Embankments are geo-structures that are mostly used to support infrastructures such as rails and roads. Therefore failures of embankments are result not only in loss of human lives but also affected the economic activities of a region. Rainfall has been identified as one of major causes to make embankments unstable and lead to subsequent failures. In fact, rain-induced embankment failures are categorized under the rain-induced landslides, the existing warning systems can not be implemented to minimize the risk successfully and economically. Therefore, in this study, a method for real-time prediction of rain-induced embankment was proposed. The method involves real-time calculation of embankment instability using numerically calculated pore-water pressures and unsaturated shear strength properties. It is necessary to have in-situ real-time pore-water pressure (or water content) measurement and rainfall data for numerical seepage analysis, in addition to laboratory-measured soil properties such as the soil-water characteristic curve (SWCC) and the hydraulic conductivity functions.

The proposed method was validated by employing to predict the failure of conducted model test. The necessary soil parameters for the analyses were obtained in the laboratory. The calculated FOS (factor of safety) was compared with the measured displacement in the model. The results suggested that the failure of the model test could be well-predicted by the proposed method. Finally, a procedure to back-calculate the SWCC from in-situ measured pore-water pressures and the results of the numerical seepage was formulated. The proposed procedure was used to back-calculate the SWCC of soil used in the model test.

As a part of this study, the SWCCs ・ and saturated/unsaturated permeability functions of test materials for different dry densities were obtained using the laboratory-designed Tempe pressure cell and permeability apparatus.　Based on the available results, following conclusions were made:

●　As the initial dry density of silty soil increases, the air entry value of the soil is higher and specimens de-saturate at a slower rate than the low-density specimens. Further, the high-density specimens have higher water contents than the low-density specimens at matric suctions beyond their air entry values. In addition, the hysteresis associated with drying and wetting becomes smaller, as density increases.

●　The SWCCs of the materials are affected by their particle distributions. Finer material has higher capacity of water retention than the coarser materials. It can be concluded that as the amount of fines in a material increases, the rate of de-saturation decreases. In addition, the specimens with high fines content have higher water content than the specimens with low fines content at the same matric suction.

●　The laboratory-designed permeability cell can be used to measure unsaturated permeability coefficient for low suction range (0 〜 8 kPa) in drying.

●　The method proposed by Fredlund (1994) and Green and Corey (1971) can predict the laboratory measured unsaturated permeability function reasonably well using the given SWCC and the saturated permeability coefficient.

The shear behavior of unsaturated silty soil for low suction range (0 〜50 kPa) was investigated by using the modified triaxial apparatus and the modified direct shear apparatus. Aiming to obtain unsaturated shear strength parameters for stability analysis using SLOPE/W (Geostudio, 2004), a series of shear tests was conducted on Edosaki and Isumi embankment soils under various combinations of net stress (confining or normal) and suction. The suction was achieved by either wetting or drying in order to investigate the effect of wetting-drying (the hysteresis of SWCC) on the shear behavior of unsaturated soils. Some tests were conducted on samples with different initial dry densities as well. The main conclusions of the investigation on unsaturated shear strength parameters are as follows:

●　The shear resistance and the initial shear stiffness increase with the increase in the net stress (net confining or net normal). The volume change of the specimen becomes more contractive as the net stress increases.

●　It was observed that soil subjected to a higher value of suction exhibits a stiffer response and higher stress-strain curve as compared to those of lower value of suction. Further, soil having a higher suction shows less compressive behavior.

●　The internal friction angle of tested materials is independent of suction and wetting-drying hysteresis of SWCC. It is apparent that the increase in the density may result in increase in the friction angle.

●　The remarkable non-linearity in the relation between the apparent cohesion and the suction was observed in the test results. The test data indicated that the apparent cohesion increases generally at a decreasing rate with the increase of suction.

●　Soil under wetting exhibited higher apparent cohesion than soil under drying at the same suction.

A series of embankment model tests subjected to rainfall was conducted in order to examine the effects of rainfall intensity, dry density of the slope, drainage condition at the slope bottom, and slope inclination on the failure of rainfall-induced embankment slope failures. Following conclusions can be drawn from the observed results.

●　Rain-induced failures of most slopes initiate near the toe upon full saturation of the lower part of the slope.

