In this section, we concentrate on the analysis of 3D structures with FEM. We try to focus on the technical considerations that when linked to Mpact recommended software, will result in high quality results.
CAD and Solid Models
Its now widely established that the CAD model is the best starting point for CAE. The commercial vendors such as AutoCad and Solid Works provide a general purpose to CAD. Yet the FEM software developers have made some effort to provide the user with a more focused , mesh generation friendly approach.
Mpact achieves this with the Mpave Mesh program which provides some automatic mesh generation that will be discussed later. The Mpave solid model is built via the solid boundary method. It has the ability to develop components. These components are then joined together by Boolean Operations. These have the options of maintaining the volume integrity of each component so that its meshes may use different material properties. In cases where special functions are required such as modelling micro lattices we have made use of the open source FreeCad program. FreeCad is Python based and all its features may be accessed by programming. Mpact has developed a program to convert finite element meshes that is then used by FreeCad to produce CAD models for 3D printing of micro lattices.
There are other methods to describe a solid. TrueGrid for example prefers surfaces of simple engineering solids with the aid of some effective intersection algorithms. With this wide range of CAD software, the result has been a multiple number of formats. We have found the TransMagic software useful for seamless integration between the CAD results.
Finite Element Meshes
Meshes are the gateway to numerical analysis of continuum mechanics. In FEA, the hexa element has been recognized as the most accurate element by far. Prof. Zienkiewicz, a pioneer of FEA, made the comment that a hexa element was worth a thousand tetra elements. For a long time practitioners were of the opinion that the choice of the element type was not important as long as we used enough of them.( because of the theorem of minimum potential energy). More recently, my colleagues and I found that the truncation errors of each element type are not necessarily negligible in the limit theoretically. These conclusions are also borne out by extensive numerical case studies [84-85]. The hexa 27-node element is a complete quadratic element. It was found to have the best performance of all the elements studied. This element also performed well in thin shell solutions so that we have concluded that there was no longer any need to study shell theory.Many of the commercial programs have modified their generation of element stiffnesses in order to obtain early convergence of shell like problems. Our work shows that this results in slow convergence for finer mesh and also larger errors in the limit[87-89].
Because of the importance of the hexa elements, we have worked extensively with Dr. Rainsberger, CEO of XYZ Scientific and author of the TrueGrid program.This is one of the few general purpose hexa mesh generation program and has demonstrated time and again that it can generate any large scale mesh. Recent updates has made it capable of generating a complete model that has the option of outputting the model for the following programs, viz. Abaqus (Simulia), Ansysy, LC-Dyna, Mpact and Nastran respectively.. In recent work, we have developed a Topological Domain Language that concentrated the modelling effort to constructive solid shapes. This restrictive geometry helps the beginning user understand the theory behind the True Grid program,