Bruno, that's why i said, his mistake is "looking back".

Look at my approach (looking forward)

Hello Zvoni,

Actually for this process routing graph, I should look both backwards and forward.

For the 1st point, I must find in column et_modeop_pere the value is null, so node 8251 in et_modeop is the 1st point(1st product process).

Then I'll check 8251 in column et_modeop_pere, to find the corresponding node 8253 in et_modeop, then 8253 is the 2nd point(2nd product process).

Follow this idea, next step to check 8253 in et_modeop_pere, find 8255 in et_modeop,8255 is the 3rd point.

In this way,can find 1st and 30th point(means product manufactured after 30 steps, it turns from component become final product), still 6 values not used in the data.

These are next branch after node 8685. Check 2nd 8685 in et_modeop_pere, find corresponding node in et_modeop, can finish the next branch.

But still the drawing after node 8685 is a little difficult.