Field Workshop on T, ~nds l i des , 1 9 8 5 , Tokyo A New Method for Predicting the Failure Time of a Slope Fukuzonu, T. Sationrl It~nearuh Ccntcr for Disaster Prrvcntion Ihnraki, Japan SmOPSI S A basic formula on the veloclty of surface displacement just before t h e f a l l ur e o f a slape is proposed and it is madc clear that the curves wrltten by each p o i n t of t h e inverse number of the v c l o c ~ t y of surface diaplacernent are divided into three k i n d s of forms, these a l e linear, convex a n d concave and decrease u n ~ i o r r n l y ~n any case. In t h e case o f a l i n e a r curve, t h e failure trme of a slope can be exactly predicted by the p o l n t at whi ch the straignt line cr osses the axls of absclssa. In the case of a convex or concave curve, the .allure tlme can be roughly pccd~ct ed by t he point at which the tangential line at each point of the curve crosscs the axls of absc~ssa. Furthermore, ~t 1s clear that the curve writtcn by each t l m e p r e d ~ c t e d roughly is linear. So t h e f a l l u r e time can be exactly predicted by t h e point a t whi ch t h e straight line crosses the axls of absclssa. The avallabilrty 01 these methods IS shown by a p p l i c a t i o n t o some types of slope failure. INTRODUCTION The new method for predicting t h e f a i l ur e t i m e of a slope proposed in thls paper is a method using t he variations of surface d~splace- ment just b e f m r e tltc failure. Saito and Uezawa ( 1 9 6 1 ) performed many creep tests of so11 and proposed t h a t t he increment O t h e loqarithm o f crecp r u p t u r e llfe (dl 1s proportional to the logarithm eE straln rate ( k ' ) in the secondary creep. That is is 1 / B. Hut t h e factor can be deci ded not befot e b u t after t h e f a t l u r e , because t he values of 5 are different at each slope failure. Tn t h ~ w papcr , a new method f o r predicting the failure time usi ng the inverse number of velocity of surface displacement 1s proposed. In the case of cquatlontz), the failure time can be exactly predicted by t h l s met hod. A l s o i n t h e case of equationtJ), tile fallure tlme can be ex ac t l y p r e d i c t e d b y us l ng more than two rough f a ~ l u r e time known at different time. . I l og[de) .= c-m.log(E)-------------------- (1 BASIC F'ORMJLA FOR VELOCITY OF SURFACE w h e r e c=2.33+0.59 and m=0.916 are constants. DISPLnCEMENT JUST BEFORE FAILURE A ~ ~ u m i n g that m equal s 1 a n d t; is the time to failure a t each time (t) in t h e st age of Thc variations of surface displacement in tertlary creep, the equatlon was appllcd to slope failure caused by rainfall was studied in prcdlct t h e f a l l us e time of a slopa ( Sa i t o , many k ~ n d s of exper i ment al models arp shown ~n 1 9 6 5 ) . That lo F i g . 1 by Fukuzono and Terashlma(l982) and log(L-k) = c-log(&)-------------------- I 2 1 40" f l at 40"mvex 30' flat 3o0con- where tl is t h e failure time and i s strain rate. Based on these formulas Ill and ( 2 1 , Saito proposed t w o ways wh i c h are r ough predlctasn methods, by the strain rate In the secondary creep range, for long term and exact predlctlon by calculation and graphical a n a l y s l s using t h e surface dlsplacement or s t r a i n lust before t h e f a i l u r e i n the t ert i ary creep range. B u t I n t h e t er t i ar y creep range, some slope failures which are not represented by t he equation(2) have been reported by Yarnaguchl(l978). In these cases, the proportional number o f t hr logarlthm of the time to fallure and t h e logarlthm of the s t r a i n rate IS not u n i t y as shown below: where A and B a r e const ant s. Based on thc cguation(31, Yamaquchi(l978) has proposed that the f a i l ur e time can be exactly predicted by multiplying the time obtained by Salto's met hod by the f a c t or of correctlcn, t h a t F i g . 