ANALYSIS ON DIKE BREAK BY SEEPAGE EROSION AND ITS TREATMENT

Wu Changyu  Ding Jinhua  Zhang Wei

Yangtze River Scientific Research Institute

Wuhan 430010

1    INTRODUCTION

In 1998, a seldom flood happened in Yangtze River basin of China. The river level higher than the emergency level lasted in 60~90 days in the area of the middle and down stretch of the river. Thousands of failures took place on the main river embankments by seepage or instability as sub-surface erosion. Among them the most dangerous result is dike break, which rushes dawn violently and often makes huge loses, since it usually happen in a very short moment that people there hardly to prepare for it or to leave in time, just like the situation of Jiujiang city dike in Jiangxi, Paizhouwan and mengxiyuan dikes in Hubei, and Anzhaoyuan dike in Hunan on this period. In order to prevent such disaster happen again, and to find out feasible and suitable treatment way while or after a danger, it is essential to make a thoroughly research on the erosion mechanism of such kind of dangers. So, the case of Jiujiang dike breaking which is the most serious one is studied in this paper.

2    BASIC INFORMATION AND ANALYSIS

2.1    Background of the dike

Jiujiang city dike broke on August 7, 1998. The breach was between the 4^th and 5^th gates, width of 62 m . About thirty million m³ of water flowed into the city and made the western part of it flooded. So it is the one of the most serious dangers in that year.

YRWRC organized a comprehensive investigation on the event in November, 1998, visiting the related unit which is near the stretch and responsible for its safety, talking to the witness of the disaster; gathering the data of regional topography and geology、design information of the dike and other structures around from the local water resource department; watching the video taken by Jiangxi TV Station recording nearly the whole process of the disaster. Meanwhile, supplementary survey and a series of seepage tests and soil property tests have been done. On the base of all these work, we get to know the basic conditions of the dike stretch and its breaking process generally, which can be summarized as follows:

(1) The breaking dike stretch was built in 1968, no serious piping happened before 1998 but some leakage phenomenon still could be seen at the toe of the inner slope.

(2) The cover layer of the ground outer the dike was excavated and clayey loam was altered by some rock blocks for constructing a dock in 1996. The depth to its bottom is 3.7m.

(3) The duration of river level over the emergency level was more than 45days before the breaking. The highest level is 23.03m (on August 2), and the breaking level is 22.87m (on August 7).

(4) There is no information about possible changes of the dike by seepage instability during the high level period.

(5) There is a pond near the inner slope about 2~20m long, which was the old Yongan river channel, and no record of the pond level. According to the topography and the fact the pond water should be pumped into the river, we can guess that the level is not lower than 18.2m or higher than 19m.

(6) The place where the dangerous sand carrying springs were first found is the toe of the inner slope. There were three springs with water column of 20cm high and loudly sound. Then the top of the dike slumped to a 2m big cave and became the breach in two hours.

(7) No termite or other harmful living things have been found.

(8) The site investigation and exploration shows that the cover layer of the beach is clayey siltation of 0.7~1m thick. The filling material of the dike is silty clay which is not uniform and weakly permeable. A discontinuous layer of silty loam is between the dike body and its foundation and is not uniform either. Below it is a silty clay layer nearly 12m thick. Then is the strongly permeable bottom layer of sand、gravel and rock, see Fig. 1,2,Table 1.

                  Table 1    Tests arrangement

(9) There is a masonry retaining wall on the inner slope. No design or construction information about it can be found. The material inside the wall is loose and its downstream face is sealed quite tight. There was no leakage on the face before the disaster starting. A thin layer of silty loam of 0~35cm was below the bottom of the wall to the toe of inner slope, and was just at the place where the springs were first found.

Fig.1    The profile up to the breach stretch

Fig.2    The profile down the breach stretch

2.2    Tests and results

A set of tests including soil property test, seepage stability test, soil erosion test and mineral and chemical analysis have been carried out by undisturbed samples from exploration troughs upper and lower the broken stretch to get the material characteristics of the dike and further understanding of the reason of the disaster. The test items and results see Table 2~5, Fig. 1,2.

             Table 2    Results of mechanical tests

            Table 3    Results of permeability tests

Results of soil property test indicate that the soil of the dike is with higher shear strength, the silty clay and silty loam are not uniform in permeability, even in the same dry density and from same place. Seepage test shows that the failure mode of both soil samples of the dike and foundation is the type of sand boiling. The critical and failure gradients of silty clay are much higher than ones of silty loam, or say the seepage strength of silty clay is much greater than the one of silty loam. Besides, by the test the content of organic and strongly soluble salt of the samples are not high, it means the materials are suitable to be used for dike construction.

