Safety Concepts and Calibration of Partial Factors in European. and North American Codes of Practice. 30/11 01/12/2011, TU Delft, The Netherlands

Transcription

1 Safety Concepts and Calibration of Partial Factors in European and North American Codes of Practice 30/11 01/12/2011, TU Delft, The Netherlands The objective of this workshop was to transfer knowledge between European and North American experts on safety concepts and calibration of Codes of Practice. The speakers were asked to describe their national approach as well as the code choices made in terms of design practice, partial factor values and target reliability levels. The overall aim was to identify advantages and limitations as well as similarities and differences between the code approaches and, thereby, to get an overview of the state-of-the-art and to generate ideas for future developments. Three summary statements: 1. Design is a matter of culture. 2. Harmonisation with caution is key. 3. New codes are difficult to use due to lack of familiarity. Day 1 Prior to the technical presentations Michael A. Hicks (TU Delft, Netherlands) welcomed all delegates on behalf of both the Geo-Engineering group of the Delft University of Technology as well as the workshop organising committee. Before opening the workshop, Michael especially thanked both Deltares and Fugro GeoServices for sponsoring this workshop. Gordon Fenton (Dalhousie University, Canada) focused his presentation on the current and future implementation of Load Resistance Factor Design (LRFD) specifications within the National Building Code of Canada as well as the Canadian Highway Bridge Design Code. Gordon pointed out that the proposed reliability based design procedure will be founded on a three level resistance factor accounting for site and prediction model understanding, as well as on a consequence factor based on three levels of failure consequences. Using Random Field theory in combination with Monte Carlo analyses, Gordon proposed a calibration scheme for both resistance and consequence factors in a rigorous manner based on statistically homogeneous random fields accounting for the inherent spatial variability of soil properties. Discussion topics included how theory will be calibrated with practice and how Canada intends to define characteristic values these will be aimed at the mean for resistances. Andrew Bond mentioned that this might make comparison with the Eurocode more difficult because the Eurocode uses a 5-percentile for the characteristic resistance. 1/6

2 In his presentation on LRFD Calibration and Implementation Samuel G. Paikowsky (University of Massachusetts Lowell, USA) started with a review and summary of the past and present design practice in the United States. Using a database of load tests on driven and bored piles, Samuel highlighted the importance of integrating the specific design method within the calibration process using bias and efficiency factors in order to quantitatively assess the safety and economic consequences of a certain design. Furthermore using a database of shallow foundations tested to Ultimate Limit State (ULS), Samuel demonstrated that the model uncertainty, e.g. in the bearing capacity factor Nγ, may be THE primary source of uncertainty within the design process. In the discussion following Samuel's presentation it was pointed out that although maybe disagreeing with national/local design practice and engineering cultures, cautious harmonisation is required e.g. using the same set of data for calibration from databases. Adriaan van Seters (Fugro GeoServices, Netherlands) introduced the Dutch approach on Geotechnical Design by Eurocode 7 (EC7) using Design Approach (DA) 3. Adriaan pointed out that in the Netherlands (NL) probabilistic design methods have been used to calibrate structural and geotechnical design codes over the past several decades. Following up on a point raised by Samuel Paikowsky (partial factors are a function of the design method in use), it was interesting to note that in the Netherlands, the partial factor on the undrained shear strength γ c u for slope stability analyses is chosen to be much higher ( ) in the Dutch National Annex than the 1.4 set as default in EC7. According to Adriaan this is due the greater uncertainty of this parameter compared to drained parameters and to the fact that experience in designing geotechnical structures using an undrained analysis is very limited in the Netherlands. Also discussed was the fact that the partial factors on cohesion and friction angle for slope and retaining wall design were not identical. Adriaan pointed out that the uncertainty of the effective cohesion was much larger than that of the friction angle and that the influence of the cohesion value on design of these structures is much greater. Therefore a higher material factor is used in NL on cohesion than on friction angle. Furthermore, variable partial factors on soil parameters γ M (EC7 - Annex A) have been defined for various Reliability Classes in slope and retaining wall design, and this was considered to be a good choice by the delegates. The French approach on implementing EC7, which is based on DA 2 and in some cases DA 3, was presented by Jean-Pierre Magnan (IFSTTAR, France). Jean-Pierre pointed out that one of the main requirements in France's migration to EC7 was to retain their site investigation scheme (new EC7 characteristic values = existing design values) as well as to retain their currently used design safety levels. In order to match the EC7 with their old global safety factors, a systematic set of model factors has been introduced, rather than adjusting the default partial factors values given by the EC7. Jean-Pierre pointed out that calibration and validation of the model factors is strongly based on experience (e.g. by using proven methods such as the pressuremeter) and case histories rather than on statistical concepts. In the discussion on the topics of day one, chaired by Brian Simpson (Arup, United Kingdom), the 2/6