1 Profiles o f four typs of experimental models. Fukuzono(l984). It has be e n made clear that the l o g a r i t h m of accelelatlon of s urf ace displacement increases in proportion to the loqarithm of vel o c ~ ty of surface displacement as shown in Fig.2 and 3 . The relatlon l a I where x is downward surface dlsplacement: d$/dt' is acceleration: dx/ dt is veloc~ty; a and 6 are const ant s. 1 (40 to 2 min. ' before failure) 3d FLAT.LCIAM ( 40 to 2 min. before failure ) a=2. 0 *. .:" / Pig.2 The relatlmship bet- the velocity and the acceleration of surface displacement in the fmal stage of slope f a l Pm on the 40' and 30' lm slapes. 40' Iun,mmxldi 40' SAND,pLAT (25min. to 30sec (2lhnin. to 30sec. 10' kf o m fail? 1 before f uur e) -. Fig.3 'Ihe relationship betm the vel aci t y and the acceleration of surface d~spla-t j us t before the fallure of the sand sl~pe with 40' mnvex and 40" flat bases. Equation ( 4 ) i s i nt e gr a t e d for t h e range of a)O as follows: I - 1- 4 a < l---e = 1- + (t+t )&----------- dt 15 -refore t h e relationship between t he v e l o c i t y of surface displacement and the time t o fallure j u s t before t h e f a i l u r e is generally represented by the equatlon ( 7 ) or ( 8 ) . A METHOD W R PREDICTTNG THE FAILURE TIME Equatlon ( 7 ) is transformed a s follows: where t,, t, and t, are constants of i n t e g r a l . d was i n a range 1.5 to 2.2 il the experiments. Also the values calculated from t he relationship between the ve l oc l t y of surface displacement and the time to failure i n ac t ual sl ape failure which were proposed by Yoshlda and Yachl ( 1 9 8 4 ) are 1. 70 to 2.13. Furthermore, the equation ( 7 1 is transformed as follows: r I dx l og (tp-t = A-B .log (-1 ---------------- (8) dt where 6=-log l a (d-11 I ; BLd-1 The form of the equation I & ) becomes similar to the equation ( 3 1 , if is substituted for dx/ dt . where l/v is l/(dx/dt), that is inverse number of v e l o c i t y of surface displacement. The equatlon shows t h a t t h e curve drawn for each p o i n t of t a nd l / v is l i ne a r i f 64=2,convex ~f d) 2 and concave if l ) d ) 2 and that t h e value u n ~ f o r m l y decreases. Ty p i c a l figures of t h e curve are illustrated i n Pig.4. In any case, the curve CainF near to tp As t h e velocity is flnite, The relationship is n o t repres ent ed by t he equation (9) at t h e near perlod of t,. But the period 1 s very short, so In t h ~ s paper ~t t r is assumed t o be t h e f a i l u r e time. me. ". Failure Ln L1 w > .. - I H t -* time: t Fig. 4 Typical figures of the changes of the Inverse n-r of velocity of surface d i s p l a mn t just before the falure. In the case of d = 2 , the failure time can be exactly predicted by a point t at which t h e straight l ~ n e of inverse number of veloclty crosses t he a x i s of abs ci s s a. Also using t he i nverse numbers of ve l oc i t y at two different times, the failure tlme can be calculated as follows; I n the case of d = 2 , t he failure t i m e pr e di c t e d by these methods equals t h e time pr e di c t e d by Saito's method In the range of error of caLculation and drawlng, because t h e basi c formula h a s the same form. In the case of ds.2. the f ai l ur e time can be roughly pr e di c t e d by t he poi nt at which the t angent line of the curve of inverse number of velocity at any time crosses the axis of abscissa. Also the failure time can be e x a c t l y predicted by the next method. he e qua t i o n ( 7 ) 1 s differentiated by t and rearranged as follow; Th i s equatlon shows t h a t the c u r v e drawn far each v a l ue of (l/vl/(dllJv)/dt) and t is linear and uniformly decreases. So the failure t r me can be exactly predicted by the time at whl ch the curve is expected to cross t h e axis of abscissa. Also using t h e i nvers e number of velocity and t he increment of oncs at two