Web tension control connects mechanics, drives, sensors and the operator interface. During winding, reel radii change, a measurement may be distorted and a command may arrive late. SLX ReelSync provides a fully simulated bench for observing these conditions in 3D, linked over Wi-Fi to a Raspberry Pi dashboard with an LVGL interface.
The ReelSync embedded dashboard project describes the system architecture. The PC calculates the line; the panel is its operator interface, not a second independent simulation.
Why reel radius changes the control problem
At a constant linear speed, a small reel turns faster than a large one. The kinematic relationship is v = ω · R: at 2 m/s, a 0.10 m radius requires approximately 191 rpm, while a 0.30 m radius requires about 64 rpm. A reference expressed only in rpm therefore cannot keep web speed constant throughout the run.
Torque also depends on radius. In an ideal static case, τ = F · R: 80 N corresponds to 8 N·m at a 0.10 m radius and 24 N·m at 0.30 m. Acceleration, inertia and friction add dynamic terms. Reduced available drive torque may therefore become apparent in some operating conditions before others.
Radius is also related to the amount of wound material. For an ideal, compact winding without voids, L = π · (R² − Rc²) / h, where Rc is the core radius and h is the effective material thickness. Deformation, actual thickness and winding conditions require application-specific characterisation on a real line.
A constant-speed, constant-force example
| Radius | Speed at 2 m/s | Ideal torque at 80 N | Ideal mechanical power, absolute value |
|---|---|---|---|
| 0.10 m | 191.0 rpm | 8 N·m | 160 W |
| 0.30 m | 63.7 rpm | 24 N·m | 160 W |
| 0.60 m | 31.8 rpm | 48 N·m | 160 W |
The product τ · ω = F · v remains 160 W in the ideal case. Its sign depends on the axis role and convention: an unwinder may brake while the rewinder pulls. These are not three electrical loads to add together, nor measurements of bench efficiency. Inertia, friction and drive conversion require their own terms in the balance.
Tension, cross-section and elasticity: 80 N does not always mean the same thing
Tensile force is measured in newtons; average material stress depends on cross-sectional area: σ = F / (b · h). On a film 400 mm wide and 25 µm thick, 80 N produces 8 MPa. The same force on a thicker material produces a different stress. In the linear elastic range, ε ≈ σ / E: with the bench PET profile's modulus of 4.9 GPa, the example corresponds to approximately 0.163% strain.
These numbers illustrate the relationships; they do not define a safe operating tension for every film. Temperature, humidity, extrusion direction, viscoelasticity and material quality can change the response. The ReelSync catalogue contains six profiles with documented properties and limits: PET, LDPE, BOPP, conditioned PA6 and two elastic reference materials.
Drives do not respond instantaneously
Requested torque does not become shaft torque immediately. The ReelSync virtual profile uses an equivalent first-order response: dτ/dt = (τrequested − τ) / Td, with Td = 20 ms in the declared profile. The dancer's equivalent force response uses 100 ms. These are virtual-machine parameters, not values identified on a commercial component.
The power limit adds another relationship: beyond the constant-torque region, available torque falls as speed increases. The virtual-line profile documents these limits. System-level representations with dynamics and torque/power envelopes are described, for example, in the MathWorks Mapped Motor documentation; this does not imply use of that product in the ReelSync core.
The dancer connects material storage, force and motion
The dancer changes the free web path and makes imbalance between line zones observable. The relationship between roller movement and accumulated length depends on path geometry and span angles. A simplified ratio valid for one arrangement should not be applied to every machine.
In the demonstrator, position, actuator force and tension belong to the same model. The actuator value is expressed in newtons, not bar. Converting it to pressure would require at least effective piston area, friction and circuit conditions: the software dial is not a pressure measurement from a real pneumatic installation.
The model dial indicates force in N. Pressure would require a pneumatic-circuit model and cannot be inferred from the displayed number alone.
A faulty sensor can still display a plausible number
Encoder loss may invalidate the measurement; a scale error can instead produce a smooth but incorrect speed. If model speed is 2 m/s and the scale factor is 1.25, the sensor reports 2.5 m/s: a difference of 0.5 m/s, or 25%. Comparing process, measurement and reference helps separate the hypotheses.
