|
Hi Al,
I was not going to bother replying to your posts but your last one pushed me
over the top. I have a couple of questions:
1) Do
you have two ECMs? Having redundant crank angle sensors is great,
feeding them all into one computer is no better then having only one crank
sensor.
2)
Have you tested limp home mode on climb out?
3) How
is your power output with a partially shorted temp sensor? I'm betting it
is not going to be too good - like none.
4) How
old is your ECM? Most people junk their cars after 10 years, where did you
get data to support the reliability of the ECM as it ages?
I do have some experience in automobile failure
modes, I am a licensed auto technician with OBDII training. I have
colleagues who work on Subaru's, one fellow works at a local dealership, and two
others work at independent garages. FYI Subaru's, like every other vehicle
ever made, do come in on a hook and they do coast to a stop from time to
time.
One
last thing, you wrote:
"Sorry for being so long winded, I
have the impression that stuff like this haven't been discussed
before."
That's
right, we are all morons that don't grasp concepts like risk and
redundancy. While I plan to have fully redundant ignition and fuel, (using
an EC-2 as primary and a megaSquirtNSpark as a backup) Tracy (and his
many customers) concluded that some inputs like a crank angle sensor did not
represent a failure risk, he seems like a pretty smart guy who considers the
decisions he makes. 1600+ hours suggests to me that his decisions are
pretty good.
You
may be a great analyst - your skills as a diplomat leave a lot to be
desired.
--
Ian
First, let's try to get a perspective.
There is no job as creative as that of Design Engineer. This guy is
making hundreds of decisions. How many inputs do I need, what size resistor,
how wide should that track be, how do I isolate that from vibration, etc etc.
It's a very very high risk activity. So easy to overlook something. Many of
the decisions are arbitrary. You are just making your best guess.
The Japanese produce superior products. When we analyzed their success 30
years ago, we found they used certain tools in the design and validation phase
that U.S. designers didn't. One of these is the FMEA (see web site). They
get a group of engineers together and say" Ok, this is our best guess on
how it should be designed, what's going to fail?". They go thru each
characteristic and rate them for risk. Then they find a way to prove how far
from failure each of those items are.
For example, they'll say"Ok, the alternator is going to fail. This will
produce an ac voltage." So then they measure how large the ac voltage can get
before the device dies. Then they take action if there is not a large safety
margin, retest. They end up with numbers that measure their safety
margin.
So I would encourage reviewing all the various failure modes of the ECM.
Deliberately subject it to experiences beyond what it will normally see.
Unplug each sensor, see how it handles it. Apply heat way beyond normal, apply
vibrations beyond normal. There are very simple ways to do this. It doesn't
have to be some long drawn out thing.
However, statistically, we know if you have true redundancy in this
particular device, then you get to multiply the probability of failure. So if
the probability of shut down is 1 time in 1000 hours, since we have two with
independent probabilities, our odds plummet to 1 time in 1 million hours.
So all you need are two independent circuits.
When in doubt, just take a look at what the auto designers have done.
They use more than one sensor to measure each characteristic. They compare the
sensor results to historical data. They instantly recognize the sensor is
providing false data, then warn you, and use tables or other sensor to keep
you plugging along. That's why you don't see vehicles sitting on the side of
the road.
Sorry for being so long winded, I have the impression that stuff like
this haven't been discussed before.
-al wick
Artificial intelligence in cockpit, Cozy IV powered by
stock Subaru 2.5
N9032U 200+ hours on engine/airframe from Portland,
Oregon
Prop construct, Subaru install, Risk assessment, Glass panel design
info:
http://www.maddyhome.com/canardpages/pages/alwick/index.html
Al,
What changes would be required? (reference your
statement below)
Bill Schertz
KIS Cruiser # 4045
----- Original Message -----
Sent: Friday, June 03, 2005 9:50
PM
Subject: [FlyRotary] Re: Rotary
risks
You bring up very important points. If you guys can develop robust
solutions for each of the challenges, then you can end up with a
powerplant that has some fabulous failure modes. Here is a great
example, I suspect your ECM shutdown risk is now somewhere around 1 time
in 1000 (maybe 500) hours. But with simple changes that make the
system genuinely redundant, you would automatically raise that to 1
time in 1000000 hours. That is fantastic for a custom low volume
ECM.
|