Evaluation of Failure Modes forConcrete Dams

(1)

DEGREE PROJECT, IN CONCRETE STRUCTURES , SECOND LEVEL STOCKHOLM, SWEDEN 2015

Evaluation of Failure Modes for Concrete Dams

LISA BROBERG & MALIN THORWID

KTH ROYAL INSTITUTE OF TECHNOLOGY

(2)

(3)

Evaluation of Failure Modes for Concrete Dams

Lisa Broberg & Malin Thorwid

June 2015

TRITA-BKN. Master Thesis 455, 2015 ISSN 1103-4297,

ISRN KTH/BKN/EX–455–SE

(4)

Department of Civil and Architectural Engineering Division of Concrete Structures

(5)

Abstract

The safety of a concrete dam is ensured by designing according to failure criteria, for all combinations of loads using safety factors. Today in Sweden, RIDAS, the Swedish power companies’ guidelines for dam safety, is used for the design of dams and is based on BKR, the National Board of Housing, Building and Planning. Swedish dams are designed to resist two global failure modes; sliding and overturning. Com- bination of failure modes, that should be considered in the design of concrete dams, is however fairly unknown. Since 2009 the Eurocodes was adopted and came into force 2011. The Eurocodes have replaced BKR in the design of most structures in Sweden where the partial factor method is used to ensure safety in the design.

The objective of this report was to examine if the design criteria for concrete dams in today’s condition are enough to describe real failure modes. The other objective was to analyse if Eurocode is comparable to RIDAS in dam design. The stated questions were answered by performing a literature study of known dam failures and analytical calculations for different types of concrete gravity dams, with varying geometry and loading conditions. The programs CADAM and BRIGADE were also used as calculation tools to further analyse if failure occurred as expected.

The results from the analytical calculations together with the performed FE analysis indicate that limit turning does occur and often generate lower safety factors compared to overturning. Limit turning is similar to overturning failure although it accounts for material failure in the rock. This design criterion is therefore, highly dependent on the quality of the rock and requires investigations of the foundation to be a good estimation of the real behaviour of the dam body.

From the compilation of reported failures the conclusion was that the current design criteria are adequate. However, the real challenge lies in ensuring that the construction of dams is correctly performed to fulfil the stated criteria. A transition to Eurocode appears to be reasonable for the stability criterion. A modification of the partial factors is nevertheless necessary to obtain result corresponding to RIDAS, especially for the overturning criteria.

Keywords: Gravity dams, concrete, design criteria, RIDAS, Eurocode, limit turning

(6)

(7)

Sammanfattning

För att uppnå säkra dammkonstruktioner, för alla lastkombinationer, dimensioneras dammar enligt bestämda brottvillkor som ska uppfylla en viss säkerhetsfaktor. Idag används RIDAS, för dimensionering av dammar i Sverige. RIDAS Kraftföretagens riktlinjer för dammsäkerhet, är baserat på BKR, Boverkets konstruktionsregler. I Sverige dimensioneras dammar för att motstå de två globala brottmoderna glidning och stjälpning. Frågan som behöver besvaras är om det finns eller kan finnas några kombinationer av brottmoder som borde beaktas vid dimensionering av dammar.

2009 antogs Eurokoderna och trädde i kraft 2011. Eurokoderna har ersatt BKR vid dimensionering av de flesta konstruktioner i Sverige. I Eurokod används partialko- efficienter för att garantera säkra konstruktioner.

Syftet med denna rapport var att undersöka om dagens brottkriterium är tillräck- liga för att beskriva hur dammar går till brott. Rapporten behandlar även möj- ligheten att övergå från att dimensionera dammar enligt RIDAS till att dimensionera enligt Eurokod. För att besvara detta genomfördes en litteraturstudie av rapporterade dammbrott. Dessutom genomfördes analytiska beräkningar för flera olika typer av dammar med varierande geometri och lastfall. Programmen CADAM och BRIGADE användes som ytterligare verktyg för att analysera brottmoderna.

Enligt resultat från de analytiska beräkningarna tillsammans med FE-beräkningarna ansågs limit turning inträffa och genererade i högre grad en lägre säkerhetsfaktorer i jämförelse med stjälpning. Limit turning kan förklars som delvis stjälpande och inkluderar brott av bergmassan. Brottmodet är dock beroende av kvalitéten hos berget och det krävs undersökningar av grunden för att kunna göra en god uppskat- tning av dammens verkliga beteende.

Sammanställningen av tidigare brott visade att nu gällande brottkriterier är lämpliga och troligtvis tillräckliga. Utmaningen är istället att säkerställa att konstruktion- erna är korrekt utförda och därmed uppfyller dessa brottkriterier. En övergång till Eurokod tycks vara möjlig för de globala brottmoderna, dock är det väsentligt att partialkoefficienterna justeras för att uppnå resultat som överensstämmer med RIDAS, särskilt för stjälpning.

Keywords: Gravitationsdammar, betong, brottkriterium, RIDAS, Eurokod, limit turning

(8)

(9)

Preface

This thesis was carried out from January to June 2015 at SWECO Energuide AB in collaboration with the Division of Concrete Structures, Department of civil and Architectural Engineering at the Royal Institute of Technology (KTH). The project was initiated by Dr. Richard Malm, who also supervised the project, together with Ph.D. candidate Daniel Eriksson and Adjunct Prof. Erik Nordström.

We would especially like to thank Richard Malm for the continuous support which has been a great encouragement. We would also like to thank Daniel Eriksson for always finding time to help and guide us throughout this project. We also wish to thank Erik Nordstöm for his guidance and advice.

We would like to thank the division at SWECO Eneriguide AB for their warm welcome and for an inspiring job environment. We would like to give special thanks to Johan Nilsson for the support and help during our work at SWECO.

Stockholm, June 2015

Lisa Broberg & Malin Thorwid

(10)

(11)

Chapter 1 Introduction

1.1 Background

Most dams in Sweden were built during 1950 to 1960 on solid good quality rock.

The dams were built under different conditions and safety regulations compared to the demands stated today (Andersson, 2012). The knowledge of rock mechanics and material properties of concrete along with the building techniques have improved.

Today the construction of new dams in Sweden is limited by regulations concerning the preservation of the environment. Therefore, the design of dams mostly involves maintenance and reparation of existing dams.

