A combined neural network/gradient‐based approach for the identification of constitutive model parameters using self‐boring pressuremeter tests

Abstract

SUMMARYThis paper presents a numerical procedure of material parameter identification for the coupled hydromechanical boundary value problem (BVP) of the self-boring pressuremeter test (SBPT) in clay. First, the neural network (NN) technique is applied to obtain an initial estimate of model parameters, taking into account the possible drainage conditions during the expansion test. This technique is used to avoid potential pitfalls related to the conventional gradient-based optimization techniques, considered here as a corrector that improves predicted parameters. Parameter identification based on measurements obtained through the pressuremeter expansion test and two types of holding tests is illustrated on the Modified Cam clay model. NNs are trained using a set of test samples, which are generated by means of finite element simulations of SBPT. The measurements obtained through expansion and consolidation tests are normalized so that NN predictors operate independently of the testing depth. Examples of parameter determination are demonstrated on both numerical and field data. The efficiency of the combined parameter identification in terms of accuracy, effectiveness and computational effort is also discussed.

Some aspects of application of artificial neural network for numerical modeling in civil engineering

“…A broad literature of this problem is known, we suggest only few papers, the further references can be found there: [24,[32][33][34][35][36]. In the solution of inverse problem, not only numerical robustness is important but also the fact that the suitably trained ANN can be used as an inverse relation between observable data and parameters of the model.…”

“…Important for geotechnics, general conclusions related to the ANN training with the aid of numerical database have been drawn. The straightforward continuation of the method developed in the mentioned above papers was identification of constitutive model parameters using selfboring pressuremeter tests [34]. This idea of identification of soil properties, basing on geotechnical laboratory tests (as well as on the above described "in situ" tests) is very important and should become the leading method in the field of numerical modeling, in author's opinion.…”

“…This is described in [33] and in many text books concerning this method of in situ testing. Another example -correlations between the data measured on the lateral, cylindrical and conical surfaces of the tip of the probe in various CPT (pressure and friction) and the properties that allow to classify the soil and to predict elements of its classical mechanical description [34]. The most promising method of data elaboration in the case of CPT is the prediction of the dimension of the pail for given bearing capacity, inserted in the soil the reaction on the CPT is known.…”

Numerical modeling and neural networks to identify model parameters from piezocone tests: II. Multi‐parameter identification from piezocone data

“…Coefficient of earth pressure ‘at rest’ : Random values of K 0 and K 0 nc were obtained likewise for the pressuremeter test presented in 5, with the exception of normal pdf assumed for random dispersals β 1 and β 2 and a uniformly sampled interval for OCR∈〈1;4〉 owing to the applicability of MCC to lightly consolidated clays (Figure 4(f)). The distribution of K 0 nc as the function of M and β 1 results in normal‐like pdf (Figure 4(g)), whereas the values of K 0 as the function of OCR, K 0 nc and β 2 result in log‐normal‐like pdf (Figure 4(h)).…”

“…Overconsolidation ratio : The values of R p were obtained using the following expression (cf. 5): resulting in normal‐like pdf between 1 and 4.6 (Figure 4(i)).…”