PDA

View Full Version : Why your electronic water pump fails



BenSTi
10-15-2015, 07:58 PM
Hey Guys, First post here.

I recently purchased a 2009 535i used with ~49k on it. My friend I purchased it from is a mechanic and told me that the water pump would fail between 60-70k miles, "sure" I said. Well, at 64k it did, and it sucked. I had to limp the car home making it about two miles at a time before pulling over and letting it cool off.

As we all know, when the water pump fails, it's immediate and without warning. I decided to tear down my pump to figure out what had failed on it and if there was any way to detect the failure or even prevent it. Below is what I found.

Here's the pump taken apart, it comes apart in three pieces, the injection molded top plastic, cast aluminum electronics carrier, and the housing with stator (motor coils). The electronics plate is connected to the stator housing with the three motor phase busbars which are ultrasonic welded together. You'll need to pry hard enough to break these to get them apart. Otherwise it's just sealed and held on with an O-Ring and a few screws. The plastic top part has the pins moulded through it. This was likely a post-process done with ultrasonics after assembly.

https://farm1.staticflickr.com/593/22016206038_db4a541784_k.jpg

Once it's all seperated we can get a good idea of the beautiful electronics inside. It's a really impressive three dimensional design, stuffed into a tight package!

https://farm6.staticflickr.com/5800/22191262262_039781342d_k.jpg
https://farm6.staticflickr.com/5692/22203960675_1216e61e34_k.jpg
https://farm1.staticflickr.com/730/21581200174_1b362add09_k.jpg

On the "top layer" there is a very large electrolytic capacitor and an inductor. The capacitor is a TDK part, EPCOS B41695-S6228, datasheet here (http://en.tdk.eu/inf/20/30/db/aec_2013/B41695_B41795.pdf). It's interesting to note the rated lifetimes according to TDK, at 125C it's rated to >3000hrs, and at 85C it's rated to >8000hrs. Since the pump is directly in contact with coolant (~100C) we can assume the cap gets to this temp. That would put the rated lifetime at ~6000hrs. At an average speed of 25mph, that's 150k miles.

https://farm1.staticflickr.com/661/22177848596_c4497c09c8_k.jpg

This assembly seems to be a filter to reduce back EMI onto the vehicle 12V buss. This is typically tuned to the switching frequency of the inverter. Doing some quick math, this supports a switching frequency of ~1500hz (dcR of the system seems to be ~50mOhm)

If we cut the three busbars and remove this assembly, we discover a very cool inverter below.

https://farm6.staticflickr.com/5617/21582942293_0946b48ecb_k.jpg

On the left side, there is +12V and Ground Input, plus gate drive and sense lines. The capacitor seen is between +12V and Ground. On the right side, there is the three phase outputs to the stator. Going forward, I call these PH1, PH2, PH3 from bottom to top in the photo.

The six large transistors are IGBTs, not sure of the manufacturer. Going clockwise from the left top: PH1+, PH2+, PH3+, PH3-, PH2-, PH1-. There seems to be very small capacitors connecting each phase together and to +12V/Ground, interesting...

You'll notice that both PH3 IGBTs have failed catastrophically. When measured +12V to PH3out, I get ~15mOhm, same with Ground to PH3out. This shows that both have lost isolation. Without this phase, the motor simply cannot work. This likely was a cascade failure that happened in one of two ways:

Possibility One: One IGBT shorted, when the control board commanded the other IGBT in that phase to turn on, it created a very low resistance short between +12V and Ground, frying both parts.
Possibility Two: The control board commanded both IGBTs in the phase to close at the same time, same result as above.

It seems more likely to have one of the IGBTs short, this is a common failure mode, and because failure of any one part causes system failure, it is much more likely to happen.

So there you have it, the pump failure in my car was caused by shorted IGBTs on one phase resulting in sudden failure.

A couple other observations of note:

The rotor is over-molded with plastic, this is cracking and almost ready to come apart!

https://farm1.staticflickr.com/594/22203981165_5f1d62aa90_k.jpg

The rotor and inside plastics on the stator are actually flooded with coolant. This avoids the need for any rotating seals. Clever.

https://farm1.staticflickr.com/622/22203983335_f988550538_k.jpg

The control board is constructed on aluminum nitride with wire-bonded components and an exposed die, very cool. It's encapsulated in a soft silicone material.

https://farm1.staticflickr.com/744/22017052919_8f308d603d_k.jpg

There is also a board which I will call the "input board" This is where all of the signals come into before they are broken out to the inverter board and control board. It also transmits signals between the inverter and control board. It includes even more capacitance between +12V and Ground, +12V and Chassis, and Ground and Chassis. It's worth noting the entire circuit seems isolated from the case, the only ground path is through the connector, not the enclosure like many other components. This could be because it has low-side switching or current sensing on the ECU end.

https://farm1.staticflickr.com/767/22177820376_612f50274a_k.jpg

TL;DR: IGBTs shorted, caused catastrophic failure of both IGBTs on one phase. Stuff doesn't work anymore. No way to detect failure prior to happening, no way to prevent it from happening on existing pumps, no way to repair after it happens.

