Figure 2. Identical pumps can actually differ by up to 10% in total dynamic head.
This is a common problem and was a factor here. While the difference wasnt 10%, it was enough to force the B pump product out of the intermediate draw nearly all the time when both units were connected. Additionally, small piping differences between the systems imposed a higher pressure drop on the B unit side, making the situation worse.
More importantly, the design suffers from a fundamental conceptual error. As product flow rates change at either unit, the one with less flow has a higher discharge pressure. With a higher discharge pressure available on one side of the combined header, more flow goes from that side of the plant to the intermediate product (seal oil).
The system as-built automatically pressure balanced to take more intermediate product from the unit that had the lower net product rate. Actual sensitivity depends upon system hydraulic balance. In this case, flow differences as small as 10% between each unit were enough to swing 100% of the seal oil draw from one unit to the other. Data showed product rates changed by more than 70% at times.
The solution was to add a second, independent control as well as to isolate the two systems upstream of the intermediate storage for seal oil (as indicated by the clouded area on Figure 1). After this change, the overall system was stable and operated very well. A bias controller between the two flow controllers (FC2 and FC2B) sets the ratio of seal oil coming from each unit. During occasional big shifts in product rates, the operators can reset this to rapidly get the system into control again.