Knowledge of why and how dam failure occurs, may help prevent or minimise the damage. The indication of how a dam behaves prior to failure is therefore of great importance in order to prevent failure. It is also important since it provides guidance on how to monitor and measure dams, what types of sensors and where these sensors should be placed to get early indications of possible failures. Risk and safety are essential in dam design due to the radical consequences a failure would cause to the surroundings, as seen in Figure 1.1.

Figure 1.1: Baldwin Hills Reservoir after the disaster 1963 (Wilson, 1963).

(14)

The consequences of failure could in the worst case scenario lead to lives lost and economic damage. Therefore, the government through the public utility Svenska Kraftnät, stated new requirements concerning higher safety demands for the existing dams (SFS 2014:114). The new requirements also concern the classification of the dams in Sweden; all dams must be classified, if a failure could result in severe consequences.

In Sweden, the dam owner is responsible in the event of a failure or an accident.

The Swedish power companies have issued the Swedish power companies’ guidelines for dam safety, RIDAS, based on the construction rules BKR (2010), the National Board of Housing, Building and Planning. Since 2011, the Eurocodes (the European construction standards) together with EKS 9 (2013), have replaced BKR in Sweden.

However, Eurocode does not account for the design of dams (Andersson, 2014).

Today Swedish dams are designed to withstand two global failure modes; sliding and overturning of the entire monolith. There are questions regarding if there are or can be any combinations of the failure modes, that should be considered in the design of concrete dams.

1.2 Aim of report

The main focus of this report is to analyse the different types of failures that have occurred and can occur in concrete gravity dams, by examining the influence of different factors. The aim of this study is to answer the following questions:

• Is analytical calculations based on the global failure modes: sliding and overturning enough to describe the failure of the dam? Are there other potential failure modes not covered by these analytical calculations?

• Is a transition to Eurocode possible for dam design? Are the design guidelines according to Eurocode comparable to RIDAS?

The stated questions will be answered by performing a literature study of reported dam failures and analytical calculations for several concrete gravity dams. In addition a FE-analysis will be performed for comparison.

1.3 Limitations

This report only include concrete gravity dams and further limitations for the literature study are stated in Chapter 4. The analytical calculations only concern dams on rock foundation. The analyses are limited to Swedish conditions regarding material properties and design parameters. The influence from seismic loads is not included due to that it is not considered for design in Sweden, while it internationally may be of great importance for the design.

(15)

1.4. STRUCTURE OF REPORT

1.4 Structure of report

Chapter 2 includes a presentation of theory behind the key concepts of concrete gravity dams. The main features for stability analyses of concrete gravity dams are presented. The different design loads for the stability analysis are stated and illustrated. A brief presentation of the causes of concrete gravity dam failure by explanation of the failure modes is presented.

Stability analyses according to the guidelines RIDAS and Eurocode are presented in Chapter 3. How the guidelines account for the design loads and explanations of the analytical calculation methods for the failure modes described in Chapter 2 are given.

A compilation of reported failures including causes and failure modes is presented in Chapter 4. Already known facts about the presented failures of concrete dams are compiled and the sources of failures that has occurred are detected.

The analytical analyses in Chapter 5 describe the stability calculations for several different dams with varying geometry and loading conditions. A parameter study is used to determine the most influential parameter and to adapt the design of dams according to Eurocode with the stability calculations in agreement with RIDAS.

The program CADAM and the software BRIGADE, are used as tools to enable a comparison to the analytical calculations. The key concepts on how to perform stability analyses with these tools are presented in this chapter. The results from the analyses described in Chapter 5, are presented in Chapter 6.

In Chapter 7, the conclusions of the different concrete dam failure analyses are presented.

Appendix A includes the full compilation about the failures of concrete dams presented in Chapter 4. Appendix B includes the results from the analytical calculations described in Chapter 5. The loads and level arms from the analytical calculations are presented in Appendix C.

(16)

(17)

Chapter 2 Concrete gravity dams

There are different types of concrete dams, which are distinguished by how the water pressure of the reservoir is transferred down to the ground. Descriptions of the different types of concrete gravity dams are included in this chapter. The two most common types are massive and buttress dams, presented in Section 2.1.1 and Section 2.1.2 respectively. Gate sections mainly consist of pillars and spillways, both are of gravity dam type and are described in Section 2.1.3. Figure 2.1, shows the different dam types included in this report.

Figure 2.1: The different types of dams included in this report, a) spillway and pillar, b) massive and c) buttress.

For other types of dams, not included in this report, concrete can also be used in embankment dams, as a central or upstream membrane, as retaining walls for spillways or used for many secondary functions. Embankment dams are usually associated with at least one concrete dam part, either intake and/or discharge facilities. Con- crete arch dams were introduced relatively late and therefore have a uniform, and somewhat higher quality. Arch dams are founded on rock and are of slender type with concaved arches (Kleivan et al., 1994).

(18)

2.1 Gravity dams

2.1.1 Massive dams

Massive or gravity dams are solid structures, designed to resist the external forces by its dead weight. Today, gravity dams are mainly constructed with concrete, compared to the previously used method of stone masonry. The development of new concrete gravity dams is ongoing and the Roller-Compacted Concrete dam (RCC) is an example of that. The RCC dam has a limited use of formwork, consists of a drier mix and is placed in a manner similar to paving, i.e. compacted with vibrating rollers. The benefits are cost beneficial with simple faster construction. The dams are built with no joints or reinforcement, with low cement content and the use of fly ash that enable less heat generation while curing (Kleivan et al., 1994).

Solid concrete structures maintain stability against loads due to the geometric shape, mass and strength of the concrete (Ali, 2012). A gravity dam consists of either a continuous or a series of concrete monoliths separated by expansion joints (RIDAS, 2011). The monolith cross section is, in principle, triangular with a dam head, an inclination of the downstream face and a vertical upstream face, which also can have a small inclination, see Figure 2.2. The benefits of concrete gravity dams are that they are easily constructed on sites with a foundation of sufficient strength to carry the weight of the dam (Ali, 2012).

Figure 2.2: Typical cross-section of a massive dam, reproduced from Bergh (2014).

The monoliths are mainly placed in a straight line or sometimes slightly curved, see Figure 2.3 and are usually of a width between 5 m to 10 m (Bergh, 2014).

(19)

2.1. GRAVITY DAMS

Figure 2.3: Stadsforsen, massive dam in Sweden (Malm, 2015).

