Optimum design of plates structures under random loadings

Projeto ótimo de estruturas de placas submetidas a vibrações aleatórias

CIVIL ENGINEERING ENGENHARIA CIVIL

Optimum design of plates structures under random loadings

Projeto ótimo de estruturas de placas submetidas a vibrações aleatórias

Elcio Cassimiro Alves^I; Luiz Eloy Vaz^II

^IFederal University of Espirito Santo, Vitória Brazil. [e-mail]

^IIFederal University Fluminense, Niterói, Rio de Janeiro, Brazil. [e-mail]

ABSTRACT

Structural optimization problems involving static loading have been studied for some time and in a way these kinds of problems are widely encountered in literature, but problems involving dynamic loading still have few studies not to mention problems involving nondeterministic dynamic loading. This paper presents a formulation of the optimization problem of plates subjected to random loadings. For the modeling of the structure, it used the bending plate element AST6 that explicitly provides these matrices, and dynamic mass reduction and rigidity matrices were used to reduce the computational cost of the problem. . The solution was achieved using the Interior Point Method and the sensitivity analysis of mass, and for the structure's stiffness matrixes the semi-analytical method was used. Three examples are presented to demonstrate the reliability of the process. The first example is an isotropic plate and the other two are problems involving sandwich plates. In all examples, an improved design in respect to the initially proposed geometry was obtained.

Keywords: Optimization, sensitivity, plates, random loading.

RESUMO

Problemas de otimização estrutural, envolvendo carregamento estático, já vêm sendo estudados há algum tempo e, de certa forma, esse tipo de problema já está bem definido na literatura especializada, porém problemas envolvendo carregamento dinâmico ainda são poucos estudados e problemas envolvendo carregamento dinâmico não determinístico menos ainda. O presente trabalho apresenta a formulação do problema de otimização de placas submetidas a carregamentos randômicos. Para a modelagem da estrutura, utilizou-se o elemento de placa à flexão AST6, que fornece essas matrizes explicitamente. Uma redução dinâmica das matrizes de massa e rigidez foi utilizada para reduzir o custo computacional do problema. A solução do problema foi obtida utilizando o Método dos Pontos Interiores e, para a análise de sensibilidade das matrizes de massa e rigidez da estrutura, foi utilizado o método semianalítico. Três exemplos são apresentados para demonstrar a confiabilidade do processo. O primeiro exemplo é de uma placa isotrópica e os dois outros são problemas envolvendo placas do tipo sanduíche. Em todos os exemplos, obteve-se um projeto melhorado, em relação à geometria inicialmente proposta.

Palavras-chave: Otimização, sensibilidade, placas, cargas aleatórias.

1. Introduction

Random dynamic loadings are only rigorously defined in statistical terms, consequently, the analysis of structural systems subjected to such loadings must be performed through statistical methods. They can be produced by natural phenomena such as wind and earthquake or by man induced phenomena like traffic or vibration of aerospace structures due, for example, to the launching of a rocket. Although the analysis of structures under random loading is well established, the correspondent optimisation problem has been only recently tackled. Some of the most important recent works in the area are quoted in the following lines. Kin and Wen (1990), present an analysis of the structural reliability taking into account the combined effect of several random loadings. Neubert (1993) studies the maximisation of structural damping in order to diminish the random vibration of the structure. Lipton (1994) considers the optimum distribution of plate stiffeners in plates submitted to multiple case of random loadings Alves at al. (2000) presents a thorough development of the equations for the analytical sensitivity analysis of the structural response of structures subjected to random loadings and validates the formulation comparing the results obtained with these equations with the ones calculated by means of the finite difference method. Alves at al. (2002) presents the complete formulation to the problem involves optimisation with random loading. Alves and Vaz (2010) presents application for truss structures. In this work, homogeneous and sandwich plates under random dynamic loadings are optimised. The AST6 finite element is used for the structural discretisation and the interior point algorithm by Herskovits (1995) is applied in the optimisation process.

