Plotting Performance Points

[impeller1]

Good day,

A colleague of mind asked me to help him plot random points on a design curve for a pump. Now I helped with the equations etc, however we can't explain why the majority of the points are far above the curve. Attached you;ll see our results. All data is included and the curve is lower down on the page. The pump is a vertical turbine pump, 3 stage, rated speed 1760 rpm, 4422 usgpm @ 388 feet of head. The pump pumps into a Reverse Osmosis membrane system and from there, it travels in a 42" header and goes to a storage tank about 200 yards away. Can anyone shed some light as to why these points are so high on the curve? We've used the affinity laws to interpolate field values back to rated speed so it can be plotted on the curve. Theoretically, it looks as if either the speed is actually higher than what the field values tell or I;m carrying a larger impeller! I don't have the system curve, however you can the points distinctly follow the pattern of the system curve. Any inputs??

Thank You.

 

[TenPenny]

Where exactly were the discharge pressure readings taken?
Are those speed readings accurate?

If I follow your data, for example the Jan 1 data point, at 1400 rpm, the pump is generating more head than the curve would suggest at 1760 rpm?

Is that the correct pump curve? Make sure it's not a single stage curve.

 

[JJPellin]

Based on your data, I would suggest five possible errors.

First, the suction head entered for all conditions is 7 feet. If you are using the Net Positive Suction Head for this value, that is not correct. NPSH subtracts off vapor pressure. Also, it is in units of absolute pressure. Since the discharge pressure appears to be in units of psig, then the suction head also needs to be in the form of static pressure, gauge. NPSH for a pump of this type is often given at the centerline of the suction impeller.

Second, I do not normally adjust the data point head and flow using the affinity laws. I normally adjust the pump curve to the actually running speed and then plot the actual data. By adjusting both the head and flow of the raw data, I think you are double dipping. I am sure someone will correct me if I am wrong with this point.

Third, you are applying the affinity laws for a large change in speed. Generally, the affinity laws are considered to be accurate for changes of 5% or less in speed or impeller diameter.

Forth, the discharge pressure values look all wrong. For a centrifugal pump, when you reduce the flow, the head goes up. Your head values go down at lower flow. It looks like a result I would expect if the discharge pressure was taken downstream of the valve used to pinch back on the flow or downstream of the RO membrane.

Fifth, I am coming up with a specific gravity of 1.03 based on your data. Your spreadsheet did not include the formula, but that is what it calculates out to. Are you performing this calculation correctly? Did you multiply by the SG rather than dividing by it? Or, is your fluid really heavier than water.


Johnny Pellin

 

[impeller1]

JJPelin

1. From what I have read, Suction head and NPSH appear to be two different entities. Since I can't get a gauge down there, all we did was use the liquid water level above the impeller centreline, which is approx. 7 feet. Do i have to add atmmospheric pressure to that value?

2. You mean adjust the flow and head of the design curve to the the flow and head of the running speed and plot on that instead?

3. Well I didn't know this, but that's all the data I have to work with anyways.

4. This gauge is located at the pump discharge so I know it has to be an accurate reading. What I am not sure, and what it could be, is some characteristic is changed on the RO membranes creating an extra backpressure. That's the only way I can explain it as the pump discharge valve is not throttled.

5. Yes that's correct, the SG is 1.03 and I did divide by it.

 

[BigInch]

Instead of giving us a bunch of data that makes no sense, a spreadsheet with numbers that make no sense, no formulas on the spreadsheet, a pump curve far too low for the data, and not telling us what you are trying to do, why not just tell us what you want to do, then we could probably tell you how to do it right. You could check your method, then you could post your spreadsheet, with formulas, and I think some person here would probably take a look and see if it made any sense. Does that sound OK?

**********************
"The problem isn't working out the equation,
its finding the answer to the real question." BigInch

http://virtualpipeline.spaces.live.com/

 

[JJPellin]

Thank you for the additional information. I have a better idea what is going on. I will follow through my original points and then come back with a few other suggestions.

1. Based on your comment, I now know that this pump is hanging in a pit (or well) with a water level 7 feet above the first stage impeller. But, I still don’t know how far this water level is below the centerline of the discharge flange. This is where it can get complicated. The pump curve is based on differential head, but what is the reference? For an end suction horizontal pump, the suction pressure would be referenced to the suction flange and the discharge pressure would be referenced to the discharge flange. But, you have no suction flange. For a well pump, the differential head is calculated with both suction and discharge referenced to the center of the discharge flange. So, rather than taking discharge pressure and subtracting suction pressure, you would take discharge pressure and add suction lift. So, if the water level is 20 feet below the centerline of the discharge, the differential pressure would be discharge pressure plus 20 feet. Now, I am not sure if this is the same reference system used for all vertical turbine pumps. I am looking at a catalog for a Byron Jackson vertical well pump. You need to look at the manual for the particular pump you are working with. In any case, if I was correct and you were using the wrong reference, then the calculated differential head would be even higher and you would be plotting performance even further above the curve. So, this is not the answer.

2. Yes. I normally adjust the pump curve flow and head for the actual speed it is runing then plot the raw data on that curve. I use a spreadsheet program to adjust the curve. I will attach a link to my spreadsheet at the bottom. But, it should also work to adjust the head and flow using the affinity laws and plot it on the full-speed curve.

3. You are going beyond the intention of the affinity laws. But, once again, this cannot be your problem. If you were off the curve by 5% or 10%, this might explain it. But your data is not even in the correct pattern. It would be better if you had a family of curves for your pump for different speeds and started with the curve that was closest to the running speed. If you only have one curve for full speed, than the best you can do is exactly what you are doing.