2.1.2 Buttress dams

Over time, there has been a strong effort towards improving concrete quality. There has been a shift from the previously dominant gravity type dam such as massive dams, towards a more slender type of dam with reinforcement, known as a buttress dam (Kleivan et al., 1994).

Buttress dams consist of two rigidly connected elements, the upstream water barrier (frontplate) and the supporting buttress on the downstream side, which together form a monolith, see Figure 2.4. The upstream water barrier transfers the hydrostatic pressure over to the buttress, which in turn transfers it down to the foundation.

The water barrier is inclined so the vertical water loads, together with the weight of the concrete, act in favour for the stability of the monolith (DOI, 2009).

Figure 2.4: Cross-section of a buttress dam (Left), Section of a buttress dam (Right), reproduced from RIDAS (2011).

A buttress dam consists of a series of monoliths, connected by horizontal struts acting as contraction joints, connecting the adjacent monoliths, see Figure 2.5. The casting arrangements and the construction are somewhat more demanding compared to most other types of dams. However, buttress dams are more suitable on

(20)

weak foundations compared to gravity dams, due to the reduced volume of concrete (Bergh, 2014), while the contact pressure between the buttress and the foundation is considerably higher.

Figure 2.5: Rätan, buttress dam in Sweden (Vattenkraft.info, 2009).

2.1.3 Gate section

Concrete gravity dams also consist of gate sections to transport water in specific directions and release water pressure on the dam structure. Concrete functions as a fastener for many different types of gate installations, with variable functions.

The gate type could be sliding, roller or radial gates, with the function of either an outlet gate, intake or daft tube (Kleivan et al., 1994). An example of a gate section in Sweden is shown in Figure 2.6

(21)

2.1. GRAVITY DAMS

For a spillway section see Figure 2.7, where the overflow is designed with a vertical upstream face. The water is able to flow over the crest along the inclined downstream side with training walls, keeping the water in place and finally the water reaches the energy dissipating structure, forming a hydraulic jump to avoid erosion of the riverbed. Pillars, non-overflowing blocks function as enclosures of a number of overflow sections, see Figure 2.7. Usually spillway sections have gates and typically, radial gates see Figure 2.7.

Figure 2.7: Spillway section (Left), plane view of spillway (Right), reproduced from RIDAS (2011).

For hydropower structures, the intake is the connection between the reservoir and the waterway connected to the turbine of the hydroelectric power plant. Intake gates are normally designed with a trash rack preventing debris, ice, fish, etc. from entering the intake. In addition, it also consists of a gate of river or tunnel type, to close of the conduit, see Figure 2.8.

Figure 2.8: Section of a typical inlet and power station, reproduced from RIDAS (2011).

The behaviour of a spillway, discharge part, and intake part of the dam is similar to concrete gravity dams, thus the geometry and loading conditions may be more complicated (Westberg and Hassanzadeh, 2007). The surface of the concrete is subjected to very high water velocities as well as abrasion, and therefore must be steel plated.

(22)

2.1.4 Support methods

There are different support methods for concrete gravity dams. Common methods are to secure the dam body to the foundation by rock bolts or tendons. Rock bolts are non-pre-stressed reinforcement bars installed in the interface between the foundation and the dam body. There are different types of fastners for the rock bolts and they can be secured to the foundation through anchors, cables, dowles or by grouting. The bolts are anchored in the dam body by adhesion. Rock tendons consist of pre-stressed cables or rods, anchorage and corrosion-inhibiting coating.

The tendons can be unbonded or bonded to the surrounding concrete. The tendon is inserted into a casing and grouted after the tendon is stressed (PTI, 2000).

Another common method used to support concrete dams is earth support fill on the downstream side of the dam, giving rise to stabilising forces. A grout curtain helps to stabilise the dam by decreasing the uplift pressure. The grout curtain is achieved by inserting cement into pores and cavities in the ground. The grout curtain might deteriorate over time and therefore the decreased uplift is usually not accounted for in the design of dams (Ferc Engineering Guidelines, 2002). The uplift pressure could also be decreased by inserting drainage pipes.

2.2 Stability analyses

Concrete dams are massive large structures since they are designed to fulfil the requirements for stability. The design should also fulfil requirements for long life spans and water tightness to withstand the permanent water pressure (Bond, 2014).

The safety of the concrete dam is assured by designing according to failure criteria, for all combinations of loads using safety factors. The safety criterion is the definition of the stress level when failure occurs. The safety factors are chosen to provide for all underlying uncertainties. Their magnitude should reflect the probability of the occurrence for the particular load, the accuracy of conditions and the method of analysis. The factor of safety is thereby higher for foundation studies, because of the greater amount of uncertainty in assessing the load-resistance capacity of the foundation.

2.2.1 Design loads

The loads included in the stability analysis should represent the actual loads acting on the concrete dam during operation. Many of the loads are unable to be exactly determined, the engineer is then responsible for estimating these loads based on available measurements, judgement and experience (Ferc Engineering Guidelines, 2002).

The required loads acting on a dam for a stability analysis are shown in Figure 2.9.

(23)

2.2. STABILITY ANALYSES

Figure 2.9: Loads acting on a dam, reproduced from (Bergh, 2014).

1. Hydrostatic pressure (P1-P2) – depends on the water level in the dam 2. Tailwater pressure (P3-P4) – depends on the tailwater level

3. Uplift pressure (P5) – hydrostatic pressure acting vertically, assumed to vary linearly from hydrostatic pressure at the heel to the tailwater pressure at the toe

4. Dead weight (P6) – the weight of the concrete

5. Ice pressure (P7) – load acting on the face of the dam due to an ice cover 6. Silt pressure (P8) – settled sediments exerting active pressure towards the dam 7. Seismic loads (P9-P11) – horizontal and vertical accelerations caused by earth

quakes

2.2.2 Failure modes

The definition of dam failure can differ between individuals, the general definition could be expressed as: "Collapse or movement of part of a dam or its foundation, so that the dam cannot retain water” (ICOLD, 1995).

Concrete dams have various failure behaviour. Sliding or shear failure is the most common failure for dams constructed on rock. The dam may fail due to crushing, i.e. the failure of its materials when the compressive stresses exceed the acceptable stresses. Concrete cannot withstand sustained tensile stress and if the tension that develops in the concrete exceed its tensile strength, it could lead to ultimate failure.

The dam may also fail due to overturning where it rotates about the toe (Ali, 2012).

