Identification of mine blast interaction pressure on excited plates

Identification of mine blast interaction pressure on excited plates: Using transient adjoint optimization for solving the inverse problems

van den Brink, Robbert-Jan (TU Delft Aerospace Engineering)

Turteltaub, Sergio (mentor)
Chen, B. (graduation committee)
Alderliesten, Rene (graduation committee)
de Jong, Bart (graduation committee)

Delft University of Technology

Aerospace Engineering

2019-10-09

Mines and improvised explosive devices are the cause of many mortalities of vehicle occupants. Both experimental and numerical research in this field aims to improve the safety of military vehicles. TNO has advanced Finite Element (FE) models to simulate such events. The numerical research is done to better understand mine blasts. Another aim is to reduce experimental costs for Defence Material Organization (DMO). For full vehicle simulations the empirical Westine model is the basis of the in-house TNO mine blast model. The model consists of a triangular pressure pulse consistent with the Westine impulse which is calibrated using the a test rig developed by TNO and DMO. Experiments are compared with numerical results using the TNO mine blast model and the jump height, representative for total impulse transferred, shows scattered results for different vehicle tests. A possible cause can be accuracy of the current mine blast model. This research will study a new technique to validate mine blast models. When this new technique results in improvements in validation of mine blast models it can be used to study and improve the current TNO model. The aim of the new validation method is to calculate the interaction pressure of an plate excited by a mine blast loading. The obtained pressure can be compared with existing mine blast models for validation of the mine blast model. The methodology is based on solving the inverse problem which requires transient Digital Image Correlation (DIC) measurements of the deforming plate for the duration of the mine blast loading. Such measurements are done with the state of the art test setup from TNO and DMO. Several methods to solve the inverse problem will be studied. The proposed best way to solve the problem at hand is by solving the corresponding optimization problem using an iterative gradient of descent algorithm. The main difficulty in this approach will be the calculation of the gradient of the objective function. To do this the corresponding transient adjoint problem has to be solved. First an algorithm to solve the inverse problem is implemented for a linear Kirchhoff Love plate model to study the behaviour of the inverse problem for a relatively simple test case. For this algorithm the transient adjoint problem is derived. The forward and adjoint problem are numerically solved. The implementation is verified using the method of manufactured solutions and a convergence study. The algorithm is verified using three different bench mark test pressures. It was found that without any exception the displacement of the algorithm converged with great accuracy to the applied displacement. The corresponding pressure converged good for smooth pressure distributions. Non smooth pressure distributions representative for mine blast interaction pressure did not converge. This shows the non-uniqueness of the solution of the inverse problem. It was realized after these tests that the forward problem acts as a low pass filter for time and spatial oscillations of the pressure distribution. This implies that the inverse problem in non-unique and that noise in the displacement data will be amplified in the obtained pressure from the inverse solution. The total force and radial position as function of time, obtained after integration over the spatial domain, for a localized load are accurately captured back. These parameters could be useful to validate a mine blast model such that research was continued for a continuum model. A non-linear elastic material model is proposed to model the deforming plate assuming monotonically increas- ing strain. This model is used and calibrated against the Johnson-Cook plasticity model. One plate simulation excited by the TNO mine blast model verifies that the non-linear elastic and Johnson-Cook material model result in almost the same displacement. The adjoint problem for a general continuum with elastic material model is derived. The same optimization algorithm is implemented for the continuum model using the calibrated non-linear elastic material model. The forward and adjoint problem are solved using the author’s transient FEM implementation which is verified using commercial software. From the benchmark tests it was verified that the displacement converged quite well however less compared to the linear problem studied earlier. The corresponding pressures behaved similar. It was concluded that the total force and centroid position of a localized load are not always in agreement for the benchmark and solved pressure. This research shows that the limitations of the inverse problem shed light on the limitations in the validation methods employed by many researchers in the field.

Mine blast
Inverse problem
Inverse analysis
Adjoint optimization
Adjoint problem
Transient Analyses
FEM
Kirchhoff Love plate
Non-linear
Non-linear FEM
Explicit

http://resolver.tudelft.nl/uuid:85e8f4ac-4e3c-485c-9b5d-f320e08fc5a7

Student theses

master thesis

© 2019 Robbert-Jan van den Brink