Smooth nonlinear hysteresis model for coupled biaxial soil-pipe interaction in sandy soils
2.4 Numerical verification
2.4.5 Uniaxial cyclic loading
The pipe was pushed cyclically along one direction that was inclined at an angle ππ’ relative to the π₯-axis (Fig. 2.11). In this case of loading, loading or unloading takes place concurrently along two directions. The pipe displacement was a 3-cycle sinu- soidal signal, π’ = π’πsin(2ππ‘/10), where π’π is the amplitude, π‘ is the time (0β30 s), ππ’ = {0, 30β¦, 45β¦, 60β¦, 90β¦}, andπ’π/π· = {0.1,0.3}. The soil experienced weak nonlin- earity whenπ’π/π· =0.1, whereas it experienced strong nonlinearity whenπ’π/π· =0.3.
Figs.2.14,2.15,2.16, and2.17present the results of BMBW, FEM, SPH, and the ASCE- recommended bilinear method. Overall, the FEM and SPH results agreed well with each other and jointly showed that the FDCs were indeed smooth, nonlinear curves. The ASCE bilinear method generated straight lines with slopes πΎ70. The proposed BMBW method
produced true smooth curves which adequately matched those of the numerical simulations.
Finch (1999) conducted a centrifuge experiment showing typical patterns of the FDC for vertical cyclic loading, as shown in Fig. 2.1. The loading, unloading, and reloading phases jointly create a hysteresis loop. The pinching phenomenon only appears during the unloading phase because the gap forms only below the pipe. The FDCs obtained from FEM and SPH in Figs.2.14,2.15,2.16, and2.17clearly depict the hysteresis loops and pinching phenomenon. In those figures, the BMBW method is capable of generating both the loops and pinching zones, in contrast to the ASCE-recommended bilinear method.
-5 0 5
u = 60Β°
This study FEM SPH ASCE
-10 -5 0 5 10
u = 45Β°
-10 -5 0 5 10
u = 30Β°
-10 -5 0 5 10
-6 -4 -2 0 2 4
u = 0Β°
Figure 2.14: πΉπ₯-π’π₯ for small pipe displacementπ’π/π· =0.1.
-10 -5 0 5 10
u = 90Β°
-10 -5 0 5 10
u = 60Β°
-10 -5 0 5 10
u = 45Β°
-5 0 5
-6 -4 -2 0 2 4
u = 30Β°
This study FEM SPH ASCE
Figure 2.15: πΉπ¦-π’π¦ for small pipe displacementπ’π/π· =0.1.
In general, the pinching is clearer along the π¦-axis than along theπ₯-axis. Specifically, the pinching length in theπΉπ¦-π’π¦curves is larger than that in theπΉπ₯-π’π₯curves (Figs.2.14,2.15, 2.16, and 2.17). The reason is obvious: the gap between pipe and soil causing pinching is due to gravity, which acts vertically. On the π₯-axis, when ππ’ = 0β¦, there is no pinching,
-20 -10 0 10 20 u = 60Β°
This study FEM SPH ASCE
-20 -10 0 10 20
u = 45Β°
-20 -10 0 10 20 u = 30Β°
-40 -20 0 20 40
-10 -5 0
5 u = 0Β°
Figure 2.16: πΉπ₯-π’π₯ for large pipe displacementπ’π/π· =0.3.
-20 0 20
u = 90Β°
-20 -10 0 10 20 u = 60Β°
-20 -10 0 10 20
u = 45Β°
-20 -10 0 10 20
-10 -8 -6 -4 -2 0 2 4
u = 30Β°
This study FEM SPH ASCE
Figure 2.17: πΉπ¦-π’π¦for large pipe displacementπ’π/π· =0.3.
as shown in Figs.2.14and2.16. Whenππ’ =30β¦, 45β¦, and 60β¦, the pinching occurs. The appearance of pinching is attributed to the effect ofπ’π¦, demonstrating the coupling effect between vertical and horizontal directions. For both πΉπ₯-π’π₯, πΉπ¦-π’π¦, the pinching length increases with displacementπ’πand angleππ’. The BMBW results captured this trend well.
