiding tools 
Decision trees can be used to aid decisions made under uncertainty. Furthermore, just by structuring the problem to be analysed as a decision tree helps in defining the problem. Thus, it is not only the numerical outcome of the decision analysis, but also the very decision analysis itself which eventually guide the decision making. In fact, even when the formal and quantitative decision analysis is performed it is always advisable to assess the outcome on it own merits. Does the decision make sense? Is there a logic, apart from the formal analysis, which can be used to support the decision) 
In some decision problems it is very difficult to evaluate the consequences and thus to use an expected value as a decision criterion. In these cases one might instead use some sort of a ranking scheme, where the different alternatives are compared to each other and ranked according to their judged desirability (without calculating the possible outcomes.) This judgement and the ranking should be made in a systematic and stringent manner in order to avoid psychological biases etc. One method to do such rankings is the Analytic Hierarchy Process (AHP).
After the formal decision analysis it is necessary to assess its reasonableness, its sensitivity to various assumptions and data and to report the findings. The sensitivity analysis should tell whether the decision is robust or very sensitive to uncertainties in data. The findings of this need to be reported along with results of the actual decision analysis. In the end it needs to be remembered that decision analysis does not replace the decision-making – it provides support for the decisions to be made. 
Acquiring information 
From a decision perspective additional information is needed when the best decision is not clear e.g. when a sensitivity analysis shows that small changes in input data can shift the best decision. The additional information can be both related to the result from further preinvestigation or information obtained from observation carried out during the excavation. 
The cost from getting the additional information shall always be compared to the benefit from the additional information. Decision theory can be used to evaluate this issue before any investigation has been carried out based on the cost and the reliability of the method to be used. 
Often in underground construction the so-called “active design” or “observational method” approach is applied. This method is based on a previous analysis of the problem and the determination (in advance) of modifications of the construction procedure to be taken, based on the observation made. 
A special type is the use of an observation system with predefined alarm threshold. The alarm threshold is a predetermined value of a single or a combination of several observation results which if exceeded will trigger pre-determined measures in order to avoid damage. In order to avoid unnecessary alarms or get failure without any warning it is essential to define the threshold in an appropriate way. 
Project management and quality control 
In light of the discussion in the previous chapters of the uncertainties involved in underground projects it is obvious that the needs for quality systems are large. Quality control is from a risk perspective to reduce the probality of failure or the consequence of an unwanted event by using some kind of quality control system. 
Quality work should always be focused upon important factors. Since many underground projects can be described as unique it will equal important to “Do the things right” as “Do the right things”. With others words everything cannot be described in advance, many important decision has to be taken during the excavation work and thus be controlled during the excavation. Depending on the problem different tools may be used like pre defined quality system as ISO 9001 or the use of technical audits. The later type is more directed to check that the right things are done than the things are carried out correctly