●　The slope failure induced by the high intensity rainfall is shallow and occurred at relatively short time after the start of rainfall as compared with the one induced by low intensity rainfall. Heavy slope surface erosion is involved in the failure induced by high intensity rainfall.

●　Denser slopes are more stable against rain-induced failures due to low permeability and high shear strength.

●　Improvement of the drainage condition beneath the toe of the slope is an effective countermeasure against rain-induced slope failures.

●　Slopes become more vulnerable to sudden collapse with the increase in their inclinations.

Parametric analyses were conducted to examine the effects of different input parameters such as SWCC, saturated permeability coefficient, initial pore-water pressure, and rainfall intensity on the numerical seepage and stability analyses. The results suggest that the effects of SWCCs (drying and wetting) and the initial pore-water pressure on the numerical seepage and stability analyses are not significant especially when the pore-water pressure increases greater than -3 kPa. However, there is a significant effect for low pore-water pressure values (at the beginning of rainfall). The effect of the saturated permeability on the stability analysis is significant. If the saturated permeability is smaller than the rainfall intensity, surface area of the slope becomes unstable within relatively short time of rainfall. Further, an increase in saturated permeability (much greater than the rainfall intensity) may increase the stability of the slope due to the improvement of drainage. With the increase in the rainfall intensity, the time required for the embankment to become unstable is shorter.

In the comparison of measured and simulated pore-water pressure time histories, it can be seen that the results agree well for high pore-water pressure (greater than -3 kPa). There is a significant difference between measured and simulated values for low pore-water pressure (at the start of rainfall). As the slopes become unstable due to high pore-water pressure, reasonably good agreement between the start of slope deformation and the drop of the FOS below unity could be observed. This result suggests that the proposed strategy may be used to predict the instability of rain-induced embankment failure.

In-situ water content measurement was conducted in a slope of Isumi-railway embankment at Otaki in Chiba. As explained in this chapter, the selected slope of the embankment was instrumented with 11 ECHO water content sensors. Further, a rain gauge was set on the slope to measure rainfall in the site and a thermometer was installed at about 30 cm below the surface of the slope to monitor the soil temperature. Data logging was done using HOBO weather station data logger which operates with batteries. Although the complete setup was on the site for more than 4 months, it was able to conduct data logging only for 1 month due to some technical problems.

During the time in which data was recorded successfully, the embankment slope did not experience any significant rainfall. However, clear responses were observed from some of the installed water content sensors during the recorded rainfall.

In order to examine the applicability of the proposed method of prediction of rain-induced embankment instability, numerical seepage analysis and the stability analysis were conducted for a certain period of recorded rainfall. Transient seepage analysis was performed assuming uniform material properties in the embankment even though the measured dry density varied with the height of the embankment. The results (pore-water pressures) of the seepage analysis and the laboratory measured shear strength parameters were used to assess the instability of the embankment.

When the measured and simulated water content time histories are compared, reasonably good agreements can be observed in the region with high-initial water content. However, there is a significant difference between the measured and simulated water content values in the region with low water content (near surface area).

The change in the stability of the embankment slope during the considered time frame is consistent with the rainfall and the water content measurements in the slope. The slope is well stable for the particular rainfall and FOS which was about 3.7 at the beginning of the rainfall event decreased by 0.35 at the end of the rainfall event.

　豪雨による斜面災害の対策手法として、アンカーや擁壁などの構造物を構築して斜面を補強する方法と、降雨量や地盤の状態を計器によって直接計測して斜面の状態を監視し、警報を発令し避難を促す方法がある。前者は、適切に施工されれば十分な効果が得られるがコストが高く、全国に無数にある不安定斜面に広く対策を施すには、後者のモニタリングの方法を開発することが適している。現在行われているモニタリングの問題点は、予測に用いられるパラメータが、比較的容易に得られる降雨量と過去の災害記録などに限られており、データ取得の困難な斜面ごとのきめ細かな地盤情報が考慮されない点である。本研究では、盛土法面にできるだけ少数の土壌水分センサーを設置し、常時計測した情報から斜面の水分特性を推定しておき、豪雨時にリアルタイムで斜面内の土壌水分の分布を推定して、斜面災害のリスクを評価する手法を提案した。