Whereas the secant stiffness and ultimate reaction force in the ASCE uncoupled bilinear model remained unchanged, FEM and SPH results showed that they indeed change depend- ing onππ’. TakingπΉπ₯-π’π₯in Fig.2.16as an example, the ultimate reaction forces were about 8, 4, 3, and 2 kN/m whenππ’=0β¦, 30β¦, 45β¦, and 60β¦, respectively. The BMBW had similar results as the numerical simulations, indicating that it captures the coupling when the pipe moves obliquely.
Additionally, the coupling effect between two directions is illustrated by drawing the lateral uplift failure envelope, as in Fig. 2.18, in which πΉπ’π₯ π and πΉπ’ π¦ π are the ultimate reaction forces along the π₯- and π¦-axes corresponding to pipe movement angle ππ’. These forces are normalized by the ultimate reaction forces corresponding to purely lateral and uplift
pipe movements, πΉπ’π₯0 and πΉπ’ π¦90. When the inclination angle ππ’ increased, the ultimate normalized horizontal reaction forces decreased and the vertical reaction forces increased, indicating the coupling effect of pipe displacements in the lateral and vertical directions.
The envelope of the proposed BMBW method was close to that of the FEM simulations.
The analytical results ofNyman (1984), and the FEM results ofGuo (2005) andDaiyan (2013) also are plotted in Fig.2.18for comparison.
0 0.2 0.4 0.6 0.8 1
0 0.5 1
BMBW FEM
Nyman (1984) Guo (2005) Daiyan (2013)
Figure 2.18: Lateral uplift failure envelope.
2.4.6 0-Shaped loading
In 0-shaped loading, as shown in Fig.2.19(a), the loading phase occurs in one direction and the unloading phase occurs in another direction. Because the lateral displacement usually is dominant over the vertical displacement in earthquakes, two ratios of displacement amplitude were considered, namely π’π¦/π’π₯ = 0.25 and 0.5. The displacement amplitude along theπ₯-axis isπ’π, whereπ’π/π· =0.1 and 0.3.
-1 0 1
-0.5 0 0.5
0.5 0.25
-1 0 1
-0.5 0 0.5
Figure 2.19: Cyclic displacement loading patterns: (a) 0-shape loading; and (b) 8-shape loading.
Again, in Figs. 2.20 and 2.21, contrary to the oversimplified bilinear ASCE model, the BMBW produced smooth nonlinear curves and captured the dissipated energy via hysteresis loops. The areas of hysteresis loops were narrow forπ’π/π· =0.1 and fatter forπ’π/π· =0.3, indicating that more energy is dissipated for higher degree of soil nonlinearity.
-10 0 10
-4 -2 0 2 4
uy/ux = 0.25
-10 0 10
uy/ux = 0.5
-2 0 2
-3 -2 -1 0 1 2 3
uy/ux = 0.25
This study FEM SPH ASCE
-5 0 5
uy/ux = 0.5
Figure 2.20: πΉπ₯-π’π₯andπΉπ¦-π’π¦for 0-shape loading and small pipe displacementπ’π/π· =0.1.
-20 0 20
-10 -5 0
5 uy/ux = 0.25
-20 0 20
uy/ux = 0.5
-10 0 10
-6 -4 -2 0 2 4
uy/ux = 0.25
This study FEM SPH ASCE
-20 0 20
uy/ux = 0.5
Figure 2.21: πΉπ₯-π’π₯andπΉπ¦-π’π¦for 0-shape loading and large pipe displacementπ’π/π· =0.3.
When the pipe was pushed horizontally along branchπ π΄in Fig.2.19(a), the ASCE method predicted πΉπ¦ = 0, whereas the FEM, SPH, and BMBW results were non-zero values for πΉπ¦ (hysteresis loop branch π π΄ in Figs.2.20 and 2.21). The πΉπ¦ value of the BMBW was different from that of the FEM and SPH models because the simplified assumed function for π(ππ, ππ π’)does not exactly capture the variation of πin numerical simulations. However, the difference is acceptable considering the simplicity and computational efficiency that the BMBW offers while reproducing the trend ofπΉπ¦-π’π¦.
In Figs.2.20 and2.21, asπ’π¦/π’π₯ increases from 0.25 to 0.5, the corners ofπΉπ₯-π’π₯ become rounder, namely the effect ofπ’π¦onπ’π₯ becomes increasingly noticeable. In addition, when
π’π increases, the coupling effect is more pronounced. This is indicated by the rounder corners ofπΉπ₯-π’π₯andπΉπ¦-π’π¦ curves in Fig.2.21compared with the corresponding corners in Fig.2.20.