(24)

Overturning

Overturning occurs when the forces acting on the dam causes rotation of the dam, see Figure 2.10. Overturning is analysed by calculating a factor of safety, which is defined as the ratio of stabilising and overturning moment. These moments are calculated around its toe or another weak point in the structure or foundation. For the overturning failure, it is also important that the resultant is located in the mid third of the base area since this will assure that the whole base of the dam is under compression. If tensile stress can be avoided it will reduce crack propagation in the concrete. The criterion is verified by application of the Navier equation for a cantilever action under combined axial and bending load (Bergh, 2014).

Figure 2.10: Overturning failure around the dam toe.

Sliding

Sliding occurs when the horizontal forces exceed the frictional resistance. Sliding can be divided in to three different kinds of failures (Gustafsson et al., 2008):

1. Failure in the interface between the concrete and the foundation (Figure 2.11 a).

2. Failure in weak planes of the foundation, such as cracks (Figure 2.11 b).

3. Failure in the solid foundation (Figure 2.11 c).

Figure 2.11: Different types of sliding failures, reproduced from Gustafsson et al.

(25)

2.2. STABILITY ANALYSES

There are different views on whether cohesion between the concrete and the foundation can be accounted for in the sliding stability. The reason is that it is difficult to quantify through borings’ and testing. Higher allowable safety factors may be applied, if cohesion is included in the calculations for sliding stability (Ferc Engineer- ing Guidelines, 2002). If cohesion is not accounted for, the concrete-rock interface is treated as unbonded giving a conservative method that may result in expensive and over-strong structures (Krounis, 2013).

Failure in the interface between the foundation and the dam body is normally accounted for in the design of dams. Though failure more often occurs in weak planes of the foundation, see Figure 2.11 (DOI, 2012). Sliding can also occur in the dam body at weak planes such as lift joints or along cracks, this failure is seldom analysed except for high dams (Ali, 2012).

Limit turning

According to Fishman (2007), the classical failure modes sliding and overturning, do not account for material failure. Classical overturning failure is unrealistic as it requires infinitely strong rock and concrete. Fishman infers that the one failure mode, either limit turning or sliding, giving the lowest stability factor should be used for the design of the structure and decisions regarding interface preparation (Fishman, 2009).

Fishman states that the stresses developing below the upstream side will result in a tensile crack along the rock, see Figure 2.12. A compressive zone will be formed in the rock, underneath the toe, due to the applied forces on the dam. When the stresses exceed the crushing resistance of the rock, a crushing zone is formed. The size of the crushing zone depends on the strength of the rock, for a weaker rock the crushing zone will extend further to the upstream side. The turn axis appears where the tensile crack and the compressive crack meet. The concrete and rock will act as a single body and failure will occur when they rotate about the new rotation point.

This failure mechanism is called limit turning, which is similar to the overturning method although it also accounts for the strength of the rock. The result gives a more reliable and conservative safety factor (Fishman, 2009).

Figure 2.12: Limit turning failure.

(26)

(27)

Chapter 3 Methods for stability analyses

In this chapter, the design methods for stability analyses are presented. There are different methods for analysing failure of dams, varying between countries or methods. In Sweden, RIDAS (2011), is used for the design of dams and is based on BKR. 2011 BKR was replaced with EKS 9 (2013). Today, Eurocode is used and function as guidelines providing a common structural design in European countries.

A revision of RIDAS, to incorporate Eurocode instead of BKR has begun. The transition work is led by Svensk Energi, responsible for RIDAS. The research com- pany, Energiforsk, previously known as Elforsk until January 2015, has also started to work with the transition from BKR to Eurocode. The investigations has so far mostly been focused on cross-section design (Andersson, 2014).

For the stability analyses, there is an ongoing project financed by Energiforsk, where a structural reliability based method is under development. In this report, the focus will be on RIDAS, using the deterministic method, compared to Eurocodes semi-probabilistic method, resulting in the use of partial factors, for calculations of stability (Westberg, 2014).

The denotations from RIDAS and Eurocode were used in Section 3.1 and Section 3.2, respectively.

3.1 RIDAS

RIDAS (2011), is based on BKR, with adjustments for specific requirements for concrete dams. BKR, as mentioned before, is not valid today however it may still be used if the contents do not conflict with the Eurocodes (Andersson, 2014). Guide- lines are given for the design of dams together with control and reconstruction of existing dams. RIDAS states requirements for stability, strength and durability of the dam and what criteria to fulfil. The requirements and criteria concern gravity dams, where RIDAS have included the most common types; massive and buttress dams, including spillway, inlet dams and pillars.

(28)

3.1.1 Design loads

RIDAS (2011) includes guidelines to determine design loads acting on concrete dams.

It states how to account for the loads presented below. RIDAS also gives guidance how to account for rock anchors (described below), temperature effects, creep and shrinkage, and traffic loads (included if unfavourable), which are not described in this report.

Dead weight

For design of new concrete dams, the dead weight for reinforced concrete is assumed to be 23 kN/m³ if no material tests are available. For existing dams, the dead weight should be determined from material tests or from information about the design.

Hydrostatic pressure

Both the water pressure on the up- and downstream side should be accounted for.

The most unfavourable combinations of up- and downstream water levels applied to the dam determine the water pressure to be used in the calculations.

Uplift pressure

The uplift pressure distribution varies for different dam types and designs with or without drain pipes. For massive dam structures without drainage where the whole foundation area is under compression, the uplift pressure distribution varies linearly from the upstream to the downstream side.

For massive dams the uplift pressure can be reduced by the use of drain pipes as shown in Figure 3.1 and Figure 3.2.

(29)

3.1. RIDAS

Figure 3.1: Uplift distribution for a dam with a drainage pipe in the rock and a drainage tunnel by the rock surface, reproduced from RIDAS (2011).

The uplift distribution in Figure 3.1 will vary from H to 0.3 · (H − h) + h closest to the drainage tunnel and varies linearly to h at the toe of the monolith, with no uplift beneath the drainage tunnel. H is the headwater level and h is the tailwater level.

Figure 3.2: Uplift distribution for a dam with a drainage pipe in the rock and drainage tunnel in the concrete, reproduced from RIDAS (2011).

The uplift distribution in Figure 3.2 will vary from H to 0.5 · (H − h) + h closest to the drainage tunnel and varies linearly to h at the toe of the monolith. H is the headwater level and h is the tailwater level.