　本論文の第1章は序論であり、上記の問題を提起している。

　第2章〜第4章は、以後に行う斜面模型実験で用いる土を使って、室内での不飽和土の要素実験を行い、斜面を構成する土の土壌水分曲線(SWCC)と、様々な含水状態での不飽和透水係数および三軸圧縮変形強度特性を測定し、豪雨時の斜面内への水分の浸透、強度の低下、斜面災害のリスクの増加を定量的にシミュレーションするための基礎データを取得した。土壌水分曲線は、Fredlund らが提案した4つのパラメータを含む数式でうまく表現できることを確認し、本研究で提案するリスク評価の方法ではこの式を用いることにした。不飽和透水試験では、透水係数をサクションでなく体積含水率に対してプロットすることで、吸水・排水過程のヒステリシスを取り除くことができ、リスク評価のための数値解析を簡略化できることを示した。三軸圧縮試験では、有効応力についての内部摩擦角は水分量や吸水・排水の状態によらず一定なのに対し、サクションについての内部摩擦角と粘着力は、水分量に依存することを示した。実験技術上の知見として、三軸供試体の端部だけでなく中央部のサクションも計測し、透水経路が長いために、一定のサクションを与えてから1週間以上時間をかけなければ、供試体内の水分分布が一様にならないことを示した。

　第5章では、模型斜面の人工降雨実験を行い、斜面内の水分の挙動と斜面が崩壊に至るまでの変位を詳細に計測した。斜面への雨水の浸透、斜面地盤の強度の低下、斜面崩壊の発生の過程を、前述の室内試験で得られた材料特性のパラメータを用いて、数値的に解析し、実測値と比べることで、提案するリスク評価の方法の妥当性を確認した。

　第6章、第7章は、本研究で提案する斜面災害のリスク評価の方法を示し、前述の斜面模型実験の計測事例に適用して、その妥当性を確認した。この方法では、盛土法面内の土壌水分分布を把握するために、斜面内の3箇所にサクション計を設置する。設置作業時に、斜面内部の土を採取し、比較的低コストの室内試験によって、密度、間隙比(飽和時の体積含水比)、飽和時の透水係数、および強度特性を測定しておく。その後、サクションと降水量を、継続的に計測する。最初の数回の降雨時の計測データに対して、これを説明できるようにパラメータを調整することで、斜面の土壌水分特性を推定する。以後、豪雨時には、このパラメータを用いて土壌水分の挙動を推定し、斜面地盤の強度の低下を推定し、斜面安定解析で安全率を計算して、リアルタイムで斜面災害のリスクを評価する。模型斜面の人工降雨実験に、この方法を適用したところ、最初の降雨時の計測値に合わせて決定したパラメータを用いて、それ以降の降雨時の挙動や、同じ材料を使った別の模型の挙動を推定することができた。また、計算にかかる時間は、実際の現象の1/3程度になり、斜面安定計算時の安全率を考慮することにより、斜面崩壊の数十分前に警報を発することができた。

　一方、実物の鉄道盛土の法面に対しても、本研究で提案した方法の適用を試みた。降雨時の土壌水分の挙動を計測し、これを説明できるパラメータを推定しようとしたが、模型斜面の場合に比べて適切なパラメータを決めることが難しかった。これは、模型では斜面地盤や水分の初期分布が一様になるように管理されているのに対して、実盛土では地盤、水分分布ともに不均一性が強いからだと考えられる。また、現時点で用いている数値解析法は、斜面内で吸水と排水が同時に生じる状態を適切に考慮できない。降雨後に排水が始まり、斜面の安定性が回復する過程を推定して警報を解除するという使い方をするには、この点の改良も必要である。これらは、実用化に向けての今後の課題として、本研究で指摘された。

　以上、少数の土壌水分量の計測情報から斜面の水分特性を推定し、豪雨時の斜面災害のリスクをリアルタイムで評価して警報を発する方法を提案した。その妥当性を検討するために、系統的な室内での不飽和土の材料実験と、小型斜面模型の人工降雨実験、実物盛土の降雨時の水分特性の計測、および理論的検討と数値解析を行った。この手法の精度の向上、適用可能な斜面の範囲の拡大、実現場での検証など、実用化に向けての課題が残されているが、低コストできめ細かな斜面防災手法の基本的な考え方を提案しており、今後の技術開発に貢献することが大である。

　よって本論文は、博士(工学)の学位請求論文として合格と認められる。