The Advanced Finite Element Modelling

(MATH 2034 (MA029))

Topic 9: Solution of Non-linear FE Models Thermal and Thermal-Stress Analyses

Solution of Non-linear FE Models

The cause of nonlinearity:
The FE solution to a time-independent problem always involves of solving a set of simultaneous algebraic equations of the following form:

[ K ]{a} = {F }
In linear analysis, both [K] and {F} are regarded as independent of {a}. Whereas in nonlinear analysis, [K] and/or {F} are regarded as functions of {a}. The followings are two examples of nonlinearity.
MATH 2034 Lecture Notes 10-2 Dr. Yan DING Dept. of Maths & Stats. RMIT University

Solution of Non-linear FE Models

Material nonlinearity:
Stiffness matrix is composed of a constant term [Ko] and a term [KN] that depends on deformation. Thus: ([Ko]+[KN]){a}={F} where: [KN] = f ({a}), i.e. depends on deformation of the structure.

Geometric nonlinearity:
Consider the plane cantilever beam, we seek the quasistatic deflection produced by loads P and ML. Assuming that the beam is slender and that its material is linearly elastic at all times. For small deflections, linear theory is adequate, and the root moment is: Mo =P LT + ML. For large deflections, the moment arm H of force P is less that LT, thus: Mo =P H + ML where H depends on P and ML.

MATH 2034 Lecture Notes

10-3

Dr. Yan DING Dept. of Maths & Stats. RMIT University

Solution of Non-linear FE Models

In the solution of non-linear problems, we will always obtain a set of algebraic equations: (a) = F - P(a)
where a is the set of discretization parameters, F is a vector that is independent of the parameters, and P is a vector dependent on the parameters, a. These equations may have multiple solutions, as shown below. Thus, if a solution is achieved it may not necessarily the solution sought. In order to obtain realistic answers, small-step increment approaches from known solutions are essential.

MATH 2034 Lecture Notes

10-4

Dr. Yan DING Dept. of Maths & Stats. RMIT University

Solution of Non-linear FE Models

The general nonlinear problem should always be formulated as the solution of: n+1= (a n+1) = F n+1 - P(a n+1) = 0 which stars from a nearby solution at: a = an, n= 0, F= Fn (2) and often arises from changes in forcing function Fn to Fn+1 = Fn + Fn the determination of the change such that an+1 = an + an will be the objective and generally the increments of Fn will be kept reasonably small in order to follow the path dependence. The solution of the problem above can not be solved directly and will always require some sort of iteration (repeated solution of linear equations). There are many iterative techniques for solving non-linear problems. Among them, the Newton-Raphson method is the most rapidly convergent process for solutions of non-linear problems.
MATH 2034 Lecture Notes 10-5 Dr. Yan DING Dept. of Maths & Stats. RMIT University

(1)

Solutions of Nonlinear FE Models

The Newton-Raphson Method

In the Newton-Raphson iterative method, to the first order, Eq(1) can be approximated as: i
i i i (an+1 ) (an +1 ) + dan = 0 +1 a n +1
(3)

Here the superscript i indicates the iteration number and usually starts by assuming a1n+1 = an in which an is a converged solution at a previous load level or time step. The Jacobian matrix corresponding to a tangent direction is given by P
KT = a = a

Eq(3) gives immediately the iterative correction as

i i i i K T dan = ni +1 or dan = ( KT ) 1 ni +1

MATH 2034 Lecture Notes

10-6

Dr. Yan DING Dept. of Maths & Stats. RMIT University

Solutions of Nonlinear FE Models

The Newton-Raphson Method

A series of successive approximations gives
i i i a n+11 = a n +1 + da n +1 + i = an + an

where

k a = dan i n k =1

The process illustrated in the figure below shows the vary rapid convergence:

MATH 2034 Lecture Notes

10-7

Dr. Yan DING Dept. of Maths & Stats. RMIT University

Thermal & Thermal-Stress Analyses

The following procedures are discussed:
How to do a thermal analysis; How to apply thermal loads in a stress analysis; How to do a coupled-field analysis.

These will be done through the following two sections:

1. Thermal analysis 2. Thermal-stress analysis

MATH 2034 Lecture Notes

10-8

Dr. Yan DING Dept. of Maths & Stats. RMIT University

1. Thermal Analysis
Thermal analyses are used to determine the temperature distribution, thermal gradient, heat flow, and other such thermal quantities in a structure. A thermal analysis can be steady-state or transient:
Steady-state implies that the loading conditions have settled down to a steady level, with little or no time dependency. Transient implies that conditions are changing with time. A typical example is a casting in the process of cooling down from molten metal to solid.