(30)

For buttress dams, the uplift distribution is assumed to vary linearly over the thickness of the frontplate. If the buttress is thicker than 2 m, the uplift pressure underneath the buttress should be included as shown in Figure 3.3.

Figure 3.3: Distribution of uplift pressure for a buttress dam with a buttress thicker than 2 m, reproduced from RIDAS (2011).

For spillways, the uplift distribution is assumed similar to massive dams. The uplift distribution in Figure 3.4 is assumed for pillars, where w is the width of the pillar.

The uplift varies from full uplift pressure to zero at the distance w from the spillway.

Figure 3.4: Uplift distribution for pillars, reproduced from RIDAS (2011).

The effect from cement grouting is not considered in the uplift pressure distribution, due to the strength of the cement decrease with time due to deterioration. The grout curtain is only considered as extra safety and should not be accounted for unless re-grouting is possible. This is seldom the case due to difficulties incorporating re-grouting tunnels in the dam body design.

(31)

3.1. RIDAS

Ice load

The intensity of the horizontal ice pressure depends on the geographic location, altitude and local conditions for the dam. RIDAS suggest the horizontal ice pressure 50 kN/m with an ice thickness of 0.6 m, for dams located in the southern part of Sweden. Dams located in the middle part, should be designed for an ice load of 100 kN/m with an ice thickness of 0.6 m. An ice load of 200 kN/m with an ice thickness of 1 m, is suggested for the rest of Sweden. The resultant of the ice pressure is assumed to be located at one third of the ice thickness, calculated from the top surface of the ice.

Rock anchors

For lower dams, it can be hard to achieve stability, and according to RIDAS in these cases it is allowed to assign a load capacity of 140 MPa to the rock bolts. This is applied for dams that have a headwater level less than 5 m and do not belong to any of the two highest safety classes.

For all other dams, rock anchors should not be considered in the stability calculations, due to the complications of verifying their strength. However, it is stated that the installation of rock anchors of the dimension φ25 − 32 is a good preventive measure.

Earth pressure

Soil may be added as downstream support fill to increase stability. The earth pressure should be determined as a at-rest pressure and it should be calculated as the lowest theoretical pressure that may occur. The soil density and earth pressure coefficient should be obtained from in-situ tests. If testing is not possible, the values in Table 3.1 may be used.

Table 3.1: Example values for unit weight and coefficients for earth pressure RIDAS (2011).

Material Unit weight density [ kN/m³] Friction angle [ ^◦] Coefficient for earth pressure [-]

Un-saturated Saturated φ At-rest K0 Active Ka

Rockfill 17.5 11 42 0.33 0.20

Gravel 18 11 35 0.43 0.27

Sand 18 11 32 0.47 0.31

Moraine 21 13 34 0.45 0.29

(32)

3.1.2 Failure modes

According to RIDAS (2011) there are three failure modes that need to be analysed for stability; sliding, overturning and the bearing capacity of the concrete and the foundation.

Dam stability analyses are performed using safety factors for overturning and allowable friction coefficients for sliding, to achieve a safe design.

RIDAS has listed different load combinations to be analysed, these are divided in to; normal load combinations, exceptional load combinations and accidental load combinations. The loads are calculated without partial factors and are analysed for individual monoliths.

Overturning

For overturning the requirement of the safety factor s is defined according to Table 3.2.

Table 3.2: Saftey factor for overturning.

Load case Safety factor (s)

Normal 1.50

Exceptional 1.35 Accidental 1.10

The safety factor defines the relation between stabilising and overturning moment, see Equation (3.1), and should not be lower than the safety factor in Table 3.2.

s = Mstab

M_over (3.1)

Sliding

RIDAS states that sliding should be analysed between the interface of the concrete and in the foundation, along potential weak planes and in weak points in the dam body. Stability against sliding is achieved if the sum of the horizontal forces divided by the vertical forces, see Equation (3.2), does not exceed the maximum allowed friction coefficient, see Table 3.3.

µ = R_H

R_V ≤ µ_max = tan δ_g

s_g (3.2)

(33)

3.2. EUROCODE

where

R_H is the resultant of the horizontal forces.

R_v is the resultant of the vertical forces.

tan δ_g is the friction angle.

s_g is the safety factor.

Table 3.3: Maximum friction coefficient, µ_max

Foundation Normal load case

Exceptional load case

Accidental load case

Rock 0.75 0.90 0.95

Gravel, sand and moraine 0.5 0.55 0.60

Silt 0.40 0.45 0.50

The maximum friction coefficient can be calculated according to Equation (3.2) where the values for the safety factor s_g are presented in Table 3.4.

Table 3.4: Saftey factors s_g for calculations of µ_max.

Foundation Normal load case

Exceptional load case

Accidental load case

Rock 1.35 1.10 1.05

Gravel, sand and moraine 1.50 1.35 1.25

Silt 1.50 1.35 1.25

Cohesion between the dam and the foundation should not be considered in the calculations for the resistance against sliding according to RIDAS.

When calculating stability against sliding, the value for the maximum allowed friction coefficient (µ_max) is calculated according to Equation (3.2), with values of the safety factor from Table 3.4 and the friction angle from geotechnical investigations.

The values in Table 3.3 can be used for dams constructed on a foundation of good quality, when calculating stability against sliding.

3.2 Eurocode

Eurocode is the European standard for technical rules in construction work, providing a common structural design tool in European countries. Eurocode clearly states that their guidelines do not cover the design of dams. This is due to the high safety required for dams and that other aspects than for usual design need to be considered. This section is therefore solely based on the authors assumptions on how to apply Eurocode to dam design. Therefore, in this report a compilation of information from the listed Eurocodes below was performed in order to obtain a method applicable for stability analyses of dams.

(34)

• EN 1990 “Basis for structural design”.

• EN 1991 “Actions on structures”.

• EN 1992 “Design of concrete structures”.

• EN 1997 “Geotechnical design”.

The same failure modes as described in Section 2.2.2 were analysed. The values and equations in the following sections were based on the assumption that dams can be considered as comparable with retaining wall structures.

3.2.1 Design loads

The safety is applied on the loads by partial factors. The load acting on the structure is multiplied with the partial factor γ to define the design load. The loads are classified according to variation in time: permanent or variable loads and whether the load is favourable or unfavourable, which would result in different partial factors.

When designing geotechnical structures, different approaches are used. In accordance with retaining wall structures, the concrete dams were assigned the design approach 3 (DA 3). The different approaches give different values for the partial factors; for load and load effects, soil parameters and the strength (EC 7, 2011).

