Journal of Water Resource and Protection, 2011, 3, 370-379

doi:10.4236/jwarp.2011.36047 Published Online June 2011 (http://www.scirp.org/journal/jwarp)

A Dam Break Analysis Using HEC-RAS
Yi (Frank) Xiong
Department of Civil and Environmental Engineering, Mississippi State University, MS
Email: xiongyiwv@yahoo.com
Received March 16, 2011; revised March 25, 2011; accepted March 28, 2011

Abstract

This paper firstly described the dam break in the aspects of theories and models. Break parameters prediction,
the understanding of dam break mechanics, peak outflow prediction were shown as the essential for the dam
break analysis, and eventually determined the loss of the damages. Secondly, as an application example,
Foster Joseph Sayers Dam break was further modeled and analyzed using USACE Hydrologic Engineering
Center’s River Analysis System (HEC-RAS) model based on available geometry data. The results show that
dam break is a complicated and comprehensive process involving lots of principles. Combination of me-
chanics and case studies, reflection of predominant mechanisms of headcut erosion, more specific categori-
zation of dam, prudent investigation and inference of dam break process are needed in developing a satisfac-
tory dam break simulation model. Foster Joseph Sayers Dam break due to piping elongates the time period of
high water surface level, which increases the duration of risk. However, the dam break does not increase the
downstream maximum water surface elevation (Max. W.S. Elev) significantly at previous design Probable
Maximum Flood (PMF). Dam break has a greater impact on the downstream location where is closer to the
dam in accordance with the comparison of the hydrographs at different downstream locations. Sensitivity
analysis demonstrates that the changes of dam break parameters had no much influence on the downstream
Max. W.S. Elev.

Keywords: Dam Break, Analysis, Model, HEC-RAS, Hydrograph

1. Introduction due to slope slides on the upstream shells and liquificati-
on due to earthquakes [2]. USACE Hydrologic Engineer-
There is an old Chinese proverb - “The Great Levee is ing Center (HEC) Research Document 13 lists a list of
ruined by ant nest”. Generally, it warns people that small more complete causes as follows: 1) Earthquake; 2)
cases might result in big troubles. However, it always Landslide; 3) Extreme storm; 4) Piping; 5) Equipment
reminds water resource engineers the probability of dam malfunction; 6) Structure damage; 7) Foundation failure;
failure. 8) Sabotage [10]. Regardless of the reason, almost all
Dam provides many benefits for our society, but failures begin with a breach formation. Basically, breach
floods resulting from the failure of constructed dams is defined as the opening formed in the dam body that
have also produced some of the most devastating disaste- leads the dam to fail and this phenomenon causes the
rs of the last two centuries. Simulation of dam break eve- concentrated water behind the dam to propagate towards
nts and the resulting floods are crucial to characterizing downstream regions [2]. Despite the fact that the main
and reducing threats due to potential dam failures. Devel- modes of failure have been identified as piping or
opment of effective emergency action plans requires acc- overtopping, the actual failure mechanics are not well
urate prediction of inundation levels and the time of understood for either earthen or concrete dams [10].
flood wave arrival at a given location. Previous assumptions of instantaneous and complete
Dam failure results from both eternal force and intern- breaches simplify the modeling process. However, the
al erosion. Case studies show that dam failure may arise assumptions are not appropriate for most cases or only
due to different reasons ranging from seepage, piping applicable to a certain stage of the dam break. In the last
(internal erosion), overtopping due to insufficient spillw- two decades, more researchers have proposed headcutti-
ay capacity and insufficient free board and to settlement ng as the predominant mechanics of dam break for cohe-