d i f f e r e n t tlmes t he failure time can be cal cul at ed by t h e next equat i on; Fur t he r mor e , a grayhlcal method bascd on the curve of i nverse numbex of veloclty is shown I n Fig.5. F i r s t , a tangent llne is drawn at any point Q, which is In a vertlcal d ~ r e c t i o n to the a x i s nf abs c i s s a from point TI, on t he curve. Tci is t h e polnt at which the tangent l l n e crosses the axi s of abscissa. A is plotted at the point whi ch is apart from the abscissa In a vert i cal di r e c t i o n for t h e same distance o f m<, In the s a m e way, P, is plotted. The abscissa t, of the point Tr at which a s t r ai ght l l n e connecting P, and P,crasses the a x l s of abscissa is t h e f a ~ l u r e t i m e . Cur ve of inverse number Time: t Fi g . 5 Graphical method f or predicting t h e time of slope failure i n the case of d f 2 . The theory is v e r i f i e d as f ol l ows ; (ma) /(mJ equals (m& / (-J in Fig. 5 .- m-m and T, 0-TI 0 are substituted for TI Tb and m, Th a t is - - - - - P I T, ' T, O - P J T x ' T ~ O ----- T + O = - - (13 PI TI - Pa Tr --- Al so, ~ i ; ~ , = ~ , / t a n ~ , , PA T p T t ~ T ~ Q, q / t a n 8 z --(I41 As t h e lines of Q, TtI and Q,Tczare t h e tangent Ilnes on the curve of inverse number o f veloclty at t h e point Q, and Q, (times are t I and t~): "('/v) I , , t a n - ~~~z- - - - ( l ~) t a n B, = - (r d t Al so, m~th m=4, m=tl, VP( ~/ V) , , mz=(l/~)I ---- ( I 6) The equations ( 1 4 1 , ( 1 5 ) and ( 16) are substituted f or the equation ( 1 3 ) and after the transformation, equat i on / 1 2 ) i s obtained. Another graphi cal method 1s s hown in Fig.6. In this method, polnts PI and P, are pl ot t e d in a vertical direction n o t at po l nt s T, and T2but a t points TC, and TI Ti m e : t F i g . 6 Another gr aphi c al method for predlctlnq the time of slope failure In t he case of 642. APPLfCATIOH FY)R PAST SLOPE FAILURE These proposed met hods were applied to some k i n d s of slope failure executed experi ment al l y In t h e pa s t to estzmate t h e availab~lity. An example whl ch was a ppl i e d to t he failure of the 4 0 " loam slope w l t h a flat bottom boundary parallel to t he surface is shown in Flg.7. T h e curve of inverse number o f velocity of surface displacement 1s convex in t h e period from 60 r nl nut es to 10 mi nutes before failure and l x n e a r from 1 0 ml nutes t o 2 0 seconds before Eallure. us i ng the tangent llne of t he curve the failure time can be predicted to wi t h i n an error of 1 mi n u t e in the period from 10 minutes to 20 seconds before failure. Al so a curve obtained by O-. Ry a method ' ' Tf f or t h e case 4 of d f 2 6 0 5 0 40 30 2 0 i o + Tl me to failure ( mi nut es) Failure F i g . 7 Var ~at i on of the Inverse number of v e l x l t y and surface displacement In the failure of the 40" loam slope with a flat bttm h d a r y parallel to the surface. a graphical me t h o d in t h e case of d+Z is shown in thc same f i q u r e . As O: 1 s not constant a l l the time, the c ur ve i s n o t l ~ n c a r . B u t the failure time predlcted by t h c curve is more exact than the time predl ct ed by t he crossing p o i n t of t he tangent line of t h e curve of inverse number o f vel oc3. t y and the axis of abscissa and t h e alarm for evacuation c an be gi ven more safely. Another example which was appl~ed to the f a l l u r e of the 30' loam sl ope with a flat bottom boundary parallel to t he surface is s h o wn in Fiq.8. In this case, t he curve 1s linear a t 30 mlnutes bef ore failure. So the faxlure time can be exactly predl ct ed from approximately 3 0 mi nut e s before allure by u s l n g o n l y the curve of inverse number o f ve l oc i t y. 2 30T40 30' FLAT. LOAH Inverse number of 2 0 " 20 1 1 1 I 1 50 40 30 20 lo Fa,,.!, Time te fai l ure [ mi n u t e s 1 fig. 8 Variatlm of the inverse n m k r