4. This point is the key. If you are measuring the discharge pressure right at the discharge of the pump with no valve, no orifice, no strainer, between the measurement point and the discharge of the pump, then something is wrong. The incorrect pattern in your data cannot be described by an issue with the RO membrane. If there was something wrong going on downstream, the pump would still follow its curve. If the membrane plugged off, the flow would drop and the pump would be force back up on the curve and the pump discharge pressure would increase. I cannot see anyway to resolve this. I have to believe that one of the main sources of your data is wrong. You need to verify that you are getting accurate data for flow, speed and discharge pressure. Of those three, I suspect speed as the culprit. You didn’t mention the driver of this pump: engine, turbine or VSD on a motor. Why is the speed changing? Is there a control system that is varying the speed? What is the measurement used to input into the speed control system? I believe your tachometer is bad and the driver is running faster than you think and speeding up as the flow increases. If that is the case, then you really are measuring and plotting the system curve.

5. If the SG is above 1.0, then I guess this is something like brine, or dilute caustic, or amine, or some chemical. It would be better if we didn’t have to guess. But, since you told us it is going to an RO system, it must be some form of dirty water that you want to purify.

Get the information from the pump manufacture and make sure you are using their curve as intended with the correct references for the suction and discharge pressures. Recheck all of the data with the thought that your instruments could be bad. Check the speed with a strobe to verify that that tachometer is working right. Replace the pressure gauge. Calibrate the flow meter.


Johnny Pellin

 

[JJPellin]

I noticed something else in your data that looks strange. Before the speed you have a column labeled as Speed (%). This looks like the raw instrument data that you are using to calculate the speed. You did the calculation based on the belief that 0% = 0 rpm and 100% = 1760 rpm. I have never seen a speed signal that worked this way. If it is a turbine, then 0% would probably be minimum governor and 100% would probably be maximum governor. If it is an engine or a VSD, then the speed signal could be set to any range depending on the capability of the tachometer. It could be set so that 0% is 1000 rpm and 100% is 2000 rpm. If this is some sort of cascading control, then the percent value could be the percent of full scale on the measurement point or the percent output to the speed controller. If it is a complex control, the speed could actually be increasing when the percent signal decreases. An example could be a level control in the downstream tank cascading to a speed signal to a turbine. When the level goes down, you want the turbine to speed up and pump more water. We have one turbine in our plant with the governor configured for reverse signal such that zero signal results in full speed and maximum signal results in minimum speed.

I think you are calculating the speed incorrectly. This is my new favorite theory. But without information about the driver or the speed control scheme, it is a wild guess.


Johnny Pellin

 

[impeller1]

BigInch, I understand what you're saying..however it's a long and complicated story why I even have to attempt to do this plotting. In my opinion, it's a waste of time..but due to certain circumstances which are out of my control, it's all we have to work with..so please forgive me.

JJPelin,

Thanks again, a few further things;

1. I'm not sure for this particular pump how the manufacturer would calculate the suction head..however, i do acknowledge what you're saying and you're right..i would end up even further off the curve. So I'll have to check this.

2. Thanks for the spreadsheet..it will come in useful..however as you said in 3..all i can do is what i'm doing now..if I were to replot the curve, I'd have to do it for each point which would take me forever..

Finally..the driver is a motor equipped with a VFD for varying speed. As you said, I also believe that the speed I'm seeing is actually higher than what the pump is actually doing as I only get the raw data in the form of a percentage. I also did some further checking and it seems the motor is rated at 1783 rpm while the pump is at 1760. Also, from speaking to a few older heads on site, on top of that, the motor apparently is being oversped! So I would agree with you that what's probably causing it is that the speed I'm using is actually a lot higher than what I think it is. I also have no idea if it's a linear relationship with speed that the VFD operates on.

So I think its fair to say that my speed values are off..and I need to recheck my suction head values. However, i do believe the main issue is the speed. What do you think?

 

[JJPellin]

I agree that speed is the likely problem. But, it may be more complicated than I thought. Last night at home I reworked your spreadsheet with all of the common errors I could think off. I inverted the relationship between speed (%) and speed. I rebuilt the relationship with the speed (%) at zero resulting a speed greater than zero. I changed the range and ratio between these two values. And, nothing I could come up with made that data look like a pump curve. On closer examination, it should have been obvious that I would fail. There are three lines of data for one speed reading. For those three points, I should be able to see the proper relationship between flow and head. But, it doesn't work. Of the three points, the highest head corresponds to the highest flow and the lowest head corresponds to the lowest flow. This is not a pump curve, it is a system curve. So, the speed values are not just non-linear, inverted or ranged incorrectly. They are inaccurate.

Johnny Pellin

 

[impeller1]

Thanks JJPelin..I didn't realize you spent so much time behind it..and yes..the points aren't meant to plot a pump curve, they're just individual points meant to show where on the pump curve the pump was operating at that time..below you'll see months associated with each point..Well..i think I know what seems to be the problem now with your help..I'll do what I can with what I have and state all observations

 

[JJPellin]

I didn't pick my words well. I knew that this was not a pump curve. But, if they were accurate data points for the performance of a pump, they should take the shape of a pump curve. Even if the pump was degraded, a line plotted through the data points should still have the shape of a pump curve. Since they do not, they cannot be an accurate set of operating data points for a pump.

Johnny Pellin