International Journal of Computational Engineering Science

A particular integral formulation is presented for the first time in a purely axisymmetric transient uncoupled thermoelastic analysis. The governing axisymmetric elastostatic equation is used as the complementary solution and the particular integrals for displacements and tractions are derived by integrating the three-dimensional formulation along the circumferential direction leading to elliptic integrals. The numerical results for three example problems are given and compared with their analytical solutions. Generally, the agreement among all of those results is satisfactory, if a few interior points are used to include the known values of temperature.

With the development of finite element method and computer technology, the complete modeling of the forming of a 3D sheet metal part using incremental approaches is becoming realistic. However, an accurate 3D simulation is usually too time-consuming to be used in the early design stage. One solution is to use simplified Inverse or One Step approaches for the sheet forming modeling. Unfortunately, these methods often based on implicit static algorithms, suffer from convergence problems because of strong non-linearities. In this paper we firstly present several initial guess methods to speed up the convergence of the implicit static solver used in the Inverse Approach, then we present some new techniques for the treatment of vertical walls (balancing, opening, matrix condensing,) which has proven to be essential when dealing with complicated parts. All these improvements have been introduced into our in-house F.E. software REFORM and applied to the drawing of industrial automobile parts. The obtained rapid convergence enables the easy determination of various process parameters for the design of sheet forming process. It is concluded that a good initial solution and the treatment of vertical walls play an important role to speed up the Inverse Approach modeling of autobody stamping processes.

It is difficult to identify some chemical processes which are processed in the complex environments and their operation conditions are frequently modified. In this paper we combine two effective identification tools, multiple models and neural networks and suggest a new identification approach. Hysteresis switch algorithm is applied, in order to select the best model of a set neuro identifiers at each time. The convergence the multiple neuro identifiers is proven. Simulation results show that the multiple neuro identifiers have better performance than those the individual neuro identifier for the pH neutralization and the fermentation process.

We have developed an improved solution of a maximum likelihood (ML)-based multitarget tracking problem by introducing a more accurate angular difference representation across the Riemann sheet into the Gauss-Newton's descend-direction relaxation routine obtaining a semi-optimal solution through shorter paths in the search space which eliminates the unnecessary relaxation calculations due to potential miss-assignments among the given and estimated measurements. We have proved that a new O(N²) 2-D data assignment algorithm based on the rotational sorting of angles maximizes the conditional likelihood function with respect to data association for fixed target states. The improved performance has been verified by implementing the improved tracking algorithm by a decentralized cooperative scheme ^1,2,3, one with the present angular formula and the other with the conventional formula used in Chen and Tokuda [2, 3] which is about twice slower even if it has a better O(N log N) complexity data assignment routine. The number of iterations needed has been reduced from an average of 39.2 to 17.3 iterations with the new formula, halving the computational time from 56 seconds to 32 seconds.

The stability problem has prevented the application of the BEM/FEM coupling procedure in dynamic fluid-structure interaction problems. The linear θ method, which was given by Yu et al. (1998), makes a significant stability improvement to the BEM scheme. With the Use of the linear θ method, the BEM/FEM coupling procedure is applied to 2-D time domain fluid-structure interaction problems. The fluid domain is acoustic and modeled by taking advantage of the BEM scheme suitable to either finite or infinite domains. The structure is modeled by finite elements that can be either linear or nonlinear. Three examples are analyzed to show the validity of the coupling procedure in fluid-structure interaction problems, and the significant stability improvement given by the linear θ method.