Digital Twins in the Real World

Digital twins have been talked about for several years, but the complexity of turbine and compressor operations has limited their real-world applications. But that is changing. Most of the large OEMs report implementations or at least large-scale trials. And a variety of software, IT, and industrial vendors have leaped into the arena to add their digitization expertise to digital-twin initiatives. This includes Ansys, Nvidia, Emerson, Vital Technology Services (VTS), Expert Microsystems, Siemens, and AnyLogic.

What is a Digital Twin?

There are various opinions of what a digital twin is, but the basic definition is relatively simple.

"A digital twin is a set of computer-generated models that map a physical object onto a virtual space with the most up-to-date data," said Anastasiya Malinovskaya, Content Manager at simulation and digital twin software provider AnyLogic.

A digital twin, then, is a digital representation of an intended or actual real-world physical product, system, or process. It can be seen as a digital counterpart of equipment or an entire facility that can be used for simulation, integration, testing, monitoring, and maintenance that accurately reflects the equipment or facility's ongoing operations. They can be used to simulate planned changes to see how the turbine will perform prior to the physical work being done. They can also be used to stay on top of performance and maintenance operations in real time. A multitude of sensors provide data on all aspects of operations, so the twin matches the equipment it mirrors.

Rainer Kurz, Manager of Gas Compression Engineering at Solar Turbines, divided digital twins into two categories: physics-based models that focus on thermodynamic and aerodynamic parameters; and digital twins that attempt to monitor vibration, temperatures, operating range, and other operational parameters. The former category is used by OEMs to simulate new designs. The latter category is directly related to operational decisions and thus demands that those running turbines trust the conclusions of the digital twin.

"The idea is to increase the reliability and the availability of machinery," Kurz said. "Physics-based models are a lot easier to implement than big-data models."

Real World Examples

Probably the best example of a digital twin in the wild is at a power plant on the Ford campus in Dearborn, Michigan. A decade-long modernization project includes office complexes, research and development facilities, and the Dearborn Central Energy Plant (CEP). Operated by DTE Energy, the CEP generates electricity and distributes chilled water, hot water, and steam to nearby buildings. The CEP contains a combined-heat-and-power (CHP) plant and a chilled/hot water plant for power generation and steam production. Digital twin and other technologies make it possible for one person per shift to manage both maintenance and operations.

"A single person looks after the entire 34-MW 87,000-square-foot facility 76% of the time," said Kevin Siess, Regional Operations Manager at DTE Energy. "Our plant operators are also our maintenance staff who can monitor the plant and its control systems while they are doing their rounds."

Digital-twin technology is courtesy of the VTS HardHAT system. As well as a digital twin, it provides 3D modeling, inventory, and computerized maintenance monitoring systems (CMMS). It is supported and integrated with SureSense asset performance management (APM) software by Expert Microsystems. In addition, Solar Turbine's Turbotronic control system is used for sequencing, control, and protection of the gas turbine package and monitoring of associated auxiliary systems. A Rockwell Automation historian and Allen Bradly PlantPAx distributed control system (DCS) are used to control the entire CEP.

Collectively, that adds up to about 12,000 datapoints that are sent to the historian. All of it is available in VTS. Trip alerts are automatically generated on key equipment running in the CEP, such as two 14.5-MW Solar Titan 130 gas turbines, a 5-MW condensing steam turbine from Siemens Energy, heat recovery steam generators (HRSGs) from Rentech Boiler Systems, and feedwater pumps.

Digital-twin technology aids operators in monitoring all aspects of running and maintaining two HRSGs. For example, the high-fired waterwall, O-type HRSGs from Rentech are designed to provide up to 225 kilo-pounds per hour (KPPH) of 200 psig saturated steam.

"Pattern management and 3D modeling enable operators to keep a close eye on hot water and high-pressure steam metrics," said Siess.

Operators can view a digital twin of the HRSGs and drill down into systems and components as needed. The CMMS pushes out upcoming maintenance actions. Parts are flagged that need maintained or replaced based on usage. A recent incident involved power loss to the HRSG. Alerts were generated immediately. Investigation detected a power supply drop off to the programmable logic controller (PLC) that tripped the HRSG. It was resolved within 90 minutes.

The Rentech HRSGs include supplemental firing. By raising the gas turbine exhaust temperature via supplemental firing, steam production can be greatly increased. Operators control supplemental firing from the booth and monitor temperature, pressure, and steam output numbers.

"This kind of automation is increasingly needed as the power generation labor pool is rapidly shrinking," noted Siess.

After a recent scheduled outage, the SureSense system generated an alert about the oil temperature exiting the shaft bearings rising from 219^oF by 10^oF. While still 40^oF below the alarm level, the software flagged the condition as abnormal. Solar Turbines has scheduled a seal replacement during the next scheduled outage unless the problem worsens.

"A sudden shift in oil temperature can lead to more varnish potential and make the unit and ancillary cooling equipment work harder," said Siess. "It is vital that we catch issues at an early stage to prevent a major failure."

Another example of the value of the digital twin concerned a problematic gas compressor. Drilling into data within the HardHat system, the operators discovered a slide gate out of calibration that caused unnecessary gas recirculation that had increased parasitic load and system wear. It was repaired before serious problems arose.

Siess reported only two trips in 2022, leading to a total of four hours of unplanned downtime for the year. Scheduled maintenance takes place in spring and fall. Each gas turbine and HRSG is taken down for two to three days. While one unit is being serviced, the other continues to operate to ensure the flow of steam to Ford is never interrupted.