Copyright © 2011 SciRes. JWARP

1985 Smart trapezoidal ailwater. Well understanding and modeling of the headcut overfall. Meyer-Peter Muller Regime type relation 1981 tailwater Meyer-Peter Muller. vegetation type and quality or riprap type and size. erosion processes. allowing studies data and/or the prediction equations are generally localized erosion that creates an overfall. timing (breach initial time. Model and Year Sediment Transport Breach Morphology Parameters Cristofano. dry slope tability Sediment. etc). In contrast. gression. All of Singh.12]. sediment FLOW SIM 1 and FLOW SIM 2 formula trapezoidal properties Copyright © 2011 SciRes. modified by Rectangular. etc) sudden collapse of the soil block. the reservoir outflow hydrograph and the routing of that U. Critical hear.6] with FLDWAV in the aspect of dam break a range of dam and reservoir configurations and erosion flood routing. ties. used to predict breach parameters. geometry and type of cular shape which improves stability of the headcut the dam can affect peak hydrograph estimation. 4) Comparative analysis. They concluded that the differences in the scenarios [13]. etc. 1981. the St. triangular. Embankment Breach Models [8. [1. As breach formation time. As material is eroded from this area. 1967.7]. Case the protective cover on an embankment fails. except Walder and O’Connor. SMPDBK. flow conditions. In overtopping failure. etc.12]. The key hydraulics. failure mode. Physically based dam break model uses principles of eled using tractive stress-based approaches [9]. material. In some cases. Breach dimensions. Critical shear stress. Venant Equations are similar [4]. saturated BEED. Ponce & Tsivoglou. others Harris & Wagner.14]. sediment Schoklitsch formula Parabolic breach shape 1977 properties Lou. issue. sediment. Gee and Brunner [4] compared USACE Hydrologic En. Many investigators have proposed simplified methods ods (Table 1 lists the major models summarized by for predicting peak outflow from a breached dam. case study data from real dam failures [13]. the mechanisms of breach development and other physic- support for the above soil mass is removed. 1965 Empirical formula Constant breach width Angle of repose. present at the downstream edge of the crest [1. Factors affect. Bureau of Reclamation [11] grouped the analysis hydrograph through the downstream valley [12]. Scarlatos and Wurbs). 3) these methods. 1985 Einstein-Brown formula Rectangular or trapezoidal slope tability Linear erosion or Schoklitsch Rectangular. The two primary tasks in the the base of the overfall and aeration of the nappe are key hydraulic analysis of a dam breach are the prediction of factors in headcut advance [12]. BRDAM. Breach dimensions. leading to cal processes (structure failure. are Predictor equations. sediment transport to simulate the develop- erosion zone once a headcut has formed is at the base of ment of the breach. magnitudes of the dam. (Frank) XIONG 371 sive embankments or rockfill embankments with a cohe. Headcut initiation takes place when in terms of locations.12].S. sediment. Headcut initiation can be mod. interpretation and approximation of river geometry are The dam break tool in HEC-RAS can simulate the Table 1.12]. data uncertainty results from limited case stud- slope. flow veloci. straightforward regression relations that predict peak Dam failure analysis models developed by National outflow as a function of various dam and/or reservoir Weather Service (NWS) such as DAMBRK. As for this prediction ing the initiation of headcutting include embankment method. vances upstream until the crest of the dam is breached Dynamic Breach geometry (breach depth and width. methods into four categories: 1) Physically based meth. JWARP . a small headcut is lations. breach pro- advance upstream it will widen and assume a semi-cir. and unit discharges. the primary source of the differences in the model simu- sive core [3. 2) Parametric models. triangular. ity analyses performed by previous researchers show that multiple stairstep headcuts form on the downstream face these parameters have varied influences on the peak flow of the dam [1. In some cases it may initiate at the knickpoint breach side slope factor). Tailwater conditions at are required in this case. with the relations developed from analysis of and FLDWAV are widely used as well as BREACH. Walder and O’Connor’s method is based upon an analy- gineering Center’s River Analysis System (HEC-RAS) sis of numerical simulations of idealized cases spanning model [5. ies and less detailed investigations. parameters. is an important cover discontinuities. Sensitiv- through arching of the soil mass [1. BREACH. Y. tailwater. flow concentrations. although the numerical algorithms for solution of typically formed near the toe of the dam and then ad.