Design approach 3 states that different partial factors should be used for geotechnical actions and structural actions. Geotechnical actions are defined as actions transmitted to the structure by the ground, fill, standing water or groundwater.

For structural actions the strength of the material is significant. For the structural actions Equation (3.3) and (3.4) are used. For geotechnical actions, Equation (3.5) is used (EC 0, 2002). The partial factors for the equations are listed below in Table 3.5.

E_d= Σγ_d· γ_G· G_k+ Σγ_d· γ_Q· ψ_0,i· Q_k,i (3.3) E_d= Σγ_d· ξ · γ_G· G_k+ γ_d· γ_Q,1· Q_k,1+ Σγ_d· γ_Q,i· ψ_0,i· Q_k,i (3.4) E_d = Σγ_d· γ_G· G_k+ γ_d· γ_Q,1· Q_k,1+ Σγ_d· γ_Q,i· ψ_0,i· Q_k,i (3.5) where

γ_G is the partial factor for permanent actions.

γ_d is the partial factor depending on safety class.

G_kg is the characteristic value of a permanent action.

γ_Q,i is the partial factor for variable action.

ψ_0,i is the factor for combination value of a variable action.

Q_k,i is the characteristic value of a single variable action.

ξ is the reduction factor.

(35)

3.2. EUROCODE

Table 3.5: Partial factors according to Eurocode (EC 7, 2011).

Load combination equation 3.3/3.4 3.5 For an unfavourable permanent load γG= 1.35 γG= 1.1 For a favourable permanent load γG= 1.0 γG= 1.0 For an unfavourable variable load γQ= 1.5 γQ= 1.4 For a favourable variable load γQ= 0 γQ= 0

Reduction factor ξ = 0.89/- -

Structures are classified into different safety classes depending on the harm a failure would cause, the definitions are stated in Table 3.6. For calculations of stability, according to the partial factor method in EC 0 (2002), the partial factor γ_dis applied and this value depends on the safety class of the structure. The partial factor for the different safety classes is shown in Table 3.7.

Table 3.6: Consequence classes (EC 0, 2002).

Consequences class Description

CC3 High consequence for loss of human life, or economic, social or environmental

consequences very great.

CC2 Medium consequence for loss of human life, economic, social or environmental

consequences considerable.

CC1 Low consequence for loss of human life, and economic, social or environmental

consequences small or negligible.

Table 3.7: Partial factors γ_daccording to safety class (EC 0, 2002).

Safety class Partial factor, γ_d

1 0.83

2 0.91

3 1.0

Dead weight

The dead weight stabilises the dam and therefore acts as a favourable, permanent load and is a structural action. For reinforced concrete with normal weight, the density γ_c= 24 kN/m³ should be used for calculations of the dead weight (EC 1, 2013).

(36)

Hydrostatic pressure

The horizontal water load (HW) acting on the upstream face, shown in Figure 3.5, is a permanent and unfavourable load while horizontal tailwater load (TW) on the downstream side and the vertical load (VW) are permanent and favourable loads.

The water loads are categorised as geotechnical actions. For analyses where the hydrostatic pressure is increased above the headwater level, the hydrostatic pressure can be classified as a variable load.

Figure 3.5: Hydrostatic pressure.

According to EC 1 (2013) the density for fresh water is set to γw = 10 kN/m³ and the loads caused by water should be determined with respect to the water level.

No combination factor is given for the water load; therefore, in this report, a value in the interval between the value for the highest snow load and the value for imposed loads on buildings was assumed. The combination factor ψ_0,w= 0.75 can therefore be used for variable water loads.

Uplift pressure

The vertical uplift pressure is considered as an unfavourable permanent load and geotechnical action. The density of water and the partial factor can be set according to the hydrostatic pressure. An additional horizontal uplift pressure will be present for monoliths without horizontal bottom surfaces. The horizontal uplift pressure can act as both an favourable and an unfavourable load, depending which direction the monolith is inclined. The horizontal uplift pressure resultant can also, depending on the location, vary between unfavourable and favourable for sliding and overturning.

Ice load

The ice load is an unfavourable variable load and is a structural action. In this report, the combination factor ψ_0,ice = 0.8 was chosen for the load combinations, based upon the highest value for snow loads.

(37)

3.2. EUROCODE

Rock anchors

Rock anchors are considered as permanent favourable loads and geotechnical actions.

The design strength of the reinforcing rock anchors may be calculated according to Equation (3.6).

f_yd = f_yk

γ_s (3.6)

where

f_yk is the characteristic strength of the reinforcement.

γ_s is the partial factor for the reinforcement.

The partial factor applied to untensioned and tensioned reinforcement bars is in this report assumed to be γs = 1.15 based on EC 2 (2011).

Earth pressure

Earth pressure is both a favourable and an unfavourable permanent load and is in this report considered as a geotechnical action. Eurocode states that the soil properties should be chosen from investigations or by theoretical or empirical correlation or from other relevant documentation. If standard values from tables are used, the characteristic values should be chosen with great care. According to the design of retaining wall structures, the determination of the earth pressure should be taken as at-rest pressure, if no movement of the wall relative the ground takes place. The lateral earth pressure coefficient, K_O is calculated according to Equation (3.7) for a horizontal backfill and according to Equation (3.8) for an inclined backfill and depend on the friction angle (EC 7, 2011).

Horizontal backfill:

K_O = (1 − sin ϕ⁰) ·√

OCR (3.7)

Inclined backfill:

K_O;β = K_O· (1 + sin β) (3.8)

where

ϕ⁰ is the effective friction angle.

OCR is the overconsolidation ratio.

β is the slope of the soil.

Values for the unit weight density and friction angle in Table 3.8, was obtained from Trafikverket (2011).

(38)

Table 3.8: Material properties for ground materials.

Material Unit weight density [ kN/m³] Friction angle [^◦] Un-saturated Saturated ϕ⁰

Rockfill 18 11 45

Gravel 19 12 37

Sand 18 10 35

Gravelly moraine 20 13 38

Sandy moraine 20 12 35

Silty moraine 20 11 33

Eurocode also states that the earth pressure should be calculated according to the chosen design approach, as shown in Equation (3.9). The design value of the earth pressure is:

X_d= X_k

γ_M (3.9)

where

X_k is the characteristic value of the material property.

γ_M is the partial factor of the material property.

