How-to
Testing HMI Simulation Against a Live PLC
Short answer
A standalone HMI simulator proves screens and navigation, not communications. Test against a live controller only through a documented online-simulation mode or a physical panel, and verify every read at two data states and every write at the controller before release.
On this page
- Can an HMI simulator connect to a live PLC?
- What does standalone simulation actually prove?
- What do I check before connecting anything live?
- How do I run the test against the live controller?
- What does a full acceptance test cover?
- Can a spare or damaged panel be the test endpoint?
- When is the HMI accepted?
- Questions asked
A standalone HMI simulator proves screens, not communications. Some HMI packages add a documented online-simulation mode that talks to a real controller — Delta exposes one at Tool -> On-line simulation (Ctrl + F4) — and Siemens pairs HMI simulation with PLCSIM over a virtual Ethernet adapter. Where your software documents a link like that, test the data path in simulation; where it does not, the live-controller acceptance test runs on a physical panel. The steps below cover both paths and the checks that decide acceptance.
Can an HMI simulator connect to a live PLC? #
Yes, but only through an online-simulation mode your software actually documents; offline simulation never touches the controller, so the first check is the help file, not the simulator window.
Two modes exist across vendors. Offline (standalone) simulation runs the project on the PC with substituted values. Online simulation runs the same project while the configured driver talks to a controller: Weintek documents online simulation against a connected PLC, Delta reaches it through Tool -> On-line simulation (Ctrl + F4), and in TIA Portal an HMI simulation can exchange data with PLCSIM, whose virtual Ethernet adapter answers a ping much like a physical CPU.
If your help system names no equivalent, stop there. A package that offers only standalone simulation cannot be turned into a live-PLC test by changing settings, and a shift spent trying alternate options is wasted. Note what the documented pairing replaces: the panel, not the controller. The machine's real I/O stays untested either way.
What does standalone simulation actually prove? #
It proves screen behavior only — layout, navigation, animations and macro logic against data you type in — and gives zero evidence about the PLC driver, addressing, cable or writes.
Offline simulation substitutes the tag's start value for controller data. Documented Siemens behavior makes the split plain: an I/O field given a start value displays it in simulation, while a tag tied to a global data block address does not track the controller. Sign off screens, text, navigation and graphics in the simulator; treat anything involving a tag as unverified until the live path runs it. A screen that animates perfectly in simulation has exercised none of what actually fails in the field: driver selection, station addressing, cable wiring and the write path.
What do I check before connecting anything live? #
Five things, in this order: the project's target panel model, the PLC driver and its addressing, the station or IP settings, the loaded controller program, and the machine's safe state before any write test.
- Confirm the project's target device equals the installed panel model exactly. A mismatch produces symptoms that look like a simulator limitation.
- Open the project's PLC connection and check the driver family, station number and tag addressing against the controller's actual configuration.
- For Ethernet, place the PC in the controller's subnet. Example: controller 192.168.1.10, PC 192.168.1.100, mask 255.255.255.0 — 192.168.1.10 AND 255.255.255.0 = 192.168.1.0 and 192.168.1.100 AND 255.255.255.0 = 192.168.1.0, so both sit on 192.168.1.0 and a ping should reply.
- For serial multidrop, give each device its own station number — the same rule that governs connecting multiple HMIs to one PLC over a 200 ft serial line.
- Confirm the controller holds the current program and is in a safe state before any write test.
Proceed only when all five pass; a symptom that survives these checks is real, not configuration noise.
How do I run the test against the live controller? #
Set the PC's address, ping the controller, start the logic, then launch online simulation or download to a physical panel, and confirm the driver reports connected before you exercise a single write command.
- Set the PC adapter as in step 3 above and ping the controller until it replies.
- Where the package supports a simulated controller, start it with the machine program loaded; PLCSIM's virtual adapter responds to pings the same way a physical ethernet board does.
- Launch online simulation from the documented menu, or download the project to the physical panel — sort out the HMI programming cable pinout first if the download cable is unfamiliar.
