Occam's Razor FTW

Those of you who work with me will already (more likely than not), be rolling their eyes at the title. That is because I've developed something of an unintended reputation of using unusual metaphors to highlight the strange and incongruous in the world of Centrifugal Pumps.

Now to be fair Occam's Razor is principle containing a metaphor, but I do tend to invoke it fairly often when working on engineering problems with colleagues and customers. For those of you not familiar with its principle, it is:

Occam's Razor states that among competing hypotheses, the one with the fewest assumptions should be selected - or in other words the simplest explanation is the one most likely to be correct...

For those of you still grappling with the principle, here is a simple worked example. Early on a Monday morning you arrive at your office cubicle fresh and ready to work, only to find a strange and unpleasant odor in the air. You have two possible explanations:

1. The Sardine and Kipper sandwich your spouse gave you for Friday lunch, which you then dumped in your trash bin (you don't like fish), has achieved an advanced state of decay.
2. A farting Wombat named Pete has taken up residence in the cubicle next to yours.

Which one is more likely ?

If you said #1, then you'd be correct as in that explanation, most of the facts are known to you. The only assumption required is that the cleaning crew did not empty the trash bin on Friday evening. In comparison #2 requires multiple assumptions in order to be valid.

At this point many of you are already wondering when I'll actually write something about pumps - so I will. Let's move on to the pump problem encountered last week.

Large Pumps Failing Lowly

A large slurry pump was on test and reported to have extremely low head, far lower than expected as well as being unable to achieve the rated flow. The relevant details were:

• 26" (660mm) full diameter impeller, trimmed to 21" (533mm), 880 RPM
• HI 14.6 2016 Grade 2B acceptance standard
• Head at shutoff found to be around 32' (9.8m) vs. 126' (38.4m) expected

Looking at the pump running it was very quiet and visually the shaft appeared to slower that I'd expect to see. I asked for the following checks of the test setup:

1. Shaft speed - shaft at the mechanical seal was checked using a strobe and the technician confirmed the speed was the same as that recorded by the optical tachometer mounted by the motor shaft.
2. Direction of rotation was correct.
3. Flow meter readings were cross checked with a second flow meter and the piping traced to verify there was no possibility of flow bypassing the flow meter.
4. Suction and discharge pressure transducers were cross checked using bourdon tube gauges and confirmed to be reading correctly.
5. Full venting of the pump and test pipework was checked and confirmed. NPSHa was confirmed to be ample.
6. The impeller and casing were inspected and found to be within normal manufacturing tolerances. The impeller to casing clearances were within normal limits.

Upon further review, the HQ data was plotted out by my colleague Tom Leibner, who noted that it was off by a factor of 4x on head and 2x on flow. See below.

At this point the possible explanations were:

1. The impeller had some non-obvious hydraulic characteristic that results in an unexpectedly massive head drop when it is trimmed. We could find no historic test data for the 21" impeller, although there was past precedent for trims larger than this on other pump sizes. The impeller parameters had been reviewed and the next recourse would have been a lengthy CFD examination of the whole pump stage.
2. The test loop had some non-obvious set of circumstances that prevented the pump being tested successfully and those circumstances resulted in both head and flow being 4x/2x off. This was the first time this model of pump (or any slurry pump) had been tested on this loop and there was some speculation that this was the reason.
3. The pump was running too slowly, half speed to be exact - except of course the speed had just been verified by two independent measurements.

So at this point, valued reader, which of the above would you have picked ?

Occam Delivers

Well if you channeled the spirit of Occam, #3 of course :)

I asked that the shaft speed be checked again, this time I picked the location and applied the reflective tape myself. Lo and behold the shaft speed reading was half what had previously been reported. Further investigation confirmed that there were two (2) strips of reflecting tape attached to the motor shaft, 180 degrees apart. This resulted in the optical tachometer reporting that the shaft speed was twice what it actually was. It seems most likely that the test technician did not notice the existing tape and applied a second strip of tape during the test setup.

Some of you may ask "what about the strobe speed check done earlier, why didn't that detect the correct speed ?". Well it turned out that the technician had aimed the strobe at a split collar next to the mechanical seal. This split collar had - of course - two (2) splits, also resulting in a false speed reading.

Morals to the story ? For me they are:

1. Even in 2017 in a world filled with the Interweb, more computing power than we can imagine and electronic sensors for everything, Friar Occam is still relevant.
2. Sensors are only as good (or bad) as how they are setup and utilized.
3. Even after 30 years of being around centrifugal pumps (resulting in the world's most boring Resume), there is always something new to encounter.

Of course I'm sure some of my readers will draw different conclusions. If you did, please feel free to share them in the comments section below.

As always and until next time Beatus Centrifuga