Exercise 5: System Optimization

Goal

Optimized a hypothetical safeguards system.

Note

Exercises so far have focused on guided examples of MAPIT functionality. Here, you will be working on your own to develop a notional safeguards system using a previously unconsidered example included in MAPIT. Solutions can be found here.

---

Objective

---

Start by loading the SNL curated dataset labeled protracted loss. Your goal is to determine the optimal uncertainties to obtain a 70% detection rate and 5% false alarm probability using Page’s trend test on Uranium. Start with the baseline assumption of 3% measurement uncertainties.

An optimal solution will consist of the cheapest set of sensors that can reach the goals stated above.

Note

For this example, we are neglecting that real world sensors will potentially have different random and systematic uncertainties. Instead, assume sensors have the same uncertainty for both values.

Tip

• Make use of the Save Error Config and Load Error Config to more quickly perform analyses. For example, when determining the FAP for a particular sensor configuration, use the Save Error Config before quitting. Then, when evaluating the sensor setup, use Load Error Config to quickly load the sensor setup without having to manually enter uncertainties again.

• For reference, there are 2 inputs, 26 inventories, and 3 outputs.

• Try to use 100 iterations or more for better probability estimates

Sensor costs

Sensor	Uncertainty (%)	Cost ($)
A	10.0	5
B	5.0	10
C	3.0	25
D	1.0	30
E	0.5	75
F	0.1	200
G	0.05	500
H	0.01	1000
I	0.005	2000

---

Baseline solution

---

For the baseline assumption, it was assumed that all measurement locations have a 3% uncertainty, implying that only Sensor Bs were used.

Cost

Sensor	Quantity	Cost ($)
A	0	0
B	0	0
C	31	775
D	0	0
E	0	0
F	0	0
G	0	0
H	0	0
I	0	0
Total Cost		775

Performance

Threshold = 1.50

False Alarm Probability (%)	Probability of Detection (%)
4.8	6.4

When finished, compare against our solution, which can be found here.