cpt_layering

cpt_layering(*args)

A function that uses the agglomerative clustering algorithm by Hudson et al. (2023) to automatically identify spatially contiguous soil layers with similar q_c1Ncs and I_c values.

Citation

Hudson, K. S., Ulmer, K., Zimmaro, P., Kramer, S. L., Stewart, J. P., and Brandenberg, S. J. (2023). "Unsupervised Machine Learning for Detecting Soil Layer Boundaries from Cone Penetration Test Data." Earthquake Engineering and Structural Dynamics, 52(11). link

Parameters

argument	format	description
qc1Ncs	Numpy array, dtype=float	overburden- and fines-corrected cone tip resistance
Ic	Numpy array, dtype=float	soil behavior type index
depth	Numpy array, dtype=float	depth to cpt measurement point
dGWT	float	depth to groundwater table
<num_layers>	integer	number of layers. optional. default = None.
<tknob>	float	constant used to define layer thickness parameter in cost function. optional. default = 0.5

Note

arguments in <brackets> are optional. If num_layers = None, the optimal number of layers is selected automatically based on a cost function that balances the distortion score with the desired layer thickness specified by the tknob parameter.

Returns

argument	format	description
ztop	Numpy array, dtype=float	depth to top of each layer
zbot	Numpy array, dtype=float	depth to bottom of each layer
qc1Ncs_lay	Numpy array, dtype=float	qc1Ncs values for each layer
Ic_lay	float	Ic values for each layer

Example

import numpy as np
import json
import pandas as pd
import requests
import io
import matplotlib.pyplot as plt
import ngl_tools.smt as smt

url = 'https://uclageo.com/coastal_database/conePenetrationTests'
headers = {'User-Agent':'XY', 'Accept':'application/json'}
r = requests.get(url,headers=headers)
cpt_df = pd.read_json(io.StringIO(r.text))
data = cpt_df['data'].values
dGWT = cpt_df['groundwater_measured'].values[0]
cpt_dict = json.loads(cpt_df.data.values[0])
depth = np.asarray(cpt_dict['depth'])
qc = np.asarray(cpt_dict['qc'])*1000
fs = np.asarray(cpt_dict['fs'])*1000
sigmav = 19 * depth
u = (depth - dGWT) * 9.81
u[u<0] = 0
sigmavp = sigmav - u
depth, qt_inv, fs_inv = smt.cpt_inverse_filter(qt=qc, z=depth, fs=fs, sigmav=sigmav, sigmavp=sigmavp)
Ic, Qtn, Fr = smt.get_Ic_Qtn_Fr(qc, fs, sigmav, sigmavp)
Ic_inv, Qtn_inv, Fr_inv = smt.get_Ic_Qtn_Fr(qt_inv, fs_inv, sigmav, sigmavp)
FC = smt.get_FC_from_Ic(Ic_inv, 0.0)
qc1N, qc1Ncs = smt.get_qc1N_qc1Ncs(qc, fs, sigmav, sigmavp, FC)
qc1N_inv, qc1Ncs_inv = smt.get_qc1N_qc1Ncs(qt_inv, fs_inv, sigmav, sigmavp, FC)
ztop, zbot, qc1Ncs_lay, Ic_lay = smt.cpt_layering(qc1Ncs_inv, Ic_inv, depth, dGWT)
fig, ax = plt.subplots(ncols=4, sharey='row')

ax[0].plot(qc, depth)
ax[0].plot(qt_inv, depth)
ax[1].plot(fs, depth)
ax[1].plot(fs_inv, depth)
ax[2].plot(Ic, depth)
ax[2].plot(Ic_inv, depth)
ax[3].plot(qc1Ncs, depth)
ax[3].plot(qc1Ncs_inv, depth)
ax[2].axvspan(1,1.31,alpha=0.3,color='darkorange')
ax[2].axvspan(1.31,2.05,alpha=0.3,color='peru')
ax[2].axvspan(2.05,2.6,alpha=0.3,color='palegreen')
ax[2].axvspan(2.6,2.95,alpha=0.3,color='mediumaquamarine')
ax[2].axvspan(2.95,3.6,alpha=0.3,color='navy')
ax[2].axvspan(3.6,4.0,alpha=0.3,color='saddlebrown')
ax[0].set_ylabel('depth [m]')
ax[0].set_xlabel(r'$q_c$ [kPa]')
ax[1].set_xlabel(r'$f_s$ [kPa]')
ax[2].set_xlabel(r'$I_c$ [-]')
ax[3].set_xlabel(r'$q_{c1Ncs}$ [-]')
for a in ax:
    a.grid(True, alpha=0.5)
ax[0].invert_yaxis()
ax[2].hlines(ztop,1,4, color='black', linewidth=0.5)
ax[2].hlines(zbot,1,4, color='black', linewidth=0.5)
ax[3].hlines(ztop,0,400, color='black', linewidth=0.5)
ax[3].hlines(zbot,0,400, color='black', linewidth=0.5)
ax[2].set_xlim(1,4)