Cyrix2k
10-15-2015, 08:56 PM
Very interesting, thanks for posting.

MarcoZandrini
10-16-2015, 08:54 AM
Wow! No wonder these pumps cost a small fortune!!!! Your description of the pump and its operation is a great read for a techie like me. Thanks!

A couple of questions:
- if you could ID and obtain the suspected failed parts, could you fix the pump, albeit not the pump in the thread.
- do you see any way to make the pump last longer than the typical 65k to 85k miles?

BenSTi
10-16-2015, 01:24 PM
Wow! No wonder these pumps cost a small fortune!!!! Your description of the pump and its operation is a great read for a techie like me. Thanks!

A couple of questions:
- if you could ID and obtain the suspected failed parts, could you fix the pump, albeit not the pump in the thread.
- do you see any way to make the pump last longer than the typical 65k to 85k miles?

The IGBT in question is reflowed or diffusion bonded onto the aluminum nitride substrate. There's no good way to remove it. Even if you could, you'd need to re-wirebond the part. Anything can be replaced, but it would be far far cheaper to just buy a new pump.

As far as ways to make it last longer, in a normal application, likely no practical way. This would likely last a very long time if it was run at a lower temperature, since the IGBTs are directly coupled to the coolant. You could also reduce time-at-temp by using the pump to cool the car after you turn off the engine. Both would likely have negative effects on other, possibly more expensive, components.

It would be interesting to see how the newer pumps hold up. It's very likely that they have changed the design, or used more reliable components which may eliminate this problem. The car should really warn you as soon as it detects the pump has failed. In my case, I made it all the way into work (~15 miles) on a cool morning without any warning of overheating, but the car clearly knew there was a problem since the cooling fan was going at 100% the whole way. It was only on the drive home where it warned me, but that didn't happen until I was on the highway.

MarcoZandrini
10-17-2015, 09:28 AM
Thanks again.

Kevlar
10-17-2015, 07:39 PM
Awesome!

Sent from my iPhone using BF.com

sar5036
10-19-2015, 10:57 AM
Awesome writeup. I have almost no experience with circuitry so please excuse the question, but there wasn't really an explanation as to why it couldn't be fixed. Seems as though the IGBTs cannot be replaced?

DavidD123
10-19-2015, 10:27 PM
AWESOME writeup! I really wasn't expecting something so detailed. Hopefully someone will come put with a quality OE replacement water pump to offset the costs of having to replace them so often. Or better yet!..... an improved pump!

bcromwell
10-19-2015, 11:46 PM
Another failure mode that occurred with mine is the brown plastic around the rotor cracks breaks free and cause the shaft to sieze. You can see the cracking in the op picture as well.

Stück
10-21-2015, 03:02 PM
Best F'in 1st post ever.

Kevlar
10-22-2015, 06:42 AM
A better pump will be needed cause I think the new m cars have like 3-4 of them.

Sent from my iPhone using BF.com

BenSTi
10-24-2015, 11:42 PM
Awesome writeup. I have almost no experience with circuitry so please excuse the question, but there wasn't really an explanation as to why it couldn't be fixed. Seems as though the IGBTs cannot be replaced?

Thanks!
It's not that it can't be fixed, but rather it would be prohibitively expensive to replace, and even if you replace the two failed parts, the other four parts have accumulated damage over their lifetime and will be more prone to fail. The best bet is to replace the whole pump.


AWESOME writeup! I really wasn't expecting something so detailed. Hopefully someone will come put with a quality OE replacement water pump to offset the costs of having to replace them so often. Or better yet!..... an improved pump!

Thanks!
I think the biggest thing that can be done is change from six IGBTs to twelve (two per polarity on each phase in parallel). This would help prevent damage from shoot through, though there would be a slight cost increase.


Another failure mode that occurred with mine is the brown plastic around the rotor cracks breaks free and cause the shaft to sieze. You can see the cracking in the op picture as well.

The plastics on mine were cracked apart, but I don't see how that would cause a catastrophic failure unless it caused the impeller to fail.


Best F'in 1st post ever.

Thanks!


A better pump will be needed cause I think the new m cars have like 3-4 of them.

Sent from my iPhone using BF.com

With more electronics it becomes harder to ensure reliability. You can be sure though that manufacturers do learn from their mistakes and design out problems on new parts. It's unlikely for them to make the same mistake twice on something like this.

petr4271
02-11-2019, 03:09 PM
Mine seems to fail on electrical issue from resistor inside, not a coolant leak in.

https://www.youtube.com/watch?v=pGIZwgjNs5Y&lc=z22aidrqvx3ot3bgn04t1aokgedhgkf00y111bkavmkhrk0 h00410

ross1
02-23-2019, 07:50 AM
Fabulous complication employed in what could be the simplest of devices.
Congrats on the diagnosis.
I hope your driving 2 miles at a time with no coolant circulation won't have you diagnosing the consequences of an even more expensive failure.
If you think a tow or this pump are expensive.............