MATH 2034 Lecture Notes

10-9

Dr. Yan DING Dept. of Maths & Stats. RMIT University

1. Thermal Analysis

Thermal Loading Conditions

Temperatures: Convections: Regions of the model where temperatures are known. Surfaces where heat is transferred to (or from) surroundings by means of convection. Input consists of film coefficient h and bulk temperature of the surrounding fluid Tb . Surfaces where the heat flow rate per unit area is known. Points where the heat flow rate is known. Regions where the volumetric heat generation rate is known Surfaces where heat transfer occurs by means of radiation. Input consists of emissivity, Stefen-Boltzmann constant, and optionally, temperature at a space node. Perfectly insulated surfaces where no heat transfer takes place.
10-10 Dr. Yan DING Dept. of Maths & Stats. RMIT University

Heat flux: Heat flow: Heat generation: Radiation:

Adiabatic surfaces:

MATH 2034 Lecture Notes

1. Thermal Analysis

Element Attributes
1) Thermal element types: Thermal analyses use thermal elements only. A thermal element has only one DOF per node. The commonly used thermal element types are:
2-D Solid Linear Quadratic 3-D Solid 3-D Shell Line Elements

PLANE55 PLANE77 PLANE35

SOLID70 SOLID90 SOLID87

SHELL57

LINK31, 32, 33, 34

MATH 2034 Lecture Notes

10-11

Dr. Yan DING Dept. of Maths & Stats. RMIT University

1. Thermal Analysis

Element Attributes
2) Material Properties: Minimum requirement is the thermal conductivity, KXX. Specific heat (C) is required if internal heat generation is to be applied. ANSYS supplied material library contains both structural and thermal properties for a few materials. Generally the analysts create and use their own material library.

3) Real Constant: Primarily needed for shell and line elements

MATH 2034 Lecture Notes

10-12

Dr. Yan DING Dept. of Maths & Stats. RMIT University

1. Thermal Analysis

Thermal Loading
Prescribed Temperatures:
DOF constraints for a thermal analysis:
Solution > -Load- Apply > Temperature

Convections:
These are surface loads:
Solution > -Load- Apply > Convection

Adiabatic surfaces:
Perfectly insulated surfaces where no heat transfer takes place. This is the default condition, i.e, any surface with no boundary conditions specified is automatically treated as an adiabatic surface.

Other possible thermal loads:

Heat flux (BTU/Hr-in2 or W/m2) Heat flow (BTU/Hr or W) Heat generation (BTU/Hr-in3 or W/m3) Radiation (BTU/Hr or W)
10-13 Dr. Yan DING Dept. of Maths & Stats. RMIT University

MATH 2034 Lecture Notes

1. Thermal Analysis

Results
The results of a thermal analysis are written to a result file, jobname.rth, as well as to the in-memory database. Review results typically consists of contour plots of temperature, thermal gradient, and thermal flux:
General Postproc > Plot Results > Nodal Solu (or Element Solu )

A useful option for contour plots in 3-D solid models is isosurfaces, which are the surfaces of a constant value:
Utility Menu > PlotCtrls > Style > Contours > Contours Style

Results validation. Are temperatures within the expected range?

You can generally guess the temperature range based on the prescribed temperatures and convection boundaries.

Is the mesh adequate?

In the case of stresses, you can plot the un-averaged thermal gradients (element solution) and look for elements with high gradients. These regions are the candidates for mesh refinement. If there is a significant difference between the nodal (averaged) and element (un-averaged) thermal gradients, the mesh may be too corse.

MATH 2034 Lecture Notes

10-14

Dr. Yan DING Dept. of Maths & Stats. RMIT University

2. Thermal-Stress Analysis
The followings are discussed:
How to apply thermal loads in a stress analysis; How to do a coupled-field analysis.

These will be done through the following sections:

A. Overview B. Sequential Method C. Direct Method

MATH 2034 Lecture Notes

10-15

Dr. Yan DING Dept. of Maths & Stats. RMIT University

Thermal-Stress Analysis

A. Overview
Thermally Induced Stress:
When a structure is heated or cooled, it deforms by expanding or contracting. If the deformation is somehow restricted, either by displacement constraints or an opposing pressure, for example, thermal-stresses are induced in the structure. Another cause of thermal stresses is non-uniform deformation, due to different materials (i.e, different coefficients of thermal expansion).