                          Table 4    Results of seepage failure test

                      Table 5    Results of mineralogical and chemical analysis

Analyzing on the above information we can get primarily conclusions:

(1) Topography survey indicates that the dike section is placed on the concave bank with silting beach. Bank erosion by the Yangtze river flow in flood season is less possible. So the possibility that dike breaks by bank erosion can be ruled out.

(2) The parameters from mineralogical and chemical analysis and property test show that the filling material is adequate, and further more there is a retailing wall back the dike. It should be hard for the slope to be failure in ordinary condition as its high shear strength. So the dike break may not be caused by slope shear failure.

(3) Since of no sign of termite or other things, the possibility that any structural defect by living things makes dike breaking is impossible.

(4) Research tests show that the dike body soil is not uniform in permeability, of which the difference from two sample troughs is more than 10 times. The silty loam layer between the dike and the foundation is not uniform both in thickness and in permeability. It is more permeable and weaker in seepage strength than silty clay, and on the exit face of the dike. So it is a questionable layer in safety of the whole dike structure.

Then how is the dike broken affected by the seepage? How does every part of the dike , every link of seepage ring function? Only by using suitable mathematical model to make quantitative analysis, we can evaluate each related element appropriately.

3    SEEPAGE ANALYSIS

3.1    Calculation model

By the type and boundary conditions of the dike, we use two dimensional FEM to analogue the problem. As the high river level acted from June of 98, the duration was long , the groundwater flow pattern can be considered as steady and saturated state. The control equation is:

                           (1)

Boundary conditions are:

, for given head boundary               (2)

 for given quantity boundary           (3)

3.2    Determination of boundary conditions and parameters

Determining suitable boundary conditions and soil parameters are important to get reasonable calculation results. Based on information gathered, exploration and tests, analysis on parameter sensitivity and geography of river beech has carried out to imitate the conditions causing the dike breaking as the shortage of sufficient data of the dike stretch before its breaking.

(1) The seepage profile is obtained by survey, see Fig.2.1. The highest river level 23.03m and level 22.87m (dike breaking) are taken as the upper level. The others such as 22m and 21.38m appeared in the season are also used in calculation to study their influence. Besides, the landside pond is taken as the lower boundary level of 19.00m. Different geometry of river beach near the dike is considered to analogue the excavation of the place.

(2) The permeabilities of body soil, subsoil and the lower retailing wall as well as their different combination are compared in the range of that is given by the tests, see table 3.1, to find the reason and possibility of the break.

3.3    Safety standard in seepage stability

The safe gradient method is adapted as the safety standard in seepage stability. That is, when a groundwater flow gradient out lower soil slope J is equal to or less than the permission gradient J_c, under which the soil mass is stable, it is safe. Otherwise we think seepage deformation will happen. The formula is  J≤J_c,

Here, J=Δh/Δl , and

  J — gradient at the point of streamline out dike slope;

  J_c— permission gradient;

Δh —water head near the same point as above;

Δl —seepage pass between two points in calculation.

The worst soil layer—silty loam is the focal point in the research and the permission gradient is taken as the one in which it begins to move.

4    CALCULATING CASES, RESULTS AND ANALYSIS

At first tens of cases have been computed to narrow the research scope. Then the final cases are determined and are listed in Table 6. The results are showed in Table 7 and part in Fig.3 to Fig.4. The reason of choice of the cases and the results analyzing are described as follows:

Fig. 3    Equipotentials of case No. 1     Fig. 4    Equipotentials of case No. 11

(1) Influence of the lower part of the dike

In case 1 and 3 the permeability of the back part of retailing wall is taken as 10^-5and 10^-6cm/s respectively, as the lower face of the wall was tightly sealed. Both calculating gradient values exiting out of the dike are over the permission one. It shows that weakly permeable material such as its permeability less than the wall to10^-1 will be harmful to the dike.

(2) Influence of silty loam layer and its thickness

In case 4 and 8, possible changes of permeability of silty loam (2.62×10^-3 to 4.70×10^-4cm/s) by the tests are compared. Computed gradients at the toe of lower slope are greater than the permission gradient. Means seepage failure is easy to happen at such soil under all the conditions, and the permeability is lower, the safety of the dike is lower since its poor seepage strength.

