1 |
On Sunday 26 April 2015 15:49:19 Alan McKinnon wrote: |
2 |
|
3 |
> I disagree. emerge really needs to have it's output redesigned from |
4 |
> scratch. Right now it arrives at the conclusion (the top) and dumps it's |
5 |
> data tree bottom-up, apparently stopping halfway and never getting to |
6 |
> output what the top is. |
7 |
|
8 |
This discussion reminds me of an AGR (Advanced Gas-cooled Reactor) power |
9 |
station being commissioned 40-odd years ago, in which a minor plant failure |
10 |
(e.g. a high oil temp in an auxiliary pump) could lead to cascades of other |
11 |
abnormal conditions, making it impossible for a reactor operator to discover |
12 |
the real problem he needed to fix, and ignore all the others. So we put up a |
13 |
second alarm screen alongside the first, showing major causes only; it was |
14 |
populated with the output of an analysis routine that was called whenever an |
15 |
alarm occurred. |
16 |
|
17 |
All those decision trees had to be designed and input by hand (on paper tape, |
18 |
like everything else). Not to mention testing, of course. And there were |
19 |
thousands of digital inputs. |
20 |
|
21 |
Maybe something like that would be useful here. |
22 |
|
23 |
-- |
24 |
Rgds |
25 |
Peter |