- Confirm the software reports the driver connected. The sequence for connecting an HMI panel to an S7-1500 PLC shows what a healthy end-to-end setup looks like if you need a reference shape.
- Watch read-only values first: displayed values must follow the controller, not the values you entered at design time.
A connection failure at step 4 is a configuration problem, not a screen problem. Fix it before continuing; testing writes over a half-connected driver proves nothing.
What does a full acceptance test cover? #
Every displayed tag verified at two data states, every write command verified at the controller, and every alarm forced on and cleared — full coverage, no sampling.
Before the first write, put the machine in a safe state: IEC 60204-1 expects the electrical supply isolated or a stop function actuated before intervention, so no command under test can move equipment.
The two data states are the bottom and top of each tag's scaled range, which is what exposes scaling and offset errors. Take an illustrative project: 12 screens, 46 displayed tags, 18 write commands, 9 alarms.
| Check | Items | Passes per item | Checks |
|---|---|---|---|
| Displayed tags | 46 | 2 data states | 92 |
| Write commands | 18 | 1 PLC-side verification | 18 |
| Alarms | 9 | 1 force-and-clear | 9 |
| Total | 119 |
The arithmetic: 46 × 2 = 92 read checks, 18 × 1 = 18 write checks, 9 × 1 = 9 alarm checks, and 92 + 18 + 9 = 119 checks total. At about 2 minutes per check, 119 × 2 = 238 minutes, roughly 4 hours. Each write check means forcing or observing the target address in the controller's programming tool and confirming the PLC-side result, never the screen response alone.
Book a half shift for a project this size. A test that gets squeezed between other jobs is exactly the one where writes get signed off on screen behavior instead of controller data.
Can a spare or damaged panel be the test endpoint? #
Yes — a spare works if it runs this project with a usable display, and a damaged panel works only after you confirm the screens and commands you must test are still visible and operable on that specific unit.
Panels with failed displays sometimes keep remote view and screen change alive, which is enough to drive a live test from a desk. That behavior belongs to the individual unit and its failure mode, not the model, so verify it on that panel before relying on it, and record the serial number that served as the endpoint. A spare with a good display is always the cleaner choice when one exists.
When is the HMI accepted? #
When the record shows all reads matched the controller at two states, all writes produced their PLC-side result, and all alarms appeared and cleared on the live path — anything tested only in simulation is retested first.
If the panel will not communicate, isolate in order: project communication settings, controller communication settings, physical link, controller status. Change one thing per attempt. If that sequence passes and data still does not move, stop and escalate to the vendor rather than editing the controller side. Never alter controller logic or machine wiring to work around a simulator limitation — that converts a test gap into an undocumented machine change, which is far more expensive than the test you skipped.
Questions asked
Can a simulated PLC drive a real panel instead?
Yes, several ecosystems pair a controller simulator with the physical HMI, which exercises the panel's driver, screens and writes. It still leaves the physical controller's wiring, I/O and scan behavior untested, so the first download to site hardware needs its own read-and-write pass before release.
Why verify each tag at two data states instead of one?
Two points, one at the bottom and one at the top of the scaled range, catch scaling and offset errors that a single mid-range reading hides. Add a third point near the alarm threshold when the tag feeds a comparison, because boundary behavior is where rounding and deadband problems first show up.
Is a successful ping enough to call the link good?
No. A ping reply proves IP reachability between the PC and the controller and nothing more; the driver, station and tag layers can still fail after it. The link passes only when the simulated or physical panel exchanges live data with the controller in both directions.
What should the acceptance record contain?
One line per check: tag or command, expected value, observed value, pass or fail, and the witness, plus the panel model, PLC model and software version used. A later failed retest is only diagnosable if the versions in the record match the hardware that ran the original test.
When should I stop troubleshooting and call the vendor?
Stop after configuration, addressing and ping checks all pass but data still does not move, or after two failed configuration passes. Bring the panel model, PLC model, software version, connection settings and the exact symptom; that set is what support asks for first.