The partial factor for material properties was chosen in accordance with design approach (DA3) to γ_M = 1.3. In most cases the earth pressure acts as a stabilising force, leading to a decreased force, resulting in a more conservative value for the earth pressure. If the earth pressure is active, the diagram in Figure 3.6 can be used to obtain the active earth pressure coefficient, K_a.

(39)

3.2. EUROCODE SS-EN 1997-1:2005 (Sv)

128

(3) Värdena på de effektiva jordtryckskoefficienterna kan hämtas från Figurerna C.1.1 till C.1.4 för Ka och C.2.1 till C.2.4 för Kp.

(4) Alternativt kan den analytiska metoden, som beskrivs i C.2, användas.

(5) I skiktade jordar bör koefficienterna K normalt endast bestämmas av skjuvhållfasthetsparametrarna på djupet z, oberoende av värdena på andra djup.

Figur C.1.1 – Aktiva, effektiva jordtryckskoefficienter Ka (den horisontella delen): stöttad horisontell markyta (E^{= 0)}

SIS fleranvändarlicens: SWECO Sverige AB. 2013-02-22

Uppdatering enligt EKS9 har gjorts av Anette Sjölund och Elizaveta Pronina. Senaste revidering 2015-04-16.

Tillägg och kommentarer i detta dokument har gjorts av Emma Persson. Teknikområde Grundläggning. Senaste revidering 2013-02-25.

Figure 3.6: Active earth pressure (EC 7, 2011).

3.2.2 Failure modes

EC 7 (2011) defines how to perform the design of geotechnical structures. The calculation model should describe the behaviour of the foundation and be reliable and give an error on the safe side.

It should be verified that ultimate limit state is not exceeded for:

• Internal failure or excessive deformation of the structure or structural elements, in which the strength of the structural material is significant in providing resistance (STR).

• Failure or excessive deformation of the ground, in which the strength of soil or rock is significant in providing resistance (GEO).

Overturning

The failure criterion presented in Equation (3.10) for ultimate limit state from EC 0 (2002), defines the safety against overturning as:

M_d,dst≤ M_d,stb+ T_d (3.10)

(40)

where

M_dst is the overturning moments.

M_stb is the stabilising moments.

T_d is the shearing resistance.

If the shearing resistance, T_d, is included, it should not have a considerable effect on the result.

Sliding

The failure criterion presented in Equation (3.11) for ultimate limit state according to EC 7 (2011), defines the safety against sliding as:

H_d≤ R_d+ R_p;d (3.11)

where

H_d is the design value of unfavourable horizontal forces.

R_p;d is the design value of favourable horizontal forces R_d is the design shear resistance.

The design shear resistance is calculated by Equation (3.12).

Rd= V_d⁰· tan δ_d

γ_M (3.12)

where

V_d⁰ is the design value of the effective vertical load.

tan δ_d is the design friction angle.

According to Eurocode, the friction angle tan δ_d, should be determined based on geotechnical investigations.

3.3 Limit turning

3.3.1 Crushing resistance

An important parameter in the limit turning calculations is the crushing resistance of the rock mass, R_cr. This is a better estimation of the resistance to shear loading compared to the shear strength parameters friction angle and cohesion. The crushing resistance of the rock mass should be obtained from geotechnical investigations and depends on the peak shear and normal stresses acting on the rock. When the

(41)

3.3. LIMIT TURNING

failure will occur (Fishman, 2007). From experiments performed by Fishman the relationship between the crushing resistance and uniaxial stress was obtained, shown in Equation (3.13).

Rcr = 1.47 · σc (3.13)

If no investigations are available, values from Table 3.9 may be used in the calculations (Fishman, 2009).

Table 3.9: Crushing resistance R_cr for different categories of rock mass (Fishman, 2009).

Category of rock Type of foundation Parameter Rcr (MPa) I Massive, large fragmental, laminated, platy, very

low and low jointed, unweathered rock

characterised by uniaxial compression strength in a sample σc> 50 [MPa]

20.0

II Medium jointed, inconsiderably weathered rock characterised by σc> 50M P a

10.0 III Intensively jointed rock with σc= 15 − 50M P a and

inconsiderably weathered and low jointed rock with σ_c= 5 − 15M pa

5.0

IV Semi-rock, platy, thin-platy, medium, high and very high jointed with σ_c= 5M P a

2.5

(42)

3.3.2 Failure criteria

The stability factor is the ratio between the sum of resisting moments and the sum of turning moments. Including the moment from the force of the peak crushing resistance S. The moments are calculated relative to the turning axis O, as sown in Figure 3.7.

Figure 3.7: Principles of limit turning, reproduced from Fishman (2007).

The position of O axis is determined as follows:

O = (a, d) = ( N

t · R_cr, [(h²+ 2 · a · e − a²)^1/2− h]) (3.14) The force of peak crushing resistance is defined in Equation (3.15).

S = (a²+ d²)^0.5· t · R_cr (3.15) The moment of the peak crushing resistance will be calculated about the O axis and added to the resisting moment:

M_p.c = S · b_cr· 0.5 (3.16)

Limit turning stability factor, relative to the O axis:

F_s = ΣM_r

ΣM_t (3.17)

(43)

3.3. LIMIT TURNING

where

a is the x-distance from the downstream toe B to the position of the turning axis O.

d is the y-distance from the downstream toe B to the position of the turning axis O.

N is the resultant of the vertical forces.

T is the resultant of horizontal forces.

t is the width of the structural section along a projected center-line or the thickness of the buttress.

R_cr is the crushing resistance of the rock.

h is the lever arm of the horizontal forces T relative the downstream toe B.

e is the lever arm of the vertical forces N relative the downstream toe B.

bcr is the length of crushing plane OB.

P Mr is the sum of resisting moments.

P M_t is the sum of turning moments.

(44)

(45)

Chapter 4 Failure modes of concrete dams

The aim of this chapter was to detect the factors that might cause concrete massive and buttress dams to fail. Especially to consider if other than the currently used design criteria could be relevant. Dams are designed against failure criteria based on sliding and overturning. By going back and studying failures it is possible to detect if the failure criteria are sufficient or if other failure modes may have to be accounted for in the design process.

This chapter includes a compilation of reported concrete dam failures across the world; how and what have caused them to fail. This study excludes China since the documentation there is incomplete. Many failures occurred decades ago, and therefore the documentation and important information regarding these failures might be inadequate.