of velocity and surface dsplacement In the failure of the 30' Iwm slope with, a f l a t h t t w n k md a r y parallel t.o the surface. Ot h e r exampl es which were applied to t he f ai i ur e of the sandy sol1 slope are shown i n Fig.9-11. Usuallyn the duration from the beginning of surf ace mo v e me n t to the final failure in sandy soil slope i s shart cr than t h e one i n loam slope and t h e p a t t e r n is vcry complex. Ne v e r t h l e s s t h e f a l l ur e t i m e can be sufficiently pr e dl c t e d by t h e curve of inverse number of veloc~ty, h f Inverse n m k r of LY f 2 + 2 0 2 $ Y 4 0 30 20 lo pailbre Time to failure ( mi n u t e s l Rg. 9 Variations of the i nve r s e nu*r of velcxrlty and surface aspl a-nt In t he f al l ur e of rhe 40" sandy sol1 slqw! with a flat h t t m boundary ~ r a l l e l to tth surface. 1. /': Inverse number of v e l o c i t y ! Tlme to failure ( m i n u t e s hFaiLufc Fig. lO Variatlms of t h ~ inver* .cudxr of velrxlty and surface d_lsplacemnt in the failurc of the 40' sandy sol1 slope wi t h a convex httan boundary. Inverw numbe r , w 0 \ 1 l D p 2 YE: ; o%s\ <4 % 8 0 "2 . + 4 0 20 'Flm to failure ( mi n u t e s 1 Fa1 lure Fig. ll Varlatlms of the lnversc nunJber of -rrelocit y and suf ace displacement i n -& failure of t k 30' sandy SOIL slope w i t h a flat bttm boundary pallel to the surface. APPLICATIONS UNDER FAILURE OF MODEL SLOPES Two slope models havlng t he ground surfaces of 4 0 and 3 0 degrees par al l el to the i mpervi ous bottom boundary w i t h mortar and both sldes bounded by vertical s t eel walls, were prepared t n estimate the availability of the proposed mcthod to predl ct the f a ~ l u r e time ag are ahawn UI Fiq.12. The sorl l ayers were made from t he coheslv~ sol1 of volcanic ash (called XANTO LOAM) having a liquid-limlt of 0 0 . 6 $, a pl a s t i c i t y i n d e x of 1 8 . 2 , a coheslan of 70 q/crn: an angl e o f internal frictlnn of 2 0 " u n d e r the dry density of 0-47 q/ cm3 and a w a t e r c o n t e n t equivalent, t o 100 8 of dry weight. The initial conditions and major results of the experiments a r e shown *n Table I. An over vlew of t he 30' slope model ~ u s t before the experim~nt 1s shown i n Photo. 1. Mater1al:KAATO LOAM Concrete foundatlo 30' LOAM, FLAT A 9 Materia1:KANTO LOAM Width:400, Uni+.:cm ---- F i n a l s l l d i n g ,q surf ace * waswing pi n& ' of surf ace 30' G. C. Pig.12 Fmfiles of s l p mxlels by loam of 40' and 30" in surface gradient w i t h a flat bottm boundary parallel kc t he surface. Ptrrto. 1 -,::e:- 3f 5P.e 3 2? ' ; ?a: slope v:c&l k f y r e t h e kyinnlna of e x p e r k n t . Artificial r a l n f a l l was s u p p l i e d continuously by a large scale rainfall s ~ m u l a t o r ln National Research C e n t e r f o r Disaster Prevention at an intensity of 2 D mm/h for the bot h slope model s. Surface displacement were measured by me a n s of extensometers having t h e accuracy of 0.1 mm at 5 rneasurlng points I n each model as are s hown In F i q . 1 2 . Ta b l e 1 I n l t l a l condrtions a nd maj or results of experiments. 4o0s l W 30"slope O r y density Ig/cm3) 0.49 0.49 f n l t i a l wat er cont ent ( % ) 100.1 99.8 Rainfall ~ntensity [mn/h) 20 20 Failure t m (mm.) 280.6 500.7 4 0 " SLOPE MODEL The 40' slope model undeswent small scale collapse 4 h o u r s 40 mlnutes 37 seconds a f t e r the b e g i n n i n g of r a i n f a l l . A final slldlng surface 1s drawn by a broken llne in Pig. 12. Three curves of the i n v e r s e numbe r of veloclty of surface displacement obtained under experiment at t he measuring point 2 , 3 and 4 ar e s hown I n Flg. 13. Th e inverse numbers. of veloclty a t each t i me w e r e c a l c u l a t e d by a dlgltal comput er from the d i g ~ t a l value lnto whlch a A-D conver t er changed the out put s of extensometers and were pl ot t e d continuously by a X-Y plotter. 