3 following topics were discussed; Jean-Pierre Magnan opened the discussion using a schematic pile installation design problem pointing out the sources of uncertainty involved. Two important points were made in the subsequent discussion; Although it was agreed that accounting for effects resulting from soil disturbance due to pile installation is crucial and that to some extent it is possible to capture these effects through experimental and numerical studies, it was argued that it is necessary to calibrate partial factor values using the models and methods which are employed in each individual national engineering practice. It was pointed out that by calibrating pile designs against CPT tests, the uncertainty e.g. in the lateral stress close to the shaft may be captured to some extent. The use of stochastic/reliability based methods in calibrating partial factor values was not approved of by some delegates, mainly due to the belief that insufficient data are available to calibrate the methods and/or due to unfamiliarity of most designers with the stochastic tools available. However, this disapproval was countered by the argument that by relying solely on experience, no sound basis exists to evaluate the reliability of a geotechnical system other than through the claim that it was safe before, especially when analysing complex design situations. The harmonisation of the EC7 code provisions within Europe was discussed, focussing on the fact that many national annexes deviate from the default partial factor values proposed in EC7. It was suggested by some that it might be better to introduce model factors according to the design methods used on a national level instead of adjusting the default factors on resistances and actions. However, it was also recognized that even if the same design methods are used in different countries, the different perceptions regarding acceptable risk levels based on national cultures and experience may prevent a full harmonisation of design codes whether that is an achievable and/or desirable goal is still questionable. (The debate on harmonisation continued in final discussion on Day 2). Day 2 Aiming to obtain a broad, but not necessarily identical, compatibility with previous designs, Brian Simpson (Arup, United Kingdom) presented the reasons for the British choice of DA 1 and associated partial factors in their implementation of the EC7. Deviating from the default values of FORM sensitivity factors on actions and resistance (EC0 - Annex C7), Brian pointed out the importance of having a process that could accommodate a wide variety of geotechnical design situations, i.e. to achieve a safe and economical design for either strongly load or resistance dominated problems. Brian further elaborated the reasons for generally applying partial factors on actions before combining them into resultant forces, and for factoring soil strength properties in preference to resitances. He explained why reliability based design calculations had not been used in developing factors for the UK National Annex. Using a pile design example Brian described the UK system for using EC7 with design based on a combination of calculation and load testing. 3/6

4 Beata Gajewska (Road and Bridge Research Institute, Poland) presented the current state and future aims of the implementation and harmonization of EC7 with the Polish Standards. Beata pointed out that in Poland standards are generally not obligatory. In addition, currently several national regulations state that designs shall be made according to `standards' without actually specifying which standards to use. With soil parameters commonly taken from tables provided in the site investigation report or the available standards, the use of expert judgement and reliability based methods is rather uncommon. Beata further stated that several design situations which have not yet been specified in the Polish Standards, e.g. uplift and hydraulic heave, will be adopted in future standards. One of the main goals of the Polish Geotechnical Standard Committee is to agree on a common approach for actions (EC0 & EC1) and resistances (EC7) in terms of structural reliability and calibration measures, leading to an improvement in the economic evaluation/judgement of designs. Following a review of the history of Geotechnical Design in Germany, Kerstin Lesny (University of Duisburg-Essen, Germany) presented the German approach to EC7, leading to their implementation scheduled for the beginning of July Kerstin noted that overlapping regulations currently in use sometimes lead to non-transparent design criteria -- on the basis of a survey amongst design engineers, Trevor Orr also discusses this same issue in a later talk. Choosing DA2 for piles, DA2* for foundations and DA3 for slopes, Kerstin showed, based on the results of a bearing capacity calculation and a pile design, the similarities between the old total factor approach and the EC7 partial factor approach using DA2. She also pointed out the differences obtained when using DA1 or DA3. Calibration of the partial factor values has been achieved by comparing the safety levels of designs to EC7 with the `acceptable' safety levels obtained using the traditional design approach. No probabilistic approaches have been used, which means that the actual reliability levels remain unknown, but presumably acceptable. Hans R. Schneider (University of Applied Sciences Rapperswil, Switzerland) emphasised the importance of accounting for the size of the potential failure zone, relative to the scale of fluctuation, in defining the characteristic soil property values X k. A simplified method for determining X k in design practice was presented which incorporates the effects of the soil's spatial variability. Hans pointed out that EC7 is well accepted in Switzerland and applied to all routine engineering design problems. However, it was felt by the designers that more guidance and details should be provided in some cases; e.g., for the use of Finite Element Analysis as well as for the design of Reinforced Earth and structures with soil nails. Furthermore, Hans emphasized that the definition of characteristic values across the Eurocode is not always consistent, e.g. when comparing concrete, steel and geotechnical codes, and this lack of consistency poses difficulties in comparing relative safety levels. Hans proposed a simplified procedure to determine the characteristic soil property values which was based on values commonly provided by the ground investigation report in Switzerland and which also considered the extent and quality of the soil investigation. 4/6

