## Please discuss the issues regarding GPM calculations.

So let's start with the assumptions of the pump math in the RTU (from pump math 1.4.doc):

� The controller algorithm, however implemented, seeks to guarantee that over time each pump will pump equal volumes. (This is up to the designer of the lift station.)

� Station wells and pump GPM ratings are sized to accommodate maximum expected flow over the life of the station. Pumps are replaced/overhauled when performance drops to ~90% of their GPM rating. This would typically take about 10 years of operation.

� Actual flow (and hence actual cycle timing) depends on weather, season, time of day and population changes. Assumptions about the actual flow are:

1. Each pump encounters flows with identical statistics. (In fact, the pump math algorithm starts having problems when this condition is not met.)

2. Flow variations between pumps will average to zero over 250 cycles (~3 days). (Best GPM accuracy occurs when things are stable for a few days.)

� A typical pump cycle is ~15-30 minutes (pump cycle = inflow time + pump out time when it's one pump's turn). The minimum would be 3 minutes, the maximum about 10-12 hours. New stations and dry times tend to have longer cycle times than older/wetter ones.

� Pumps are sized so that they can completely drain a full well (with 0 additional inflow) in 15 to 240 seconds (_ - 4 minutes). Pumps sizes outside this range may have a loss of accuracy or precision in their reported efficiency. (Oversized pumps cause very short pump out times. Undersized pumps cause very long pump out times. Either will perturb the accuracy of the algorithm.)

Specifically, if the pump can drain the well in less than 15 seconds, the efficiency figure will peg at a value corresponding to the highest reportable 15 second drain rate. If the pump takes longer than 4 minutes (considered the more likely case), then there will be a gradual loss of significant figures in the reported figure. Until, if the pump takes longer than 8 hours the reported value becomes zero.

� Pumps may be forced on when the well is not actually full in order to perform maintenance, tests, or when the fill time has exceeded sepsis limits. (These may perturb the pump performance algorithm. Another period of stability is required before the algorithm can return to giving reliable GPM values.)

� Pumps are never run dry.

In general terms, one must remember that the pump algorithm is adaptive over a period of several days. Any big change, such as one pump failing or being off-line, can really upset the result. Low overall station flow conditions, exhibited by short pump on times, and therefore long inflow times, also perturb the algorithm. Customers should be cautioned that during some periods of upset in stability, GPM is not very well coordinated to pump on times. Furthermore, the algorithm may, at times, substitute GPM values that seem to have no relation to runtimes. But it does usually recover. In a small number of instances the pump math has needed to be restarted before the station statistics returned to reliable numbers.

During periods of instability, or afterwards while the pump algorithm is still working to get back to delivering reliable numbers, one should not compare old GPMs from a stable period to today's GPM for a similar run time and expect them to be the same.

Review data:

- Starts per day versus runtimes over time
- Starts per day versus weather
- Runtimes constant?
- Pump Controller installed?
- Pump load is balanced. Is there a problem with one pump?
- Starts increasing while runtimes staying the same?