both upstream and downstream boundary Susquehanna River Subbasin. “without Dam break”. Study Area analyzed because the spillway has adequate capacity to prevent the dam failure due to overtopping. The adjustment under PMF. on Bald Eagle Creek about ted full formulation time. The objective of this study is to le Maximum Flood (PMF) conditions: “without the dam”. This 100-foot high and 1. JWARP . Topo map of sayers dam and vicinity area.00 in the previous simulation as the water elevation can not reach Foster Joseph Sayers Dam Break Analysis was complet.372 Y. totally three scenarios were simulated for Probab- structure such as a levee.00 instead of 664. Figure 1. break with adjusted breach depth. dam is a unit of a comprehensive flood control project In this Dam Break Analysis. The detailed dam breach data are shown as HEC-RAS geometry data. understanding the Foster Joseph Sayers Dam Break mechanisms and impacts. (Frank) XIONG breach of an inline structure such as dam. Also. conditions (inflow hydrograph and rating curve.2. PA. 2.3 mile adjusted side slope factor under PMF. using mixed flow regime for the protection of communities in the West Branch simulation. 664. In addition. some dam breach parameters 2.1. or a lateral break. The break analyses of adjusted In order to quantify the incremental effects of a dam breach parameters are based on these basic data as well. only piping break mode was 2. for better geometry data. Figure 3 and Figure 4. Copyright © 2011 SciRes. break with mouth at Lock Haven. Foster Joseph Sayers Dam Break Simulation were adjusted to simulate the dam break and analyze the sensitivities. and break without long dam forms Foster Joseph Sayers Reservoir. Dam Break Simulation and Analysis using Assume the Piping failure Starting WS at the HEC-RAS Headwater Peak Stage of 660. apply HEC-RAS to a dam break analysis based on given “dam break”. Figure 2 represents the condition. Five scenari- Foster Joseph Sayers Dam (Figure 1) is located on Route os of break simulation were compared: break with adjus- 150 in Centre County.00 right behind the dam based on the given flow ed using HEC-RAS model. break with adjusted breach 1mile upstream from Blanchard and 14 miles above the width. respect- ively) and the gate opening height were identified.

(Frank) XIONG 373 Figure 2. Dam breach plot. JWARP . Copyright © 2011 SciRes. Y. Hec-ras simulation of sayers dam break on bald eagle creek. Figure 3.

“Without dam” has a much hydrograph dialogue.942). “dam break” and “without dam”. It looks that dam break elongates the time area in terms of dam failure. 1999 as shown in Figure 6.18). For PMF conditions. scenarios results again shows that no significant changes Before the time of Feb 22. and three scenarios are compared at four differ- drawdown of the reservoir in the case of “dam break”. 1999. the stage and flow are found as described above except for the increased hydrographs are overlapped for both “dam break” and duration of high stage and flow in the case of “without “no dam break”. period of higher water surface level. Blanch. while after that the “no dam break” flow break”. comparisons of the “without flow and stage are less significant than those of RS dam”. In this study. The inflow hydrograph is increased to 1. respectively) near the dam are four typical areas of con. “with break” and “without break” results and sen.41 effectively. Then. which might indicate the 22. The Sayers dam attenuates the peak flow at designed PMF by setting the multiplier as 1. two downstream key locations. Lock Haven) as shown in Table 2. both the stage the “dam break” and “no dam break” except at downstre- and flow of “no dam break” drop dramatically after Feb am of the dam (RS 81084. JWARP . and higher stage and flow for all three scenarios. and greatly weakens Obviously. while flood Figure 5 shows the stage and flow hydrographs at RS situation is much more severe at Lock Haven without 81849. 3.18.2 times the Apparently.374 Y. dam failure has less impact on the further downstream flood control. 81084. the scenario without dam poses much the flood protection capacity of the dam and at the greater risk on further downstream area (Blanchard and downstream area near the dam. It does not make any difference between At downstream of dam (RS 81084.41 (immediate upstream of the dam) for scenarios dam as shown in Figure 8.18). (Frank) XIONG Figure 4. and breach reaches its maximum depth. As shown in Figure 7. approximately 20 the two cross sections (RS 81849. compared with those of downstream area near the dam is the critical and sensitive “dam break”. sitivity analysis at these locations were used to ascertain The peak stage of “without dam” is much higher than RS the additional hazard caused by the breach.41 and RS 81084. 81084.96) and Lock Haven (RS 659. hours’ delay. the effects of dam break on the cern. a comparison of the three lower stage than both “no dam break” and “dam break”. the peak outflow occurs after the ard (RS 76002. Copyright © 2011 SciRes. As for “dam break”. It shows that there is a significant analysis. drops suddenly and the “dam break” stage is lower than Table 2 gives the incremental results of dam failure “no dam break” stage. Results Similar to Blanchard area. of “no dam break”.18 which is nearer the dam.18 and the advantage of dam in the flood control at Blanchard is coming out.2 in flow RS 81849. ent locations. Dam breach presumed progression.

