Source code for pylife.materialdata.woehler.analyzers.likelihood

# Copyright (c) 2019-2021 - for information on the respective copyright owner
# see the NOTICE file and/or the repository
# https://github.com/boschresearch/pylife
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

__author__ = "Mustapha Kassem"
__maintainer__ = "Johannes Mueller"

import numpy as np
from scipy import stats


[docs]class Likelihood: def __init__(self, fatigue_data): self._fd = fatigue_data
[docs] def likelihood_total(self, SD, TS, k, N_E, TN): """ Produces the likelihood functions that are needed to compute the parameters of the woehler curve. The likelihood functions are represented by probability and cummalative distribution functions. The likelihood function of a runout is 1-Li(fracture). The functions are added together, and the negative value is returned to the optimizer. Parameters ---------- SD: Endurnace limit start value to be optimzed, unless the user fixed it. TS: The scatter in load direction 1/TS to be optimzed, unless the user fixed it. k: The slope k_1 to be optimzed, unless the user fixed it. N_E: Load-cycle endurance start value to be optimzed, unless the user fixed it. TN: The scatter in load-cycle direction 1/TN to be optimzed, unless the user fixed it. fractures: The data that our log-likelihood function takes in. This data represents the fractured data. zone_inf: The data that our log-likelihood function takes in. This data is found in the infinite zone. load_cycle_limit: The dependent variable that our model requires, in order to seperate the fractures from the runouts. Returns ------- neg_sum_lolli : Sum of the log likelihoods. The negative value is taken since optimizers in statistical packages usually work by minimizing the result of a function. Performing the maximum likelihood estimate of a function is the same as minimizing the negative log likelihood of the function. """ return self.likelihood_finite(SD, k, N_E, TN) + self.likelihood_infinite(SD, TS)
[docs] def likelihood_finite(self, SD, k, N_E, TN): if not (SD > 0.0).all(): return -np.inf fractures = self._fd.fractures x = np.log10(fractures.cycles * ((fractures.load/SD)**(k))) mu = np.log10(N_E) sigma = np.log10(TN)/2.5631031311 log_likelihood = np.log(stats.norm.pdf(x, mu, abs(sigma))) return log_likelihood.sum()
[docs] def likelihood_infinite(self, SD, TS): """ Produces the likelihood functions that are needed to compute the endurance limit and the scatter in load direction. The likelihood functions are represented by a cummalative distribution function. The likelihood function of a runout is 1-Li(fracture). Parameters ---------- variables: The start values to be optimized. (Endurance limit SD, Scatter in load direction 1/TS) zone_inf: The data that our log-likelihood function takes in. This data is found in the infinite zone. load_cycle_limit: The dependent variable that our model requires, in order to seperate the fractures from the runouts. Returns ------- neg_sum_lolli : Sum of the log likelihoods. The negative value is taken since optimizers in statistical packages usually work by minimizing the result of a function. Performing the maximum likelihood estimate of a function is the same as minimizing the negative log likelihood of the function. """ infinite_zone = self._fd.infinite_zone std_log = np.log10(TS)/2.5631031311 t = np.logical_not(self._fd.infinite_zone.fracture).astype(np.float64) likelihood = stats.norm.cdf(np.log10(infinite_zone.load/SD), scale=abs(std_log)) non_log_likelihood = t+(1.-2.*t)*likelihood if non_log_likelihood.eq(0.0).any(): return -np.inf return np.log(non_log_likelihood).sum()