The following process is given as an example of a convergence criterion mentioned in
paragraph 5.4 of appendix 1.
1 In total, a minimum of 50 different simulations should be carried out for
each of the benchmark cases. This will yield, for each case, a total of at least 50
values of tA. More than 50 simulations may be required according to the
outcome of the convergence test (3 and 4 below), which requires to increment the number
of simulations one by one (see 3) and to test the criterion every batch of 50 simulation
runs (see 4).
2 These simulations should be made up of at least 10 different randomly
generated populations (within the range of population demographics specified in
paragraph 3 of appendix 1). Simulations based on each of these different populations
should be repeated at least 5 times. If these 5 repetitions produce insignificant
variations in the results, the total number of populations analysed should be 50 rather
than 10, with only a single simulation performed for each population.
3 Observed 95th centile of tA
3.1 For each case, the evaluation of the 95th centile is an incremental evaluation which
is performed every simulation run using all available tA previously
calculated from the first to the last simulation run of the case studied.
3.2 The value of the 95th centile of all calculated total assembly times
(noted T0.95) is taken which is higher than 95% of all the previous
calculated values (i.e. for each of the four cases, for each simulation run increment,
indexed on letter "i" below, all available values of assembly times tA of the
case are ranked from lowest to highest and 
is selected for which 95% of the ranked values are
lower. Consequently, at the simulation number i, there is a series of i values of
).
4 Convergence criterion
4.1 For each case, the convergence test is an evaluation of the following criterion which
is performed every batch of 50 simulation runs. N denotes the number of simulations that
have been run every time the criterion is tested (i.e. N = 50 for the first batch, N =
100 for the second batch etc.)
4.2 The distance between the maximum to the minimum of
obtained over the 50 last simulation increments should
not exceed the distance (in absolute value) of the mean of
over the 50 last simulation increments, to the maximum
allowable assembly time (T
lim):
- Where:
with n, E, and L, as defined in Annex1, §5.1
(1),
with I between (N − 49) and N,
with i between (N−49) and N,and
with i between (N−49)and N.
4.3 For each of the four cases, the following iterative method should be followed to
determine the travel Time Tcase:
- - If the criterion is not met, another batch of 50 simulations should be
run;
- - If the criterion is met, sufficient number of simulations has been
run for the case.
(for the first N which satisfies the criterion) is
selected as the travel time Tcase; and
- - If a total of 500 simulations have been run for the case, the process
should be stopped and
is selected as the travel time Tcase.
5 The value of the travel time to comply with the performance standard T is the highest
of the four calculated travel times Tcase (one for each of the four
cases).
6 The same procedure for a convergence criterion for case 5 and the travel duration in
case 6 (travel duration from assembly stations to the LSA entry points) can be based on
the same principle (paragraph 1 to 5). For case 6, the procedure requires to adapt the
notations (tA) and to take into account (E+L) ≤ 30' (see annex 1, paragraph
5.1 (2) for the definition of Tlim.