Without Dam Flow .41 670 50000 Legend Stage .18 680 50000 Legend Stage .Dam Break 650 640 30000 Flow (cfs) Stage (ft) 630 20000 620 610 10000 600 590 0 18 19 20 21 22 23 24 Feb99 Time Figure 5.Without Dam Flow . Copyright © 2011 SciRes. Y.No Dam Break 660 40000 Flow . (Frank) XIONG 375 River: Bald Eagle Reach: Loc Hav RS: 81849.Dam Break 660 Stage .Dam Break Flow .Without Dam Flow .No Dam Break Stage .18 (Downstream of Dam). Stage and flow hydrographs at river station 81084. JWARP . Stage and flow hydrographs at river station 81849.No Dam Break 40000 Flow . River: Bald Eagle Reach: Loc Hav RS: 81084.Dam Break 640 30000 Flow (cfs) Stage (ft) 620 20000 600 10000 580 0 18 19 20 21 22 23 24 Feb99 Time Figure 6.No Dam Break Stage .41 (Upstream of Dam).Without Dam Stage .

JWARP .No Dam Break Stage .Without Dam Stage .Dam Break Stage .Without Dam 565 Flow .Dam Break 560 30000 555 Flow (cfs) Stage (ft) 550 20000 545 10000 540 535 0 18 19 20 21 22 23 24 Feb99 Time Figure 8. Stage and flow hydrographs at river station 76002.No Dam Break 598 40000 Flow .942 570 50000 Legend Stage .Dam Break Flow .Dam Break 596 594 30000 Flow (cfs) Stage (ft) 592 590 20000 588 586 10000 584 582 0 18 19 20 21 22 23 24 Feb99 Time Figure 7.96 (Blanchard). Copyright © 2011 SciRes.No Dam Break 40000 Flow .376 Y.Without Dam Flow .942 (Lock Haven).No Dam Break 600 Stage . River: Bald Eagle Reach: Loc Hav RS: 659.96 602 50000 Legend Stage . (Frank) XIONG River: Bald Eagle Reach: Loc Hav RS: 76002.Without Dam Flow . Stage and flow hydrographs at river station 659.

25 666.38 0. the Max. Only in the case of adjusted full formulati- controls the flood.53 6. W.25 666.S.27 590.56 Adjusted Breach Width Adjusted Breach Depth Adjusted Breach Side Slope Increase 20% Decrease 20% Decrease 20% Increase 20% Decrease 20% W.S. (Frank) XIONG 377 Table 2 also depicts that the construction of the dam sensitive to the adjustments of given breach parameters changes the velocity in the Bald Eagle Creek.41 666.01 feet higher than Table 3 gives the dam break sensitivity analysis 1 Max.S.68 2.41 666. Table 2.45 1.S. For Sayers Dam break analysis.27 590. but brings the problem of sediment on time at the upstream of the dam (RS 81849. There is no significant difference among water behind the dam might be the reason.04 Lock Haven 659. Incremental results of dam failure analysis.S. Max.45 1. “adjusted breach width”.56 555.29 3.45 1.25 0.27 590. W.27 590.S.56 555.00 612.36 feet) in –50%.11 Downstream of the Dam 81084.38 0.25 666. Elev Vel Chnl W.38 0.56 Copyright © 2011 SciRes.18 666.86 the Dam Downstream 81084. deposition and erosion as well.37 0. “adjusted breach depth” (±50%) and no significant change is observed either.25 666.S.36 1. JWARP . compared with Table 4.25 666. Max.27 590.37 0.38 0.18 666.45 590.11 666. Elev (666.25 6.45 590.96 590.25 Blan chard 76002.S.11 666.37 0.25 666.37 0. No Dam Break Dam Break Without Dam Cross Elev Elev Location W.38 0. W.29 of the Dam Blanchard 76002.53 6.10 10. Elev (666. Max.00 5.45 0. higher than Max. Elev Vel Chnl Vel Chnl Vel Chnl W. Elev Vel Chnl 666.S. Elev) and Velocity at each of “adjusted full formulation time” at the upstream of the identified location. formulation time delaying the release of impounded nt locations.45 1.942 555.56 555.11 0. dam.S.25 6.93 13.11 666. “adjusted full formulation Table 4 shows the dam break sensitivity analysis 2 time”.96 590. Elev Vel Chnl W.S.11 666.37 0.11 666.27 555.00 568.02 610.29 0. “break without adjustment”.37 0.942 555.25 6. Greater full (±20%).30 0.95 -56. and three scenarios are compared at four differe. W. W.11 666.38 feet) in +20% is 0.53 6.29 0.11 666. Elev of “break without adjustm. in addition to the slight change in the scenario Surface Elevation (Max.36 1. and “adjusted breach side slope” in Maximum Water However.25 590.53 6.11 666.37 0.37 feet) in –20%. Elev is not so given dam breach parameters. Elev is insensitive to the change of the Table 3 also shows that Max.56 555. Elev Vel Chnl W.53 6.56 0.S.05 -53. Elev Vel Chnl Upstream of the Dam 81849.30 0.01 feet However. Dam within ±20%.56 555. Adjusted Full Formulation Time Break without Adjustment Increase 20% Decrease 20% Location Cross Section W.37 0.56 555.S.36 1. W.S.S.S.S. W.56 555.S.30 0.45 1.27 0. No further new information is shown in Table 5 as It sounds that there exists a critical set of breach parame.37 feet) in +50% is 0.41).45 1.45 1. ters corresponding to a “minimum Max.38 0.53 6. W.27 Lock Haven 659.00 600. Elev Vel Chnl W.S.25 666.44 13. Elev (666. Elev”. ent” is a bit less than that of all other adjusted scenarios. Elev (666.32 Table 3. Y. Elev Vel Chnl W. Dam break sensitivity analysis 1 (±20%). Elev Vel Chnl Section Increase Increase Upstream of 81849. W. Elev Vel Chnl W.53 6.