2. The optimisation problem

The proposed optimisation problem consists of the minimisation of the structural mass (volum) of an homogenous or a sandwich plate subjected to the condition that the probability that the displacement and/or acceleration at a given point of the plate should not exceed, respectively, given bounds for the displacement and/or acceleration must be less or equal than a given bound for the probability. This problem is translated as:

In Eq.1, ρ_i, A_i and h_i are, respectively, the mass density, area and thickness of the i^th plate finite element; u_j and ü_j are, respectively, the vertical displacement and acceleration at the j^th nodal point, u_j,max and ü_j,max, respectively, the bounds for vertical displacement and acceleration at the j^th nodal point and P_d,max and P_a,max are given bounds for the probabilities (for example 1%, 2% or 5%) associated with the displacement and acceleration respectively. Also in Eq.1, Pd ( u_j > u_j,max) is the probability distribution function associated with the probability of the vertical displacement u_j be greater than the given bound u_j,max,

Where p (u_j) is the probability density function of u_j. A corresponding definition holds for Pa (ü_j > ü_j,max).

The probability density function adopted in this work is the Gaussian distribution:

The ergodicity and zero mean assumptions transform Eq.3 into.

In Eqs.3 and 4 σ_uj is the standard deviation of u_j. From the theory of the stochastic response of linear Multi Degree of Freedom systems σ_uj is obtained by the following expression:

Where B is the matrix whose terms are the n_j mode components and S_y (Ω) is the spectral density matrix of the generalised co-ordinates. This matrix is defined as:

In this equation S_p (Ω) is the spectral density matrix of the generalised loads and H(Ω) is a diagonal matrix whose generic term is the complex frequency response function of normal mode n given by:

Where K_n, M_n, Ω_n and ξ_n are, respectively, the generalised stiffness, mass, frequency of the exciting force and damping ratio of normal mode n. The design variables considered are the plate thickness in the isotropic plate and the sheet and core thickness in the case of a sandwich plate.

3. Sensitivity analysis

The gradients calculation to obtain the search direction is performed through the analytical method for the constraints. Alves (2000) presents a thorough development for the sensitivity analysis of the structural response due to random loading. The developed analytical equations are validated by a comparison with finite difference solutions. In the next section an example of the sensitivity analysis of a plate structure is presented.

4. Examples of optimization

Example 1. Isotropic Plate

In this example, the isotropic plate displayed in Figure 1 is optimised. The plate properties are the same as indicated in the previous example. The following constraint is imposed: the probability that the displacement in the centre of the plate, u_c, is less than 1mm, should be less than or equal to 1%. Therefore:

Two cases are considered regarding the number of variables:

1. Only one design variable (constant plate thickness) for all the mesh elements in Figure 2A.

2. Four design variables (four different plate thickness) as indicated in Figure 2B.

The results of the optimisation problem for the design variable values, objective functions and constraints are given in tables 1 and 2 for cases 1 and 2, respectively. The thickness configuration for the optimum design of case 2 is displayed in Figure 3.

Example 2. Sandwich Plate

The sensitivity analysis and the optimisation of a sandwich plate, Figure 4, are presented here. This type of structure is largely employed in satellite structures as they present high bending rigidity and low weight. The plate is modelled with a 4x4 mesh as indicated in Figure 1 and the random loading is the same as in Example 1. The material properties are given in Table 3. Two design constraints are considered associated with the displacement u_c and the acceleration ü_c at the centre of the plate as expressed by the following equations:

Results for the sensitivity analysis of the standard deviation of the displacement with respect to the design variables taking into account as design variables the thickness of the upper and lower sheets and the core thickness are given in Table 4. These results indicate the fair convergence of the sensitivity values calculated by Finite Difference Method (FDM) compared to the values calculated by Analytical Method (AM). Finally, these conclusions support the reliability of the optimisation method for sandwich plates.

