The standard response to insufficient head for a centrifugal pump is to put a larger impeller in the pump, add pumps in parallel, or replace the pump with a larger one. However, less brute force solutions may do the trick.
Consider the product pipeline system from a small plant to a pipeline network (Figure 1).
Figure 1. The length of the pipeline means that small fluid changes have a big impact. (Click to enlarge.)
The pump handled either one of two feeds: a light stream (0.76 specific gravity) or a heavy one (0.91 specific gravity). The feeds were never mixed together. Due to downstream pipeline system constraints, the operation had to maintain a continuous feed rate of 540 gpm. The amount of time on each feed was varied as required.
Changes to the feed blending were being made upstream. The total average feed rate to the pipeline would slightly increase. The relative amounts of the products also would shift, to a greater amount of light feed. The most significant alteration was a boost in heavy feed viscosity from 11.4 cP to 14.1 cP.
Normally, relatively small changes in viscosity have only modest impacts in plant piping systems. However, two factors combined here to make the viscosity change extremely important. First, the pipeline length is 32 miles so, even small effects count for a lot. Second, the system has a drag reducing agent (DRA) added to cut pressure drop.
A DRA decreases pressure drop by making the laminar flow regime more stable. This extends the Reynolds number range for laminar flow. Laminar flow has much lower friction factors (and pressure drops) than turbulent flow. In laminar flow, pressure drop is proportional to viscosity. The viscosity change from 11.4 cP to 14.1 cP would increase head losses in the pipeline system by 24% in laminar flow.
The final analysis of the flow system was much more complex because DRA performance was tested on the new feeds and the exact DRA blend and concentration was changed to optimize overall performance. However, at the end of this, the pump still was significantly short of the head required.
Naturally, at this point, brute force solutions were discussed. The pump already had the maximum size impeller. Reducing the flow rate with parallel pumps has little impact, as moving back on the pump curve (Figure 2) only increases the pump head by 150 ft., which isnt enough.
Figure 2. Matching speed to the head requirements of the feed provides fast payback.
Replacing a large pump such as this is an expensive proposition. Setting the pump operation based on the heavy product also incurred a large operating cost. The light product, having a viscosity of 0.65 cP, required much less head. So, the existing installation for many years had been wasting power when pumping light product.
Pump speed is one other factor that can be modified to change pump performance. Pump head varies with the square of pump speed. The affinity law for total dynamic head (TDH) and pump speed (N) in rpm is:
This relationship holds when pump efficiency remains constant.
Pump speed can be varied using adjustable-speed or multiple-speed drivers. In adjustable-speed motors, electronic control enables rpm to be changed to any point within specific ranges. In contrast, multiple-speed motors can operate at more than one particular speed but arent continuously variable. For applications with two vastly different feeds, a multiple-speed drive often provides a very good fit.
However, here, the extra head required for the heavy feed didnt match up well with available standard speed combinations. So, an adjustable-speed driver was selected. It typically operates at either 3,550 or 3,700 rpm. The power savings for the time on the light product more than make up for the lower efficiency of the adjustable-speed motor system. The motor replacement would have had a modest, but justifiable, return simply as an energy-saving project. Factoring in avoided capital costs made the overall economics very attractive.