Monte Carlo Simulation.

[Tom Nifuku]

Hi OQ users,

I just started to use OQ so I would appreciate it if you guys would help me to handle with this.

Today, I have a simple question for Scenario Damage Assessment (using fragility curve and consequence function).

Can I conduct Monte Carlo Simulation at Fragility curve to simulate damage state?  If so, what kind of command is needed in **.ini file?

OR

Can't I perform Monte Carlo Simulation at anywhere other than Ground Motion Field?

I'm using Version 1.9.1-2 (& lubuntu1404).

Thank you for any helps in advance,

Tom

[anirudh.rao]

Hi Tom,

The OpenQuake-engine Scenario Damage calculator does not perform Monte Carlo sampling at the fragility level. The sampled ground motion fields are used to obtain the probabilities for the different damage states for each asset. These probabilities for each asset are then multiplied by the number of buildings (or structural units) for that asset to get the expected number of buildings in each damage state.

On the other hand, if you are calculating losses using the Scenario Risk or Stochastic Event-Based calculator, and you use either the lognormal or Beta distribution or a probability mass function to model the uncertainty in the conditional loss ratio given the ground motion intensity, the calculator will use Monte Carlo sampling also at the loss level. No additional commands are required in the .ini file; this is the default behavior.

Cheers,

Anirudh

[Tom Nifuku]

Hi Anirudh,

 

Thank you very much for your straightforward response.  And, please let me make one thing clear as follow.

 

If the damage distribution (not loss) according to all damage states of each asset needed to be simulated and the pre-estimated median value of seismic intensity in addition to fragility curve & exposure data could be prepared, the only way to conduct MCS for simulating damage distribution is that the sample sets (according to MCS times) of seismic intensity have to be generated based on a probability distribution and be converted to an independent gmfs.xml.  Then, the OQ risk calculation (Scenario Damage Assessment) can do MCS for damage distribution by applying this GMFs file.

 

Is this correct?  Is this the only way in the current OQ?

 

 

Thank you for your help in advance,

Tom

[Anirudh Rao]

Hi Tom,

If the damage distribution (not loss) according to all damage states of each asset needed to be simulated and the pre-estimated median value of seismic intensity in addition to fragility curve & exposure data could be prepared, the only way to conduct MCS for simulating damage distribution is that the sample sets (according to MCS times) of seismic intensity have to be generated based on a probability distribution and be converted to an independent gmfs.xml. Then, the OQ risk calculation (Scenario Damage Assessment) can do MCS for damage distribution by applying this GMFs file.

That is correct. Currently, this is the only way to obtain the distribution of damage states for a scenario.

Best regards,
ANIRUDH RAO    seismic risk engineer        +39-0382-5169-936
GLOBAL EARTHQUAKE MODEL    working together to assess risk
GEM -  globalquakemodel.org     @GEMwrld      GEMwrld

--
You received this message because you are subscribed to the Google Groups "OpenQuake Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openquake-use...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Tom Nifuku]

Hi Anirudh,

Thank you for your reply.  You helped me a lot!!  I'll try to do this.

Best wishes,

Tom

[Tom Nifuku]

Hi Anirudh,

 

Today, I have a question about this topic again.

 

I created an independent GMFs.file for doing Monte Carlo Simulation (MCS) and run Risk-Engine with other all necessary input data.  Currently, I can and cannot get results.  This is because I can successfully run it by a GMFs file consisted of 100 Monte Carlo sets, but I cannot unfortunately run it by a GMFs file consisted of 10,000 Monte Carlo sets.  Then, I tried to find the exact maximum number of Monte Carlo sets and found it is 110 sets.

 

I do want to make sure this is correct or not.  Is this 110 Monte Carlo sets the maximum in this case (running Risk-Engine with an independent GMFs file)?  Or, can I conduct MCS with more Monte Carlo sets (I hope so strongly!!)?

 

 

Thank you for your help in advance,

Tom

[anirudh.rao]

Hi Tom,

Within the OpenQuake-engine, there is no hardcoded limit on the number of ground motion fields you can use to run a scenario damage or loss analysis. The number of GMFs you can use would most likely depend on the amount of memory available on your machine for the computation.

Regards,

Anirudh

[Tom Nifuku]

Hi Anirudh,

 

Thank you very much for your quick response.

I carefully ran OpenQuake Risk-engine with my model again, and it seems I might have executed it wrongly.  Now, I can run a risk analysis based on 10,000 sets of GMFs and obtain proper results.  I’m sorry for confusing you.  By the way, the maximum times with my computer seems to be between 10,000 and 25,000 times.

 

 

Best wishes,

Tom