A Physics-Informed, Corrosion-Aware Digital-Twin Framework for Condition Monitoring of Saline-Service Centrifugal Pumps

Imagine a massive pump working underwater, moving seawater through the pipes of an oil rig or a ship. It is a hard-working machine, but the salt in the water is a slow, silent enemy. Over time, the salt eats away at the metal, the vibration of the spinning parts changes as the metal weakens, and the pump becomes less efficient at moving water. For engineers, watching these machines is a challenge because these three problems—rust, shaking, and slowing down—happen at the same time but look different on a monitor. If you only watch the vibration, you might miss the rust. If you only watch the speed, you might miss the shaking. The goal is to see all three at once to know exactly how sick the machine is before it breaks.

This is the problem Nsini Ignatius Udo set out to solve in a new study. The researcher took a method they had developed earlier, which combined measurements of rust, vibration, and efficiency into a single score called the Pump Degradation Index, and tried to fit it into a "digital twin." A digital twin is simply a computer model that acts like a virtual copy of a real machine. It takes data from sensors on the real pump and compares it to what the computer model expects to happen. If the real pump starts to act differently than the model, the system knows something is wrong. The challenge is that building a perfect computer copy of a pump is incredibly difficult, especially for old equipment on offshore rigs where data is scarce. Udo's work proposes a simpler, lighter version of this digital twin that focuses on the specific damage caused by salt water, rather than trying to simulate every single physical detail of the machine.

The researcher built a framework with five layers, moving from the physical pump up to a decision-making system. First, the system gathers data from the real pump, such as how much water is flowing, how fast the shaft is spinning, and how much the pump is vibrating. It also looks at the chemistry of the water and the temperature. Next, the system cleans this data and figures out exactly what the pump is doing at that moment. This is crucial because a pump might vibrate more simply because it is working harder, not because it is broken. The system separates these normal changes from actual damage. Then, the computer model predicts what the pump should be doing based on those conditions. By comparing the prediction to the real sensor readings, the system calculates a "residual," which is just the difference between what was expected and what actually happened.

The heart of this new system is the Pump Degradation Index, which acts as a transparent window into the machine's health. Instead of hiding the data inside a complex, unexplainable computer score, this index keeps the three main problems separate but visible. It tracks the corrosion, the vibration, and the loss of efficiency as three distinct numbers that are then combined into one overall score. In the study, the researcher tested this idea using data from a previous experiment where a stainless-steel pump was run in salty water for 200 hours. The data showed the pump slowly getting worse over time. The digital twin framework successfully recreated this path, showing the degradation score rising from zero at the start to a value of one at the end of the 200 hours. The computer model was able to predict the pump's head, efficiency, and vibration levels with very small errors, proving that the mathematical structure works.

However, the researcher is very clear about what this study has and has not achieved. The work is a proof of concept, meaning it shows the idea works on paper and with a small set of existing numbers, but it is not yet a fully tested tool for the real world. The study did not use new field data from actual oil rigs, nor did it include records of pumps actually breaking down. The corrosion data used in the test came from small metal samples called coupons, not from measuring the pump's own parts directly. Because of this, the researcher states that the system cannot yet predict exactly when a pump will fail or calculate how much useful life is left. It is a way to organize and track the signs of wear, not a crystal ball that tells the future.

The study also explicitly rules out the idea that this system is a replacement for traditional inspections. It does not claim to solve the mystery of why a pump is vibrating or rusting; it simply provides a better way to watch those things happen together. The researcher argues that for this system to be useful on an offshore rig, it needs to be tested on many different pumps, with real-time data and actual maintenance records. Until then, the digital twin remains a promising blueprint. It offers a way to keep corrosion, vibration, and efficiency in the same conversation, ensuring that engineers do not miss the slow, silent damage that salt water inflicts on the machines that keep the world moving. The next step is to take this lightweight framework out of the computer and into the field, where it can be tested against the messy, unpredictable reality of the ocean.