U.S. ATLAS HL-LHC Upgrade Level 2 and Deputies-NSF only Management Mailing List

[Gustaaf Brooijmans <gusbroo AT nevis.columbia.edu>]

Hi Reinhard,

In general, I think we shouldn’t include in talks plots we haven’t had time to understand, so I think we should take the next week or so to get that understanding and only include them in back-up at the last moment if every L2 is comfortable with their contents.  If one reviewer asks a question that shows there’s a problem it will discredit our full contingency calculation, an outcome we do not want.

These plots are by L2, but the schedule contingency task is sometimes before you end date, and usually only in one L3, so doesn’t this bias the “impact on end date” as the end date of one L3 might be artificially delayed?

Then I don’t understand the numbers in the sensitivity plots.  Can you explain what they mean, eg for the LAr “delay in launch of FEB2 production”?  The schedule delays are O(10d), adding up to 119d, but the needed contingency is 200+ days.  Maybe we should include these sensitivity plots in slides at all as they seem hard to explain.  Even the driver plots I think we should be careful with.

Best,

Gustaaf

On Jul 10, 2024, at 05:43, Schwienhorst, Reinhard <schwier AT msu.edu> wrote:

Dear L2s, deputy L2s,
I have finally completed all of the tornado charts for drivers and for sensitivity for all L2s. Here are the plots, obtained from the risk-only simulation (since the uncertainty numbers don’t have such a clean and obvious interpretation). You could include these in your backup slides in case anyone asks for details during the review. There are five plots per system, two each for drivers and sensitivity, plus one more combined one for sensitivity, which are the names used by Safran. There is one for the schedule delay at L2, and one for the cost increase at L2.

- Drivers: This is computed as the correlation coefficient between a given risk and the schedule delay (or cost increase) for that L2.
- Sensitivity: This is computed by re-running the entire simulation without this risk, and then taking the difference between the full simulation and this run. I.e. this shows how much each risk individually would increase the delay (or the cost) of that L2, with all other risks in place.

These show more-or-less the same information (the important risks are at the top in both cases), but one or the other might be more useful.
A few more comments:
- All calculations are for the 90% level. All subtractions are linear.
- Not all plots include all risks, only the most relevant ones are shown for more clarity.
- I also include a sensitivity plot showing both cost and schedule at the same time, it’s condensed but might work for some L2s.
- For trigger, I’m including all risks in the cost plots so that you can see the opportunities.
- The sum of the sensitivities for a given system is smaller than the total delay or cost increase due to correlations between risks from the simulation.
- I have also updated the risk register in docdb, #196.
Let me know if you have any questions or comments. 

Thanks,
Reinhard

<lar-drivers-riskonly-cost.png>

<lar-drivers-riskonly-schedule.png>

<lar-sensitivity-riskonly-cost.png>

<lar-sensitivity-riskonly-schedule.png>

<lar-sensitivity-riskonly-cost-schedule.png>

<tile-drivers-riskonly-cost.png>

<tile-drivers-riskonly-schedule.png>

<tile-sensitivity-riskonly-cost-schedule.png>

<tile-sensitivity-riskonly-cost.png>

<tile-sensitivity-riskonly-schedule.png>

<muon-drivers-riskonly-cost.png>

<muon-drivers-riskonly-schedule.png>

<muon-sensitivity-riskonly-cost-schedule.png>

<muon-sensitivity-riskonly-cost.png>

<muon-sensitivity-riskonly-schedule.png>

<trigger-drivers-riskonly-cost.png>

<trigger-drivers-riskonly-schedule.png>

<trigger-sensitivity-riskonly-cost-schedule.png>

<trigger-sensitivity-riskonly-cost.png>

<trigger-sensitivity-riskonly-schedule.png>