5 Based on a survey amongst design engineers and code developers, Trevor Orr (Trinity College Dublin, Ireland) presented a summary of the EC7 implementation experiences in Europe. Trevor stated that most of the CEN member countries implemented EN (EC7) in 2010 and that it is now commonly used for routine designs. However, the code was criticised by respondents from some countries for not being easy to use, mainly due to unfamiliarity with the code, and more details and guidance were commonly desired, e.g. on the use of numerical methods, on soil characterisation, and on the design of specific structures such as dams and tunnels. All three Design Approaches are in use by various countries, with DA2 and DA3 being most popularly adopted for foundations/retaining structures and slopes, respectively. However, to date, only a few countries have linked the Nationally Determined Parameters (NDP) with risk and reliability using Reliability Classes (Denmark and the Netherlands), Consequence Classes (Austria and partially Finland) or Safety Classes (Sweden). This fact may, however, relate back to the national experiences with using reliability based approaches in structural designs. Arising from the subsequent discussion, one specific point Trevor made is the fact that in some countries, mainly due to legal reasons, the characteristic soil property values are not chosen by the geotechnical design engineer. Andrew Bond (Geocentrix, United Kingdom) discussed the proposed future and evolution of EC7. Further developments of EC7 envisaged were addressed with respect to the different stages involved, up to the Final Vote on the new and revised Eurocodes in With an emphasis on harmonisation, sustainability, simplification and extension, Andrew, as chair of the EC7 committee, addressed some of the points of discussion raised during this workshop. For example, Andrew discussed the work of the EC7 Evolution Groups (Egs): To date 11 EGs have been established (with a further two under consideration) who work on the development of the new and revised second version of EC7 via regular web-meetings. Andrew invited everybody interested to join the EGs and to participate in the evolution of EC7. In the final discussion, chaired by Andrew Bond (Geocentrix, United Kingdom), the following topics were discussed; Continuing the discussion on harmonisation from Day 1, one of the main points arising from the presentations is that although EC7 provides a set of default partial factors, most countries end up using their own factors. This may be partially due to two reasons. National code developers commonly want to retain traditional safety levels which have been established over years of engineering tradition, Different investigation/design methods are used in different countries, e.g. the pressuremeter in France or the CPT in the Netherlands. Since the partial factors should not be invariant with regard to the investigation method used, the use of different partial factors is reasonable. There was some discussion about how these differences should best be accommodated and the general consensus was that the use of model factors would be preferable to varying the partial factors. Model factors can be related specifically to the overall investigation/design model used 5/6

6 and this has the added benefit of helping all countries assess or compare their various approaches. With reference to examples shown during Andrew Bond's presentation, it was pointed out that this divergence is one of the major concerns in the current evaluation process of EC7. Based on the presentation by Andrew Bond on the evolution of EC7, the question about whether the EC7 is too complicated was discussed. One obvious way to reduce its complexity is to reduce the number of Design Approaches (DA's), e.g. from three to two, but how to do this remains an open issue. It was commonly agreed that expert judgement by the geotechnical engineer should be involved in defining the characteristic soil property values. The method proposed by Hans Schneider, where the characteristic value takes account of the variability of the soil parameters and the amount of information in the ground investigation report as well as the extent of the failure zone, was generally believed to be very reasonable. Frans Barends pointed out that the term Engineering Factor may be used to account for design uncertainty based on subjective choices made by the geotechnical engineer as a result of both individual knowledge and personal experience. The organising committee would like to thank the presenters, the sponsors, as well as the participants for making this workshop a success. We hope that this workshop was a first step in the initiation and extension of the transfer of knowledge on safety concepts and Codes of Practice between Europe and North America. Delft, 15/01/2012 Patrick Arnold Updates to previous version from 15/12/2011: 1.) Rewording - Presentation of Beata Gajewska with respect to legal status of codes of practice in Poland. 2.) Addition - Last item of final discussion regarding the Engineering Factor in geotechnical design. 6/6