45 1.25 666. Dam break is a complicated and comprehensive process which increases the duration of risk.36 0.27 590.25 Blan chard 76002. Conclusion on given geometry data.S.25 590.45 1.S.41 666.11 666.S.38 0.27 555. Dam break sensitivity analysis 3 (±75%). The dam break due to piping elongates the time period of high water surface level.38 0.45 1.45 1.37 0.25 666.53 6.S.53 6.45 1. Elev Vel Chnl W.27 590.45 1.30 0.56 555.27 590.11 666.29 0.38 0.25 666.53 6.11 666.56 Adjusted Breach Width Adjusted Breach Side Slope Increase 75% Decrease 75% Increase 75% Decrease 75% W.41 666.53 6.56 555.56 555. similar situations.45 1.53 6.25 6.25 666.11 Downstream of the Dam 81084.56 The comprehensive dam break analysis is a very and the actual failure mechanics are not well understood. However.25 666.27 590. the dam Copyright © 2011 SciRes. Adjusted Full Formulation Time Break without Adjustment Increase 75% Decrease 75% Location Cross Section W.37 0.11 Downstream of the Dam 81084.53 6.56 Table 4.53 6.36 1. (Frank) XIONG Table 4.37 0.942 555.25 590.56 555.45 590.11 666.11 666.56 555.942 555.45 1.53 6.37 0.45 1.56 555.S.27 590. The method models could fully explain dam break mechanisms and used in this study can be applied to other cases with impacts.38 0.25 666. Elev Vel Chnl W.25 666.29 0.30 0.53 6.18 666. The dam break tool in HEC-RAS was applied to Foster Joseph Sayers Dam break simulation and analysis based 4.37 0.56 555.27 555.25 Blan chard 76002.38 0.37 0.38 0.53 6.378 Y.25 666.37 0.45 1.53 6. Adjusted Full Formulation Time Break without Adjustment Increase 50% Decrease 50% Location Cross Section W.27 590.27 590.53 6. Elev Vel Chnl Upstream of the Dam 81849.53 6.37 0.96 590.11 666.45 1. Elev Vel Chnl W.56 555.45 1.S.38 0.38 0. Elev Vel Chnl 666.45 1.S.18 666.11 666. Elev Vel Chnl Vel Chnl Vel Chnl W.11 666.37 0.27 Lock Haven 659.37 0.37 0.S.S.11 666.S. Elev Vel Chnl W.S. important step in assessment and decision-making of the Neither current physical based models nor empirical damage following a dam break accident.56 555. Elev Vel Chnl 666.96 590. Elev Vel Chnl Upstream of the Dam 81849. Dam break sensitivity analysis 2 (±50%).45 590. Elev Vel Chnl W.37 0. Elev Vel Chnl W.S.25 6.27 590.36 1.37 0.25 666.S.56 Adjusted Breach Width Adjusted Breach Depth Adjusted Breach Side Slope Increase 50% Decrease 50% Decrease 30% Increase 50% Decrease 50% W.25 666.38 0.27 590.11 666. Elev Vel Chnl W.11 666.11 666. Elev Vel Chnl W.56 555.S.37 0.11 666.27 Lock Haven 659.25 666.38 0.38 0.56 555. Elev Vel Chnl W. JWARP .

