DGLOBAL

Bulk Data Entry Defines input parameters required for the Global Search Option (GSO).

Format

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
DGLOBAL ID NGROUP NPOINT SPMETH NOUTDES DESTOL      
+   MAXSP MAXSUCC MAXWALL MAXCPU        
+ GROUP SID1 NPOINT1 SPMETH1          
+ GROUP SID2 NPOINT2 SPMETH2          
+ etc etc              

Definitions

Field Contents SI Unit Example
ID Each DGLOBAL card must have a unique ID.

No default (Integer > 0)

 
NGROUP Number of groups of design variables. 1
AUTO (Default)
blank

(Integer > 0)

 
NPOINT Number of starting points for each group of design variables. 1
AUTO (Default)
blank

(Integer > 0)

 
SPMETH Method used to generate the starting points. 2
OFFSET (Default)
The lower and upper bounds are not considered as starting points.
EXTREME
The lower and upper bounds are included in the list of starting points.
blank
 
NOUTDES Number of unique designs to be saved. 4
AUTO (Default)
blank

(Integer > 0)

 
DESTOL Unique design tolerance. 4

Default = 1% (Real > 0.0 or blank)

 
MAXSP Maximum number of starting points. 5

Default = 20 (Integer > 0 or blank)

 
MAXSUCC Maximum number of consecutive starting points without finding a unique design. 5

Default = 10 (Integer > 0 or blank)

 
MAXWALL Maximum amount of WALL time (in hours). 5

Default = infinite (Real > 0.0 or blank)

 
MAXCPU Maximum amount of CPU time (in hours). 5

Default = infinite (Real > 0.0 or blank)

 
GROUP Indicates that design variables grouping information is to follow. 3  
SID# Design variables SET identification number.

No default (Integer > 0)

 
NPOINT# Number of starting points for the current group of design variables.
AUTO (Default)
blank

(Integer > 0)

 
SPMETH# Method used to generate the starting points for the current group of design variables.
AUTO (Default)
OFFSET
EXTREME
blank
 

Comments

  1. Design variables are automatically organized in groups, and design variables within a given group are assigned the same relative starting points, such as their lower or upper bound. With the AUTO option, OptiStruct determines NGROUP and NPOINT so as to generate a reasonable and manageable number of starting points. For small optimization problems, each design variable might be assigned its own group. By default, the number of groups is equal to the number of independent design variables with an upper limit of 10. For larger optimization problems, design variables are grouped together in order to consolidate the potential starting points. The values of NGROUP and NPOINT vary based on the model being solved. If the number of groups (NGROUP) is set too low, many design variables will be changed at the same time and variation in the optimum result is minimal, and therefore, does not serve the purpose of using the Global Search Option. If too many groups are defined, it may lead to unnecessary runs that converge to the same local optimum. The number of starting points is equal to (NPOINT)NGROUP, unless MAXSP is set to a very low value or MAXSUCC is attained (MAXWALL and MAXCPU can also affect the number of starting points).
  2. With the EXTREME and OFFSET options, the starting points are distributed evenly.
  3. In situations where finer control is required, design variables can be grouped manually by creating DESVAR SETs. NPOINT# and SPMETH# can also be defined for individual groups. If those parameters are not defined for a specific group, they inherit their value from the generic NPOINT and SPMETH parameters.
  4. The unique design tolerance DESTOL provides the threshold under which two designs are considered identical. It is measured as the average of the relative differences between the design variables at the last iteration.
    The following equation illustrates how DESTOL is applied. Two designs are considered identical only if the normalized average of relative differences between design variables at the last iteration of two designs from different starting points is lower than DESTOL. Two designs are considered unique if the following value is lower than DESTOL.(1)
    D V S P i + 1 F D V S P i F D V U B D V L B < DESTOL MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaWaaSaaaeaaca WGebGaamOvamaaDaaaleaacaWGtbGaamiuamaaCaaameqabaGaamyA aiabgUcaRiaaigdaaaaaleaacaWGgbaaaOGaeyOeI0IaamiraiaadA fadaqhaaWcbaGaam4uaiaadcfadaahaaadbeqaaiaadMgaaaaaleaa caWGgbaaaaGcbaGaamiraiaadAfadaahaaWcbeqaaiaadwfacaWGcb aaaOGaeyOeI0IaamiraiaadAfadaahaaWcbeqaaiaadYeacaWGcbaa aaaakiabgYda8iaabseacaqGfbGaae4uaiaabsfacaqGpbGaaeitaa aa@5141@
    Where,
    D V S P i + 1 F MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiraiaadA fadaqhaaWcbaGaam4uaiaadcfadaahaaadbeqaaiaadMgacqGHRaWk caaIXaaaaaWcbaGaamOraaaaaaa@3D03@
    Design Variable of the final iteration of the current starting point
    D V S P i F MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiraiaadA fadaqhaaWcbaGaam4uaiaadcfadaahaaadbeqaaiaadMgaaaaaleaa caWGgbaaaaaa@3B66@
    Design Variable of the final iteration of the previous starting point
    D V U B MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiraiaadA fadaahaaWcbeqaaiaadwfacaWGcbaaaaaa@3968@
    Upper bound of the Design Variable
    D V L B MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiraiaadA fadaahaaWcbeqaaiaadYeacaWGcbaaaaaa@395F@
    Lower bound of the Design Variable

    If such identical designs are found, the best occurrence is preserved, and other results are discarded. Up to NOUTDES unique designs are saved in subdirectories named <filename_#s_#u>.

    Where,
    #s
    Starting point
    #u
    Rank of the unique design
  5. The global search option stops searching for optimal designs when any of the following criteria has been met:
    • the maximum number of starting points (MAXSP) has been reached
    • the maximum number of consecutive starting points without finding a unique design (MAXSUCC) has been reached
    • the maximum amount of WALL time (MAXWALL) has been reached or exceeded
    • the maximum amount of CPU time (MAXCPU) has been reached or exceeded
    • all possible starting points have been explored
  6. In general, it is recommended to run the global search option with the default parameters, except for the termination criteria.
  7. The DGLOBAL Bulk Data Entry is referenced by the DGLOBAL command in the I/O Option section of the input data.
  8. The Global Search Option (GSO) in OptiStruct is supported for Size and Shape Optimization.