Today I am going to show you an example of an industrial application of a digital twin with co-simulation: a scale aqueduct that we built as a reference plant to demonstrate how Arvantti-Emulator®, Arvantti-DigitalTwin® and Arvantti-Fractal® work together. It is a demonstration plant with plausible estimated data, not a real installation, and I am going to tell you honestly what works today, what is in validation and what is coming next. Arvantti Labs is a brand of Sequoia Space.

This is the route. First, the challenge of operating a water network and the scale plant that reproduces it, with its service rule. Second, how it works: the emulator with virtual PLC, the 3D twin, the control room and the Arvantti-Fractal® studios. Third, what we verified in the tests. Fourth, what can an aqueduct operator earn. And fifth, the next steps. Each title is a link to its section.

Any aqueduct operator lives with four tensions. The pressure: if it is missing, the furthest or highest house is left without water; If there is excess, the network loses more due to leaks and wears out. Leaks: technical losses, fraud and illegal connections, that is, water that is pumped and not billed. Variable demand: consumption changes throughout the day, and the appropriate pressure in the early morning is not that of midday. And the pumping energy: each meter of pressure that is pumped extra is paid for and then dissipated at the regulating station. The daily question is how much pressure is enough, and no more. That is the problem that our plant reproduces at scale.

It is a demonstrative example. This is the plant: a scale test bench, mounted on a building, that reproduces a city aqueduct. Two multistage pumps in parallel lift water to at least 70 meters of water column, with peaks up to 80. Mca means meters of water column; 10 mca is approximately 1 bar. The regulating station lowers this pressure before entering the network, through an automatic or manual branch. The network is a four-level grid with 16 houses per level, 64 in total, each with its consumption valve, plus two leakage valves per level, eight in total, which represent technical losses, fraud and illegal connections. Between levels there are regulation valves. And the water recirculates: it falls into gutters, goes down to the return tanks and the recirculation pumps raise it back to the pumping tank. Important: This is a reference plant with plausible estimated data, not a real installation. The View in 3D link opens the floor plan at digitaltwin.arvantti.online; It is a good idea to check that it is open before the meeting.

The service rule is the same that a real water company would apply. It is not only about the pressure at the outlet of the regulating station, but about the most critical house in the entire network, normally the furthest or the highest: it should never drop below 15 meters of water column, plus or minus 1.5 bar. The automatic control points to 15.45 mca, that is, a 3% margin over the minimum. The pumping delivers at least 70 mca and the regulating station lowers what is left over. There are graded alerts in the critical house at 20, 18 and 15 mca, and the display distinguishes target, warning, alarm and minimum violation. With this the demonstration shows pressure management: less excess pressure, fewer leaks, without leaving anyone without service.

This is how co-simulation works, with three products, all registered trademarks of Arvantti Labs. First, Arvantti-Emulator®: calculates the hydraulics of the network and produces the signals that a real PLC would receive; The virtual PLC executes the 24-hour plan and the permitted controls. We use it when the plant is still a project or when it is necessary to emulate field signals. Second, Arvantti-DigitalTwin®: the 3D plant moves with the PLC data, saves the history and shows the simulated future separately, without mixing it with the operation. Third, Arvantti-Fractal®: its own independent hydraulic calculator, which does not use the emulator; reads PLC signals in read only, projects 24 hours, compares current plan with alternatives and recommends. The decision is up to the operator: accept or reject, and only then does the PLC apply the adjustment. Arvantti-Fractal® never writes directly to the PLC. In this demonstration the PLC is virtualized; the same schematic would connect to a real PLC in read only.

Each daily scenario is described with 24 time points, and the emulator moves from one point to the next gradually and continuously, without instantaneous steps. The example in the graph is the initial progressive scenario: the houses start at 30% open, rise to 100% at 12:00 and go down again; Minimum, maximum and peak time are configurable. The demand is edited in three ways: global, with the same opening for all 64 houses; per level, with an opening for each of the four levels; or per house, with an individual opening. There is also a seed-reproducible random demand mode, which never touches leaks, and versioned operating templates in addition to proprietary scenarios. The graph is illustrative of the profile shape, not a measurement.