2.4.7 8-Shaped loading
In 8-shaped loading, as displayed in Fig. 2.19(b), both loading and unloading along the π¦-axis occur in either loading or unloading along the π₯-axis. Figs. 2.22 and 2.23 show results forπ’π/π· =0.1, 0.3 andπ’π¦/π’π₯ =0.25, 0.5. Similar to the case of 0-shaped loading, the hysteresis loops were thinner forπ’π/π· =0.1 and fatter forπ’π/π· =0.3. This reconfirms that energy dissipation increases with the degree of soil nonlinearity. Moreover, when π’π¦ increased, it increasingly affected πΉπ₯-π’π₯, creating a bulkier hysteresis loop. The FDC predicted by the BMBW model was in good agreement with the results of FEM and SPH.
-10 0 10
-4 -2 0 2 4 u
y/u x = 0.25
-10 0 10
uy/u x = 0.5
-2 0 2
-4 -3 -2 -1 0 1 2 3 u
y/u x = 0.25
This study FEM SPH ASCE
-5 0 5
uy/u x = 0.5
Figure 2.22: πΉπ₯-π’π₯andπΉπ¦-π’π¦for 8-shape loading and small pipe displacementπ’π/π· =0.1.
-20 0 20
-6 -4 -2 0 2 4 6
uy/ux = 0.25
-20 0 20
uy/ux = 0.5
-10 0 10
-8 -6 -4 -2 0 2 4
uy/ux = 0.25
This study FEM SPH ASCE
-20 0 20
uy/ux = 0.5
Figure 2.23: πΉπ₯-π’π₯andπΉπ¦-π’π¦for 8-shape loading and large pipe displacementπ’π/π· =0.3.
In Fig.2.23, there are two points at the corners ofπΉπ₯-π’π₯, and one point atπΉπ¦ =0 ofπΉπ¦-π’π¦
where the slope of FDC becomes discontinuous. This occurs because at those points, ππ¦ switches between positive and negative values, leading to changes in πΉπ’ π¦ and πΎπ¦, as in Eq. (2.14). Subsequently, this causes discontinuities in πΒ€π₯ and πΒ€π¦. This is inevitable when πΉπ’ π¦1, πΎπ¦1 and πΉπ’ π¦2, πΎπ¦2 are combined into one single spring. Nevertheless, the discontinuities vanish when πΉπ’ π¦1/πΎπ¦1 = πΉπ’ π¦2/πΎπ¦2. Furthermore, this happens only at isolated points on the FDC and does not affect the general trend ofπΉπ₯-π’π₯ andπΉπ¦-π’π¦. 2.4.8 Transient loading
The reaction force of the soil acting on the pipe was investigated for the case of biaxial transient loading. The 1995 Kobe earthquakeππ6.9 ground motions at OSAJ station, 8.5 km from the fault rupture, were used for this purpose (Ancheta et al.,2013). Fig.2.24shows the time histories and pattern of displacement. The lateral displacement was dominant, with a peak value of 88 mm, whereas the peak vertical displacement was 23 mm. The SPH method was chosen for the reference simulations to appropriately model large displacement of the pipe. Fig.2.25compares the response time historiesπΉπ₯-π‘ andπΉπ¦-π‘ predicted using the BMBW, SPH, and ASCE bilinear model. The proposed BMBW method captured the responses obtained by SPH to a high degree of accuracy. In contrast, the ASCE model tended to overestimate the πΉπ₯ in regions of large lateral displacement and ignored the coupling effect. Accordingly, πΉπ¦ was considerably different from the values predicted by SPH and BMBW. Fig. 2.26 shows the FDCs πΉπ₯-π’π₯ and πΉπ¦-π’π¦ estimated using the aforementioned three methods. Clearly, the shape of FDC from the BMBW matched well that from SPH, with round corners expressing the coupling between lateral and vertical directions.
0 10 20 30 40 50 60
-100 -50 0 50 100
ux
uy
-100 -50 0 50 100
-20 0 20
Figure 2.24: Kobe earthquake signal.
-10 -5 0 5 10
0 10 20 30 40 50 60
-15 -10 -5 0 5
This study SPH ASCE
Figure 2.25: πΉπ₯-π‘andπΉπ¦-π‘from BMBW, SPH, and ASCE model for Kobe earthquake.