Greater incidents of concrete dams were also included. Known dam failures without information about either the foundation or the failure cause were excluded.

4.1 Documentation of failures

The aim of the engineering industry today is to take responsibility of establishing a global collaboration as well as openness to share and increase the general knowledge of the industry by creating formalised channels such as registers and organisations.

The Committee on Dam Safety (CODS) in particular is working with this. However, it is difficult to obtain information about particular failures, especially in cases where failure took place long ago. Other reasons could be that some dam owners are not willing to admit failure and do not make the records public, or due to a legal policy preventing publication of records. This has slowed down the technical development of the industry, limiting possibilities of understanding earlier generations’ thoughts behind their solutions and designs. The sharing of information and the ability to talk about dam failures could help increase our knowledge of the field as well as provide the opportunity to learn from the experience of others, which would greatly increase the knowledge of the industry (Isander, 2013).

(46)

4.2 Compiled failures

The study includes failures of 19 concrete dams. These are divided into 12 massive dams, one gravity spillway dam and six buttress dams, presented in Table 4.1. The included failures have, to varying degrees, documented information about the failure and the dam. For a full compilation of the studied dams, see appendix A.

Table 4.1: Concrete massive and buttress dams included in the study.

Dam name Country Dam type

Height over lowest foundation

Year commissioned

Year of failure

Bayless¹ USA Massive 17 1909 1910(1911)

Camara² Brazil Massive 50 2002 2004

Eigiau³ GB Massive 10 1911 1925

Elwha river¹ (hydro-power)

USA Massive 51 1912 1912

High Falls⁶ USA Massive 9 1910 1999

Marquette no 3⁶ USA Massive 10 1924 2003

Shih-Kang dam⁵ (gravity spillway)

Taiwan Massive 22 1977 1999

St Francis¹ USA Massive 62 1926 1928

Torrejon-Tajo¹ Spain Massive 62 1967 1965

Upriver dam⁶ USA Massive 12 1937 1986

Warrensburg⁶ USA Massive 8 1909 1976

Xuriguera¹ Spain Massive 42 1902 1944

Zerbino⁴ Italy Massive 16 1925 1935

Ashley¹ USA Buttress 18 1908 1909

Cascade lake dam⁸ USA Buttress 5 1908 1982

Komoro¹ Japan Buttress 16 1927 1928

Morris Sheppard⁷ USA Buttress 58 1941 1986

Overholser¹ USA Buttress 17 1920 1923

Stony creek¹ USA Buttress 21 1913 1914

1(Douglas, 2002)

2(Shaffner and Scott, 2013)

3(J Andrew et al., 2011)

4(Luino et al., 2014)

5(Kung et al., 2001)

6(Reegan, 2015)

7

(47)

4.2. COMPILED FAILURES

4.2.1 Comparison of properties

In Section 4.2, Table 4.1 show the variation in year commissioned, age at failure and height, these properties are compared in Figure 4.1 and Figure 4.2. The majority of the studied dams were commissioned before 1940 according to Figure 4.1.

0 1 2 3 4 5 6 7

Massive dams Buttress dams

Figure 4.1: Year studied dams were commissioned.

0 2 4 6 8 10

During construction During first filling * During first five years After five years

Buttress dams Massive dams

Figure 4.2: Variation in age at failure of the analysed buttress and massive dams.

The buttress dams had according to Figure 4.2 slightly a higher tendency to fail during the first five years while the massive dams generally failed after five years.

The majority of the failures did however occur within the first years. Failure is less feasible for older dams, where the possibility of failures per year decrease with the age of the dam.

∗First filling is the first time the dam was filled

(48)

4.2.2 Failure type

Information about foundation material and geology of the foundation is presented in Table 4.2. In some cases information is missing, and this is denoted with the symbol "-". The failure of each dam is referred to a certain failure type, the used failure codes are defined below.

F f, failure due to dam foundation.

F b, failure due to the structural behaviour of the dam body.

F a, failure due to appurtenant works.

F m, failure due to dam materials.

Table 4.2: Failure types for massive dams.

Dam name Foundation material

Geology Failure code

Bayless Rock Sandstone horizontal layers with shale and clay between

Ff

Camara Rock Plane of micaceous silty clay Ff

Eigiau Clay Hard blue clay containing boulders of granite overlain by a layer of peat

Ff/Fm Elwha river Soil/rock Fluvioglacial and conglomerate Ff

High Falls Rock - Fm

Marquette no 3 Rock - Fa

Shih-Kang dam Rock Top deposition layer: unconsolidated gravel, sands, silts and clay. On Soft bedrock:

slate-gray, sandy-shale and silty-sandstones

Ffb

St francis Rock Conglomerate and schist Ff

Torrejon-Tajo - - Fa/Fm

Upriver dam Soil - Fa

Warrensburg - - Fa

Xuriguera Rock - Ff

Zerbino Rock Schist and hornfeld Faf

Ashley Soil Fluvioglacial Ff

Cascade lake dam Soil Glacial terminal-moraine sediments Ffa

Komoro Rock Tuff Ff

Morris Sheppard Rock Shale Ff

Overholser Rock - Ffa

Stony creek Soil - Ff

Evaluation of Failure Modes forConcrete Dams

Evaluation of Failure Modes for Concrete Dams

Evaluation of Failure Modes for Concrete Dams

Lisa Broberg & Malin Thorwid

June 2015

TRITA-BKN. Master Thesis 455, 2015 ISSN 1103-4297,

ISRN KTH/BKN/EX–455–SE

Abstract

Sammanfattning

Preface

Contents

Chapter 1 Introduction

1.1 Background

1.2 Aim of report

1.3 Limitations

1.4 Structure of report

Chapter 2

Concrete gravity dams

2.1 Gravity dams

2.1.1 Massive dams

2.1.2 Buttress dams

2.1.3 Gate section

2.1.4 Support methods

2.2 Stability analyses

2.2.1 Design loads

2.2.2 Failure modes

Chapter 3

Methods for stability analyses

3.1 RIDAS

3.1.1 Design loads

3.1.2 Failure modes

3.2 Eurocode

3.2.1 Design loads

3.2.2 Failure modes

3.3 Limit turning

3.3.1 Crushing resistance

3.3.2 Failure criteria

Chapter 4

Failure modes of concrete dams

4.1 Documentation of failures

4.2 Compiled failures

4.2.1 Comparison of properties

4.2.2 Failure type