According to the information by the exploring trough, the influence on the failure by the thickness of silty loam layer (4~23cm) under the wall is computed in case 2 to 5. The result shows the gradient values are 0.72 to 2.38, increasing with the thickness decreasing, most of the results are greater than the critical gradient and part of them even over the failure gradient. So the thin layer of silty loam under the wall especially with a exit face can be favorable to seepage failure.

(3) Influence of body soil and its compaction

The bigger value of permeability in tests is used to imitate the looser dike body of silty clay in calculation. The result of gradient is bigger then a tighter condition. But both are over the permission value. See case 6 and 7, that is the influence of poor compaction.

(4) Influence of river level

Comparisons are done with different river level to check their influence on dike breaking. The results of gradient are 1.48 and 1.54 in breaking and highest level, over the permission one. It shows the dike can not withstand the river level in 1998 with its structure and such inflow and outflow boundary condition. See case 4 and 6.

For other river levels the most of results are over the critical gradient but one is 0.76 near the critical when level is 22.38m. It indicates the dike stretch would be in danger when river level is over 21.5m or so and lasts long time.

 For other river levels the most of results are over the critical gradient but one is 0.76 near the critical when level is 22.38m. It indicates the dike stretch would be in danger when river level is over 21.5m or so and lasts long time.

               Table 6    Calculating cases of seepage analysis

note：lower level is taken as 19.00m, unit of k：cm/s

Natural condition before excavation of river beach is considered in case 9 and 10. The results are safer than the excavation condition. It shows excavation of beach and no recovery makes the condition be worse. In the other hand both results of gradient are greater than the permission. This is coincide with the investigation that there was leakage along the toe line before.

(6) Danger treatment and dike reinforcement

In order to get efficient way for danger treatment and similar dike reinforcement, a filter structure instead of masonry wall is put on the back of the dike, with k of 10^-3cm/s in case 11. The result of exit gradient is 0.69, means the seepage stability of the dike is basically ensured. Besides, special research in test proves that seepage strength of the material near lower part of dike can be increased over two times when it is protected by another material which is stable itself and fulfill the filtration principle. So it is possible to treat deformation problem in landside in or after flood season.

(7) The necessity of the dike break

The results of exit gradient are greater than the one soil can withstand in all cases described above, it proves that the reason of Juicing city dike break is caused by tests.

                  Table 7    Results of seepage calculation

5    CONCLUSION AND DISCUSSION

(1) Two-dimensional steady and saturated state AEM groundwater model is used in this paper to compute various cases of boundaries and parameters by exploration and tests. Results obtained shows seepage failure can happen on the lower part of the dike, which is consistent with the break record. It means the choice of computing model and parameters used is suitable and the result is reasonable.

(2) The weak permeability of the back part of retailing wall increases seepage force on to the exit face of the dike and decreases stability of the dike. The thin silty loam layer under the wall is the main passage for seepage failure starting and developing as its small thickness and poor strength. The excavation and no recovery in riverside worsen the situation. These are reasons insight the dike structure itself making the disaster. The outer reason is high river level in long period, here when river level higher than 21.5m in more than two months and combined with other conditions, seepage deformation can take place at downstream part of the dike till the whole dike failure without any treatment in time.

(3) It should be taken seriously note that the deformation in lower part of dike by seepage may cause local seepage failure. A right treatment is “block the upper part and drain the lower part of dike”. Moving the upper cover layer away or blocking the exit part of dike arbitrarily is inadequate. This is often ignored in reality, especially there is a misunderstanding of how to deal with the lower part of dike correctly. In fact as the research result shows, treating the deformed place in the landside by draining and filtering in time, if treatment in the riverside is difficult in flood season, it still could be possible to get efficient result.

(4) In reinforcement of a dike, it should be paid great attention to improve the condition of lower part of dike. The research indicates that even if under the worse conditions like in Jiujiang, 50% magnitude of gradient at toe can be reduced and unstable situation can be changed into stable in seepage, if altering the less permeable retailing wall to a permeable structure. This way is easier to be carried out and is cheaper than the way “to block the upper part”. So suitable material and type of the structure is worth to be concerned from now on.

(5) Some suppositions in the research such as geometry of river beach before and after excavation, parameters of retailing wall, are determined by inference. This works well in analyzing in right point generally but is more or less different from the reality. Besides, the original state of the breaking stretch is not yet imitated accurately and the related survey and test is still limited. All these need to be improved in similar research in the future.

References

[1]    The seepage stability and control of soils, Liu Jie, Water Conservancy and Eletric Power Press, 1992. (in Chinese)

[2]    The seepage analysis and control, Mao Changxi, Water Conservancy and Eletric Power Press, 1990.5. (in Chinese)