MATH 2034 Lecture Notes

10-16

Dr. Yan DING Dept. of Maths & Stats. RMIT University

Thermal-Stress Analysis

... Overview
There are two methods of solving thermal-stress problems in ANSYS. Both methods have their advantages and disadvantages: Sequential coupled field:
Older method, which uses two element types mapping thermal results as structural temperature loads. Efficient when running many thermal transient time points by few structural time points Can be easily automated with input files

Direct coupled field:

Newer method, which uses one element type to solve both physical problems Allows true coupling between thermal and structural phenomena. May carry unnecessary overhead for some analyses.
MATH 2034 Lecture Notes 10-17 Dr. Yan DING Dept. of Maths & Stats. RMIT University

Thermal-Stress Analysis

B. Sequential Method
The sequential method involves two steps of analyses: 1) First, do the Thermal Analysis 2) Then, do the Structural Analysis

1.

The thermal analysis (steady-state or transient) : - Refer to slides No. 10-9 ~ No. 10-14: Model with thermal elements Apply thermal loading Solve and review results (the results file: jobname.rth)

MATH 2034 Lecture Notes

10-18

Dr. Yan DING Dept. of Maths & Stats. RMIT University

Thermal-Stress Analysis

... Sequential Method

2. The static structural analysis: Switch element types to structural. Define structural material properties, including thermal expansion coefficient. Apply structural loading, including temperatures from thermal analysis. Solve and review results (the result file: jobname.rst). The GUI paths for the above procedure are given in the next slide.

MATH 2034 Lecture Notes

10-19

Dr. Yan DING Dept. of Maths & Stats. RMIT University

Thermal-Stress Analysis

... Sequential Method

The Procedure of the Structural Analysis :
a) Move to PREP7 and switch element types from thermal to structural:
Preprocessor > element Type > Switch Elem Type
Select Thermal to Struc, then [OK] Caution: switching element types will reset all element options back to their default settings. For example, if you used 2D axisymmetric elements in the thermal analysis, you may need to re-specify the axisymmetric option after the switch. Therefore, make sure to verify and set the proper element options:

Preprocessor > Element Type > Add/Edit/Delete > [Options]

b)

Define structural material properties, including the coefficient of thermal expansion (ALPX).
Caution: If ALPX is not defined or set to zero, no thermal strains will be calculated. By the way, this technique can be used to turn off temperature effects, if it is needed !

c)

Specify static analysis type. This step is needed only if the thermal analysis was a transient:
Solution > -Analysis Type- New Analysis Solution >-load- Apply > -Structural- Temperature > From Therm Analy
10-20 Dr. Yan DING Dept. of Maths & Stats. RMIT University

d) e)

Apply structural loads and include temperatures as part of the loading:

MATH 2034 Lecture Notes

Solve and review the results.

Thermal-Stress Analysis

C. Direct Method
The direct method involves just one analysis that uses a coupledfield element type containing all necessary degrees of freedom. The procedure is:
First prepare the model and mesh using one of the following coupled field element types:
PLANE13 (plane solid) SOLID5 (hexahedron) SOLID98 (tetrahedron)

Apply both the structural and thermal loads and constraints to the model. Solve and review both thermal and structural results. Only produces one result file: jobname.rst.

MATH 2034 Lecture Notes

10-21

Dr. Yan DING Dept. of Maths & Stats. RMIT University

Thermal-Stress Analysis

Sequential vs. Direct

Sequential:
For coupling situations that do not exhibit a high degree of non-linear interaction, the sequential methods is more efficient and flexible because the two analyses are performed independently of each other. Sequential thermal-stress analysis provides more flexibility in applying thermal load for a stress analysis. For example, you can perform a nonlinear transient thermal analysis followed by a linear static stress analysis, during which the nodal temperatures from ANY load step or time-point in the thermal analysis can be used as loads for the stress analysis.
MATH 2034 Lecture Notes 10-22

Direct:
Direct coupling is advantageous when the coupled-field interaction is highly nonlinear and is best solved in a single solution using a coupled formulation. Examples of direct coupling include piezoelectric analysis, conjugate heat transfer with fluid flow, and circuit-electric analysis.

Dr. Yan DING Dept. of Maths & Stats. RMIT University