is used in the “HAT example”. Besides, based on UserManual_Kriging, the term “Gaussian” can be substituted by other types of “.Kriging.Corr.Family” such as Matern-5_2.
Finally, I was wondering to know what are the other specific options of APCK-MCS reliability method using “.AKMCS.PCK.” command which can be used in the PCK form of an adaptive reliability-based approach?
Thanks for your question. I am not sure I completely understand what you are asking. Are you asking what options you can specify for PCK when you use it as metamodel in AKMCS? According to the Reliability Analysis user manual, Section 2.3.6.2 and Table 11, it seems you can specify any options that you would specify for a simple PCK metamodel. Did you run into any problems with this? Or do you rather want to know what are reasonable settings for APCK-MCS?
Thanks for your response.
As you mentioned, based on the UserManual_Reliability, we could have specific options of PCK metamodel using the below command:
.AKMCS.PCK.
Thanks to the “Hat example”, we can compare the performance of AK-MCS and APCK-MCS. I am really enthusiastic to know about the other specific options which could only used in APCK-MCS reliability analysis.
Could you please list the unique options of APCK-MCS which are not in the AK-MCS?
As far as I understand, the options of AK-MCS and APCK-MCS are identical, except for the options specifying the metamodelling method, which is Kriging in the first case and PC-Kriging in the second. Have a look at table 11 in the Reliability analysis user manual. The difference between the methods is the choice of the metamodelling method.
I hope this answers your question. If not, could you please tell us in more detail what you are planning to do with APCK-MCS? Maybe we can figure it out with the help of a concrete example
Thanks for your quick answer and guidance.
Let us consider the uq_Example_Reliability_02_hat example as the case study.
It seems there are just two differences between the AK-MCS and APCK-MCS methods.
1- Obviously, between the type of the metamodeling method which we must allocate the PCK term for using the APCK-MCS method:
APCKMCSOpts.AKMCS.MetaModel = ‘PCK’; %for PCK
2- Just one another option for the APCK-MCS which is relevant to the correlation family of the PCK approach:
It seems there is no other unique option for the APCK-MCS method which could be only assigned to this approach by the below command:
.AKMCS.PCK.
I am a little bit curious about the availability of any other specific option for the APCK-MCS method using the term ".AKMCS.PCK."due to the discussion of UserManual_Reliability, p.35, and 45.
If you want to use AKMCS, you would specify MyOpts.AKMCS.MetaModel = 'Kriging';
and then specify whatever options you want for the Kriging model in MyOpts.AKMCS.Kriging. Please check the Kriging user manual, section 3.1, for all available options.
Similarly, if you want to use APCK-MCS, you would specify MyOpts.AKMCS.MetaModel = 'PCK';
and then specify whatever options you want for the PC-Kriging model in MyOpts.AKMCS.PCK. Check the PC-Kriging user manual, section 3.1, for all available options. Basically, what you would assign to the struct MetaOpts to create a simple PC-Kriging model, you now assign to MyOpts.AKMCS.PCK.
No: the line you quote from the manual,
is just an example. You can specify here any options that are valid for PCK-Kriging. Maybe it would be clearer to reformulate the sentence in the manual to Options specific to the PCK model can be added in the field .AKMCS.PCK.
I suggest you look up some options for PCK from the PC-Kriging manual and modify the hat example by adding them to APCKMCSOpts.AKMCS.PCK. Does this clarify the issue?