The optimisation is initially performed considering the thickness of the sheets as the only design variable. The results are presented in Table 5. Then, both the thickness of the sheets and of the core of the sandwich plate are taken as design variables leading to the results of Table 6. It must be noted from Table 7 that the first constraint becomes active when the core thickness is a design variable.

Example 3. Sandwich plate with random loading with variable PSDF

The sandwich plate of the previous example is now submitted to a random load with the Power Spectral Density Function (PSDF) indicated in Figure 5 and applied at the centre of the plate.

Now the material of the sheets is constituted by orthotropic carbon fibres with their properties given in Table 7.

The imposed constraints for this problem are:

The first optimisation is considered with only the sheet thickness as design variables. The results are shown in Table 8.

Both the thickness of the sheets and of the core are considered as design variables in the second optimisation whose results are given in Table 9. It can be observed, as in the previous example, that the first constraint becomes active when both thicknesses are the design variables. On the other hand, the first optimisation leads to a better design. The reason for this is that the mass density of the core is smaller than that of the sheets, thus having negligible influence on the total mass.

5. Conclusions

The following conclusions are drawn from the analysis of the results of the examples:

A better design in relation to the initial design is obtained in all the analysed cases. In Example 1, with 4 design variables, the variable less altered is variable 1 (h). This is to be expected as the elements near the point of application of the load must be more rigid. This example is illustrative since the real plate has a constant thickness.

In the sandwich plate example the optimisation process is faster when two design variables are used instead of one. It must be noted that in this case there is an inversion relative to which constraint becomes active when compared to the first case with one design variable. The acceleration constraint tends to be active when one design variable is taken into account whereas the displacement constraint is active when two variables are considered. It must be noted that, in these examples, the best design is obtained when only the sheet thickness is taken as design variable. This is due to the fact that the material of the core is much lighter the sheet material. If the mass densities of the sheet and of the core are the same, the best design is obtained with both thicknesses as design variables.

6. Bibliography

Artigo recebido em 29 de maio de 2012

Aprovado em 16 de dezembro de 2012

• ALVES, E. C., VAZ, L. E. Otimização de estruturas reticuladas submetidas a vibrações aleatória. Revista de Engenharia Estudo, v. 10, p. 54-61, 2010.
• ALVES, E. C., VAZ, L. E., KATAOKA FILHO, M. Análise da sensibilidade e otimização de estruturas submetidas a vibrações aleatórias São José dos Campos, Brazil: National Institute of Space Research, 2002. (Doctor Thesis).
• ALVES, E. C., VAZ, L. E, KATAOKA FILHO, M. Análise da sensibilidade da resposta de estrutura submetida a carregamentos aleatório. XXI CILAMCE - Rio de Janeiro Brazil, 2000.
• GOTO, S. T., NETO, E. L., KATAOKA FILHO, M. Um elemento triangular para placas e cascas laminadas Brazil: Divisão de Mecânica Espacial e Controle (DMC/ETE), Instituto Nacional de Pesquisas Espaciais, 2000. (Proposta de Dissertação de Mestrado).
• HERSKOVITS, J. Advances in structural optimization Kluver Academic Publishers, 1995.
• LIPTON, R. Optimal design and relaxation for reinforced plates subject to random transverse loads. Journal of Probabilistic Engineering Mechanics, p. 167-177, 1994.
• NEUBERT, V. H. Optimization of location and amount of viscous damping to minimize random vibration. Journal of Acoustical Society of America, May, p.2707-2715, 1993.
• KIM, S. H., WEN, Y. K. Optimization of structures under stochastic loads. Structural Safety, v. 7, p. 177-190, 1990.
• SZE, K. Y., ZHU, D. AND CHEN, D. P. Quadratic triangular co plate bending element. International Journal of Numerical Methods in Engineering, v. 40, p.937-981, 1997.

Publication Dates

• Publication in this collection
  03 Apr 2013
• Date of issue
  Mar 2013

History

• Received
  29 May 2012
• Accepted
  16 Dec 2012