This is the aqueduct control room, powered by the Arvantti-Emulator® virtual PLC. On a single screen you see the pressure and flow profile from the pump outlet, through the regulating station and each level, to the critical valve; the levels, inlets and outlets of the tanks; the pumps with their flow rate, drive speed, pressures and power; the regulating station in automatic or manual; the 64 houses with opening, pressure and flow, with the most critical highlighted live; and alarms, events and 24-hour trends. All data and control is marked as simulated: there is no connection or writing to a physical PLC here. The same scheme is intended to connect to a real PLC in read only.

Arvantti-Fractal® carries out future studies based on what the PLC publishes. The base projection answers what happens in the next 24 hours if the PLC plan is executed as is. The alternative scenario answers what happens with a hypothetical change in the operator, for example in demand or in the regulating station, and is compared with the base. The future is shown in 3D as a separate view, translucent and labeled as a simulated future, not applied, so that it is never confused with the operation. The twin also saves history: you can review past hours without pausing or rewinding the PLC. The clock runs by default at ×240: a simulated hour every 15 real seconds, a day in six minutes. Optimization, which would propose the best plan that meets the service rule, is the next step and today it is turned off on purpose, until the forecast is validated. No study changes the operation on its own.

To trust Arvantti-Fractal® we compare it against the emulator, which acts as reality. On the left, pressure: In a test scenario with the station on automatic and a target of 18 mca on the furthest house in level 4, we checked all 64 houses every second for 24 hours. None fell below 15 mca; the minimum was 17.52. Fractal also evaluated 72 cases, 24 hours a day with demand at 90, 100 and 110%, and all were over 15 mca. The floor pressure curves of both models differ by less than one hundredth of a mca. This view zooms in on the difference between real operation, co-simulation, and optimization. On the right, tanks: this view allows you to visualize the discrepancies between the plan and the actual operation. In the continuous 24-hour run, the level of the supply tank separated up to 34.27 percentage points at hour 18. There is no lack of water: the total volume of the three tanks is almost the same in both models; What changes is the moment in which it starts and stops recirculation. The graph on the right is for the critical point scenario, with a lag of the same type. They are model results, not measurements of a physical installation.

What can an aqueduct operator earn? I'm talking about potential, not measured results. First, less excess pressure: by regulating through the most critical house and not through the station outlet, the pressure can be lowered when demand drops, with fewer leaks and less pumping energy, without leaving anyone without service. Second, anticipate: see the next 24 hours before they happen. Third, decide with data: compare the current plan with alternatives before moving anything. And fourth, without risk to the operation: the connection is read-only and the decision is always up to your team. As a reference, in the scale plant model, the automatic branch of the station reduced leaks by 6.4% compared to the manual branch with the fixed output of the day. It is a simulation result on estimated data, not a measurement; The impact on your network is estimated in a diagnosis.

These are the next steps. One: validate the tanks, repeating the 24-hour run with the stable PLC sample and closing the difference in levels. Two: achieve ×240 speed live with full physics; today the entire day at that speed is played precalculated. Three: connect a real PLC in read-only, mapping your signals to the model. And four: the launch in a sector of your network, where we validate the forecast with real data and only then turn on the Arvantti-Fractal® optimization. If your network or your plant is still a project, we start with Arvantti-Emulator® and its virtual PLC simulation. All are registered trademarks of Arvantti Labs.

We close with an open question: what sector of your network do you want to see before it happens? Based on your answer, we propose the scope of the first step with Arvantti-Emulator®, Arvantti-DigitalTwin® and Arvantti-Fractal®. My contact information is on the screen: Andrés Quiceno, CTO of Engineering, [email protected], and the LinkedIn sites and pages of Arvantti Labs and Sequoia Space.