EM 1110-2-1420.” Journal of Hydraulic Engineering. “Engineering and Design Breach Parameters– A Literature Review and Needs As.” U. pp. “HEC-RAS River Analysis [14] R. November 2006. JWARP . [9] D. E.” ACER Technical Memorandum Induced Flood Management.” Proceeding Dam Breach Parameters.1061/(ASCE)0733-9429(1987)113:1(29) [6] US Army Corps of Engineers.S. pp. Earth-Dam Failure. “HEC-RAS River Analysis System. [12] Dam Safety Office. August 1997.” Jour- System. Ralston. elevation. References cultural and Biological Engineers. “Dam Break Flood Rout. Hanson. Y. Applications Guide Version 3. A. Gee and G. Bureau of Water Resources Research Laboratory. pp.0 Beta. “Prediction of Embankment Dam [10] U. C. Engineering Center. Brunner. “Overtopping Flow on Low Embankment Parameters –A Literature Review and Needs Assessment. Denver. Water Resources Man.S. D.” American Society of Agri- 5.S. P. 1988. II: Hydraulic and Design Con- accordance with the comparison of the hydrographs at siderations. “Mechanics of Overflow downstream location where is nearer the dam in Erosion on Embankments. “Dam-Breach Flood Wave Models. In accordance with the sensitivity doi:10.1061/(ASCE)0733-9429(1989)115:8(1056) analysis. Vol. Bureau of Reclamation. different locations. Scarlatos. December 1988. 115. August 1988. pp.S. 389. M. Wurbs. Elev. P. 23-25 May 2007. 389-397. W. 130. Dodge. Vol. “The Role of Dam Safety in Dam-Break sification Guidelines. M. A. 29. 11.1061/(ASCE)0733-9429(1988)114:1(21) hydrograph and gate opening height. Dincergok. W. ver. “Analysis of Gradual much influence on the downstream Max. User’s Manual Version 4. surface elevation significantly at previous design PMF. pp. (Frank) XIONG 379 break does not increase the downstream maximum water gineering Center. 2004.” Dams – Summary Report of Model Tests. [13] T. Foster Joseph Sayer dam break has greater impact on the Clopper and T D. 8. July 1998. P. [1] Dam Safety Office. 113. 1.1061/(ASCE)0733-9429(2004)130:5(389) [5] US Army Corps of Engineers. L. 1989. Y. 10. on River Basin Management. 29-46. “Downstream Hazard Clas- [2] T. [7] G. –Hydrologic Engineering Requirements for Reservoirs. pp. Vol. 1056-1075. agement. which may result from starting trigger water surface Vol. Temple and G. boundary conditions including the inflow doi:10. Singh and P. “Headcut Development in Vegetated Earth Spillways. 5. 27-28. doi:10. emple. “Predicting Embankment Dam Breach [3] R. D. [11] U. San Francisco. 2005.” International Conference NO. pp. Army Corps of Engineers.” Journal of Hydraulic Engineering. Chen. Powledge. 1994. Den. No. M. the changes of dam break parameters had no [8] V. 21-42.” sessment.” Journal of Hydraulic Engi- of Environmental Engineering. 31 October 1997. No. R. “Uncertainty of Predictions of Embankment ing using HEC-RAS and NWS-FLDWAV. 1987. 683-691. J. Kos. Wahl. [4] D. Denver. 114. No. 677-682. No. World Water Congress. Miller.” Hydrologic En- Copyright © 2011 SciRes.” Water Resources Research Laboratory. neering.1. Reclamation. Vol. doi:10. November 2002. No. H.” Hydrologic nal of Hydraulic Engineering. Assistant Commissioner-Engineering and Research.