Features built into the Rentech HRSG provide further assurance of steam flow. A fresh air firing capability utilizes a forced draft fan so the HRSG can be operated when the gas turbine is offline. The fan draws in ambient air and the duct burner provides heat to generate 80 KPPH of steam. A louvre-type diverter valve installed between the HRSG and the turbine facilitates steam/electricity switching. If there is no current steam demand, flue gas from the turbine can be diverted up a bypass stack instead of through the HRSG, generating electricity and not steam.

Siess added that automation enables DTE to untether the operator from the chair to go on maintenance rounds for 45 minutes or more without anyone in the control room. Data on plant operation is available via mobile device, which brings the control room into the field. Alerts are issued if anything needs attention.

"The right technology gives people enough confidence in automation so they can walk away," said Siess. "Our operations staff is now comfortably performing 80% of all maintenance activities."

Digital Twin in Turbo-Related Functions

Few have gone as far as DTE in trusting their digital twins to run plant operations in real time, but there are other examples of digital twins being used for various turbomachinery-related functions:

Aluminum Bahrain (Alba) boasts the world's largest single-site aluminum smelter. Powering these smelters requires 800 MW to heat the metal to nearly 1,000^o C. During product evaluation for this new facility, GE's HA-turbine lacked the operating hours to be approved for the project. Alba voiced concern about how the turbines might impact the Bahrain grid if they suffered an unscheduled outage.

GE Digital applied its Predix software platform to run a digital twin of the HA using data from the first HA running at a power plant in Bouchain, France, as well as numerous simulations to show how the turbines would react to the different situations. This demonstrated that if the turbines shut down, the effect on Bahrain's national grid should be minimal. Alba moved ahead with an order for three 9HA turbines as well as three GE steam turbines and three HRSGs.

"The digital twin helped us show that the HA should continue to operate as planned, even in the most challenging conditions," said Khalid Salem, a manager at GE Digital.

Siemens Energy is active in digital-twin technology in applications like transformers and gearing. There was also a pilot project running a few years ago using the AnyLogic Agent-based Turbine Operations & Maintenance (ATOM) model. This digital-twin technology was implemented to address operational disparities between Siemens aeroderivative units and a unit it acquired from Rolls-Royce (now known as the SGT-A65). The ATOM digital-twin approach proved more effective than the Excel-based forecasting tools previously used to monitor the SGT-A65. ATOM simulated turbine conditions, temperature, engine characteristics, failure modes, and maintenance operations across the fleet; identified bottlenecks; and supplied the simulations needed to determine better investment decisions related to upgrades and turbine replacements.

Emerson and Mitsubishi Power are working together on digital-twin technologies to optimize performance and reliability and enable predictive and AI-driven maintenance strategies. They are collaborating on solutions using Mitsubishi Power's Tomoni digital platform and Emerson's Ovation automation includes an embedded digital twin. The collaboration is aimed at enhancing the performance and reliability of power plants operating with Mitsubishi Power gas and steam turbines.

"The simulation solution will seamlessly receive data and operate in parallel with the plant's integrated control systems and other enterprise platforms to support commissioning and training," said Marco Sanchez, the former Vice President of Intelligent Solutions at Mitsubishi Power.

Beyond the big turbine and control OEMs, Ansys helps organizations across many verticals to build their own digital twins. Ansys Twin Builder software combines physics-based data with analytics. The company reports that one utility used its model to predict turbine stresses, saving more than $100,000 in a year by avoiding unplanned downtime.

Siemens, too, (as distinct from Siemens Energy) offers digital-twin software that has been implemented in aerospace, drones, fusion energy, electronics, and semi-conductors. The closest to a turbomachinery application is a project for SANY Heavy Energy. This real-time digital twin is said to have increased energy efficiency and lowered the levelized cost of electricity by 10%, improved wind turbine efficiency and wind farm design by 50%, and reduced maintenance costs.

Solar Turbines also uses digital models extensively, both in the development and improvement of existing products, as well as in its Insight condition monitoring system.

"The former involves detail, components and full engine simulations, supported and augmented by test data, while the latter provides a framework to interpret and act upon operational data," said Kurz.

Elliott Group currently uses advanced multi-physics based digital twins in its controls and fleet-monitoring systems. These digital twins for compressor performance, steam turbines, and vibrations rely on accurate modeling of the function of the turbomachine rather than big data and AI.

"They have been proven to be reliable and accurate," said Klaus Brun, Director of R&D at Elliott Group. "Nonetheless, for applications where significant test data is available either from factory or field operations, such as for bearings and seals, AI is being currently explored for utilization of subsystem digital twins."

Digital Twins and Compressors

Digital twins, then, have many applications. Brun believes AI and statistics-based digital twins add more value with gas and steam turbines that have hundreds and sometimes thousands of similar turbines operating around the globe. Those managing large fleets and OEMs monitoring their own units in the field can gather massive amounts of operational parameters and use the data to improve performance, lower emissions, spot degradation, and avoid unplanned downtime.

But data-based digital twins don't work well in every situation. The compressor market, in particular, is so geared to one-of-a-kind manufacturing that digital twins have limited applicability.

"Even if you have 1,000 compressors out there, each one of them is going to have different impellers and bearings inside and be tailored to a very specific application," said Brun. "As each one behaves differently, you can't correlate data from one machine to another no matter how big your data set is."

Author: Drew Robb, former editor of Turbomachinery International, is a freelance writer specializing in engineering and technology. Email Drew at drew@robbeditorial.com.

Check out the TurboTime podcast on applications of digital twins and pairing them with AI.