Centrifugal Pump Wear
Background
When a pump experiences internal wear, its performance drops. The primary mechanisms for such wear are:
- Corrosion: This is a gradual destruction of metals by chemical or electrochemical reactions with their environment.
- Erosion: Mechanical abrasion due to suspended solids in the pumped liquid.
- Cavitation: Caused by the implosion of cavitation vapour bubbles leading to erosion.
The affected parts can include wear rings, shaft sleeves, packing, mechanical seal faces, lip seals, pump casting, and the impeller. Such conditions can be difficult to detect as the pump may be inaccessible or it may be just uneconomic to take the pump out of service for inspection.
The Solution
As can be seen from the pump curve, the relationship, for a fixed speed, between flowrate, differential pressure and power consumption is not a straightforward one. Should the pump have a variable speed drive then the relationship becomes even more complex.
Nevertheless, it is a problem that lends itself to data science. It is possible for an algorithm using historical data, such as suction and discharge pressures, combined with flowrate and power consumption, to benchmark the performance of the pump. Effectively you create an operating signature for the pump, which is assumed to represent efficient performance. This signature can be used to predict power consumption based on flow rates and pressures. 
To operationalise this system, it is necessary to feed the signature with real-time data, perhaps from an OPC server. This then compares the predicted power consumption, calculated by the signature, with the measured power consumption. Should the actual power consumption exceed the predicted value then an alarm is raised.
In practice, there are some complexities around pre-processing the data to ensure that the acceleration and deceleration phases are ignored and the measured parameters are represented as moving averages.