13:OO 1 1 13:30 , 1 I L4.00 1 I 4 1 , ' I L +. Time ~ a l - u r e pig.13 Variations of the inverse n m h r of ~ l c e i t y under falPure of t k 40' loam slupe. collapse 8 h o u r s - 2 0 r n i n u t ~ s 41 seconds a f t ~ r the beqinnlng o f rainfall. A f i n a l sliding surface is drawn by a broken Line I n Fig. 12. Thr ee curvnR oE inverse number of velocity of s u r f a c e di ~pl acer nent nbtained under experi ment at t h e me a s u r ~ n g points 2 , 3 and 4 are shown In Fig. 14. Ths val ues of each t i m e w e r e obtained t h e same way as in t he case of 4 0 ' slope model. In this case, the curves were linear E r o m 30 minutes before failure. I n splte of t h e fact that t h e gradients of t h e t hree curves were different each othcr, a l l curves came toward a same p o i n t on the a x j s oP a bs c j s s a , that i s the failure t i m e . The variation of t h e lnverse number of velocity at measuring polnt 3 was A l l curves were convex, so t he failure time could not be predicted exactly, b u t roughly. I f one of the methods proposed in t h e case of 6 %2 was a p p l i c d , it mlght be posslble to predict the failure time e xa c t l y. 30- SLOPE WDRL The 30" slope model u n d e r we n t mi ddl e scale CONCLUSION fig.14 Varlations of the inverse n-r of velocity under farlure of the 30"loam slope. monitored wi t h t h e vi e w of f a i l u r e on a television. These are shown in Photograph 2 and 3 . Using t hese me t hods t h e exact failure time coul d be predicted continuously from approximately 30 minutes before fallure and t h e experiment was closed successfully, M . 2 V s w of X-Y plotter plotL ~ r q ~r>vr~r.-r-. n W r of velaclty under expermnt. f n this paper , a new method for predlctinq the failure tlme of a slope has been proposed. The major dl s t i nc t l v e f eat ure of t h e rnpthod i s that the i nv e r s e number of velocity and not the displacement or strain rate in s urf ace 1s used to predict the failure trme. The f ai l ur e t i m e c a n be easlly predicted by using a curve drawn continuously by t h e lnverse numbers of velocity of surface displacement at each tame. Also in cases that cannot be exactly pred~cted by Saito's method, t h e rough fallure trme can be e a s i l y predicted by usi ng the tangent llne of the curve and exactly predicted by a s t r a ~ g h t l i n e dr awn by the rough predicted time. It is hoped that the proposed method will apply in practical sl ope f a i l u r e . The experiments were performe8 by uslng a large scal e rainfall s l r nul at or i n National Research Center f or Disaster Prevention. H , Aoki , a t e c h n l c l a n assisted In t he operation of t h e simulator. Thl s assistance i s gratefully acknawledgod. REFERENCES Fukuzono, T. and Terashima, H. (1982) Experimental study of t he process of f a i l u r e ln cohesive soil slope caused by rainfall (in Japanese), the Report of the National Research Center for Disaster Pceventlan, 2 9 , pp. 103- 122. Fukuzono, T. ( 1984) A method f o r predicting the failure time of a sandy soil sl ope u s ~ n g t h e i nver se number of velocity [in Japanese), Proc. of 23rd Meeting of Japan Landslide Society, pp. 80-&1. Saito. M. and Ue z a wa , H . (1961) Failure of soil due to creep, Proc. of 5 t h I. C. S. M. F. E . , pp. 315-318 Sai t o, M. I 19651 Forcasting the time of occurrence of a slope failure, Prec. of 6th I . C. S. M. F. E . , pp. 5 3 7 - 5 4 1 . Yamaguchi , S . ( 1 9 7 8 ) Some notices of countermeasure for landslide and slope failure ( ~ n Japanese), Landslide Prevention and Slope Stability 2 , Soqo Doboku Laboratory, pp. 14- 24. Yoshida, T. and Yachi, M. (1984) On the velocity of landslide (in Japanese) , Proc. of 23rd Meeting of Japan Landslide Society, pp. 136- 139. m . 3 Yl cw of t . r l c v ~ ~ l o n showing inverse n m k r of wl ml t y untkr exprirrent.