Connection of the pump with its digital twin. © Ansys/PTC

For this purpose, the pump in operation was equipped with pressure sensors at its inlet and outlet. In addition, acceleration sensors were installed on the pump and bearing housing to measure vibrations and flow meters were fitted on the discharge side. An actuator controlled the discharge valve, while the valve on the suction side was operated manually. The sensors and actuators were connected to a data acquisition device that recorded the data at 20 kHz and relayed it to a Hewlett-Packard Enterprise (HPE) IoT EL20 edge computer system. PTC's ThingWorx platform created an ecosystem to connect devices and sensors to the IoT, demonstrate the value of IoT data, develop enterprise-level IoT applications, and empower end users through augmented reality. ThingWorx was used as a gateway between the sensors and digital data - including the simulation model of the pump. A layer of machine learning in ThingWorx, running on the EL20, controlled the sensors and other devices and automatically learned the normal state of the pump in operation. Machine learning was also used to identify operational anomalies and generate insights and predictions. The ThingWork platform was also used to create a web application that displays sensor and control data as well as analytics. For example, the app indicated input and output pressure and predicted bearing life. An augmented reality front-end overlaid an image of the pump displayed on the user's smartphone, tablet or data glasses with sensor data and analytics, as well as parts lists, repair instructions and other personalized information.

A Flowserve pump. © Ansys/PTC

To demonstrate the added value of the digital twin, the pump was initially operated normally. An anomaly was initiated manually by closing the suction valve to 50 percent. The sensor readings showed that the suction pressure, discharge pressure and discharge volume decreased drastically, while the acceleration sensors vibrated strongly. Red warning messages were displayed, while predictive analysis predicted that the pump bearing life would drop to just a few days if this condition continued. However, it was not clear from the sensor readings and analysis how the irregular flow conditions were affecting the operation of the pump, i.e. why the pump was vibrating or what possible solutions could be considered. Under reduced inlet and outlet pressure and flow rate, the connected system model showed the same MQL values as the physical pump. A 3D simulation model, which was connected to the physical pump in the cloud, was activated via the "Simulate 3D" button. The team wanted to find out why the vibrations occurred and how the change in flow condition affected pump operation. The 3D simulation showed that the pressure drop inside the pump caused cavitation and vapor bubbles formed. In the higher pressure ranges of the pump, the bubbles imploded and generated vibrations.

Combining the real and virtual worlds

By separating the system model from the physical pump, several potential corrections could be tried out with the system model's MMS. For example, the system model had predicted that opening the inlet valve would solve the problem. To validate the potential solution, a second 3D simulation was performed on the offline system model with the valve open. The 3D results showed no vapor bubbles. The problem was thus solved by opening the inlet valve on the physical pump and normalizing the performance. The IoT allows simulation models to connect to products that are already in operation via a platform such as ThingWorx. This gives industrial companies the opportunity to better understand and optimize product performance. The digital twin enables companies to identify and isolate faults, carry out diagnostics and troubleshooting and suggest corrective measures. In addition, an ideal maintenance plan can be determined based on the specifics of the individual plant, plant operation can be optimized and insights can be generated that can help improve the next generation of products.

The advantages of digital twins are enormous. Short-term successes include maintenance optimization and operational troubleshooting. However, as soon as customers start demanding results and not just products, the digital twin has the potential to unlock this added value for both product manufacturers and their customers.

Chris MacDonald, Senior Director, Analytics GTM Strategy & Business Development, ThingWorx Analytics at PTC; Bernard Dion, Chief Technical Officer, Systems Business Unit at Ansys; Mohammad Davoudabadi, Principal Engineer at Ansys / ag