****** Every line starting with * is ignored. * Separate data values with a space or a new line. * Maximum length of a data line : 90 characters. ******************************************************************************* * Dimension of the space (2 or 3) * Number of domains (K <= 4) * ---> The domains are assumed to be numbered counterclockwise. * Context for each domain * in dimension 2 : * 1 : isotropic elasticity * 2 : axisymetric orthotropic elasticity (C11=C22) * 3 : orthotropic elasticity * 4 : general elasticity * in dimension 3 : * 1 : isotropic elasticity * 4 : general elasticity * Cij coefficients for each domain. According to the context, give * in dimension 2 : * 1 : lambda mu (that is : C12 C66) * 2 : C11 C12 C66 * 3 : C11 C12 C22 C66 * 4 : C11 C12 C16 C22 C26 C66 * in dimension 3 : * 1 : lambda mu (that is : C12 C66) * 4 : the 21 coefficients in the following order : * C11... C16 C22... C26 C33... C36 C44... C46 C55 C56 C66 * Angles (in degrees) defining the geometry of the domains, * that is omega(i), i = 0,...K * Boundary conditions at omega(0) and omega(K) * 1 : Dirichlet * 2 : Neumann * 3 : Mixed normal Dirichlet - tangent Neumann * 4 : Mixed tangent Dirichlet - normal Neumann * 5 : Transmission * Action to be done : integer 10*a1 + a0 such that * a0 = 0 (Manual computation), 1 (Automatic computation (= AC)), * 2 (AC + rotation of axes), 3 (AC + variation of the geometry) * 4 (AC + orientation of the materials), 5 (AC + edge tracking) * 6 (AC + variation of a coefficient) * --> a0 = 2, 4 or 5 in dimension 3 only * a1 = 0 (Kernel not computed), 1 (Computation of the kernel of A(nu)) * 2 (1 + computation of the singular functions) * Rectangular domain where to search for the roots (necessary if a0 > 0) * Give xmin xmax ymin ymax such that 0 < xmin < xmax and ymin < ymax. ********************* Mutually exclusive specific actions ********************* * -> Do not modify the keyword "ACTIONi" that marks the beginning of data. * ACTION2 (a0 = 2 : Automatic computation + rotation of axes) * Definition of a rotation * 1) starting angle, end angle, step (in degrees) * 2) number of the rotation axis ACTION3 (a0 = 3 : Automatic computation + variation of the geometry) * Variation of the geometry of a domain * 1) number of the concerned boundary (0 <= i <= K) * 2) starting angle >=omega(i-1), end angle <= omega(i+1), step (in degrees) ACTION4 (a0 = 4 : Automatic computation + orientation of the materials) * 1) For each domain, give on the same line : the starting angle, * the step (in degrees) and the number of the rotation axis * 2) Give the number of steps to do and the domain number that defines * the abscissa angle on the graphs ACTION5 (a0 = 5 : Automatic computation + edge tracking) * For each computation point and up to end of file, give the value of the * parameter governing the computation, the 9 coefficients of the rotation * matrix defining the position of the local basis (row-wise), the angles (in * degrees) defining the geometry of the domains, that is omega(i), i = 0,...K ACTION6 (a0 = 6 : Automatic computation + variation of a coefficient) * Give the number of the Cij varying coefficient, then the number of the concerned * domain. The number of the Cij coefficient is one of those given previously * with respect to the context and the dimension (see 4th group of data). * Then, give the values taken by this coefficient, one value per line, * up to end of file.
In this section, we give informations about the settings of the program, that is the value of the main parameters governing the execution of the program that may be modified. If it happen that one of these parameters is modified, the program has to be compiled and linked again by the administrator of the program.Note that some of the parameters are critical with regard to precision and even the way the program runs, since they may interact. So modifying them must remain exceptional.
Except internal constants depending of the algorithm used, all the parameters that configure the program are grouped in a single file, which is taken into account at compilation time via the include instruction.
Here follow the main parameters, with their default values.
General parameters
Parameters concerning precision
- nbdomx = 4
- Maximum number of subdomains. It can be increased without any trouble.
- pimin = -2*toler
- The program computes only the exponents that have their imaginary part greater than this value. The default value is chosen in order to obtain the real exponents. Only the exponents with positive imaginary part have to be computed since their conjugates are also singularity exponents. However, pimin may be set to any arbitrary value, depending on the user's needs.
- prmin = toler
- The program computes only the exponents that have their real part greater than this value. The default value is chosen in order to avoid the zero solution. Only the exponents with positive real part have to be computed since their opposites are also singularity exponents. However, prmin may be set to any arbitrary value, depending on the user's needs.
- toler = 1.e-9
- Tolerance defining the precision requested to compute the exponents. This is the maximum distance between two successive estimations of an exponent.
- nbpcer = 64
- Number of points used to compute a Cauchy integral over a circle. It is advised not to reduce the default value. Increasing it will result in more accurate approximations of the integrals and will generally lead to a larger global computation time.
- nbprec = 10
- The number of points used to compute a Cauchy integral over a rectangle is roughly 4 times this value. It is advised not to reduce the default value. Increasing it will result in more accurate approximations of the integrals and will generally lead to a larger global computation time.
- taimai = 5.e-5
- Minimum dimension allowed for the smaller length of the rectangles used in the splitting procedure of the algorithm. Increasing it may result in a faster computation, but the precision on some exponents may be affected, especially when there is multiplicity.
| Summary |