A load-cell bias shifts the reading; a frozen cell may hold it constant while the process changes; a delay adds a timing problem. Charts need consistent units, axes and observation intervals, without turning invalid data into an apparently correct zero.
A software view comparing process values with simulated measurements makes sensor errors observable. The displayed values belong to the virtual model.
Fault injection and diagnosis are different kinds of information
ReelSync can introduce nine fault families, including sensor errors, friction, drive derating, loss of actuator force and material breakage. The laboratory knows the injected cause; diagnostic logic must instead detect observable conditions and threshold crossings.
A scale error or bias may exist before an alarm is raised. One cause can have several consequences, such as speed disagreement, tension deviation and torque saturation. The HMI therefore distinguishes the test cause, active alarms and latched records of cleared alarms. Red, amber, text and symbols make that difference readable on the panel and in the 3D scene.
| Test | What to observe | Mistake to avoid |
|---|---|---|
| Encoder scale error | Comparison between model speed and sensor reading | Treating a smooth curve as an accurate measurement |
| Load-cell bias or freeze | Difference between process force and simulated measurement | Accepting a value simply because it is stable |
| Reduced available torque | Persistent saturation and tension response | Attributing every error solely to the regulator |
| Material breakage | Line state, visual material path and permitted recovery | Confusing alarm acknowledgement with a new run |
| Connection loss | Data age and suspension of remote commands | Equating Wi-Fi connectivity with valid control authority |
Raspberry Pi dashboard: when the network becomes part of operator interaction
The dashboard uses embedded Linux, Buildroot and LVGL on Raspberry Pi 3 Model A+. The simulator link uses an application contract over WebSocket and TLS 1.3: identity, pairing and assignment of control precede command submission. A router connection alone does not establish that the process is reachable or that the panel has control.
Requests are correlated with results and run context. Transport freshness uses real monotonic time, separate from simulation time, which can stop during a pause. Stale data suspends commands; reconnection resynchronises state and authority without automatically repeating previous operations.
The panel does not transfer a certified safety function over Wi-Fi. Machine integration requires specific assessment of the network, I/O, drives, interlocks and stop architecture. In ReelSync, the line remains virtual and the PC retains responsibility for the process.
Identifiers, deadlines and simulation time
In the ReelSync contract, run_id identifies the run and command identifiers are decimal strings, preserving 64-bit precision in JavaScript clients as well. expires_at_sim_s belongs to model time. Connection freshness uses real monotonic clocks: pausing an experiment does not make an old measurement fresh.
The demonstration transport uses a queue with a 500 ms expiry; the panel suspends commands when data is more than one second old. A command without a result within two seconds closes the session and requires the state to be checked. These values describe bench configuration, not guaranteed Wi-Fi latency. The telemetry and command contract explicitly separates these deadlines.
TLS 1.3 protects the channel, while WebSocket provides bidirectional message exchange. Neither replaces command-semantic validation, controller availability or verification of the effect on the process.
Preparing a repeatable test
A useful test declares material, geometry, references, initial conditions and expected results. The laboratory countdown allows a fault to activate after the overview has been restored: the PC shows the 3D component while the dashboard presents the process diagram and alarm page.
The timer uses model seconds: a pause freezes it and an accelerated run changes its relationship with real time. Removing the cause, waiting for recovery conditions and acknowledging recovery are separate steps. CSV and JSON recordings and replay support comparisons between runs with declared configuration and seed.
The two-zone elastic model and software checks make the bench useful for development and analysis. They do not describe every real-line behaviour: slip, plastic deformation, specific materials and functional safety require dedicated data and project work.
From a demonstrator to custom industrial HMI development
For machine builders and companies developing embedded products, the bench makes commands, measurements, faults and communication recovery verifiable before integration. Firmware, embedded Linux, industrial interfaces and protocol adapters can be developed around the equipment's requirements.
The Windows 1.5.0 download includes the executable, documentation and evaluation SDK. Core and 3D authoring sources are not distributed. For a dedicated bench or custom operator application, a technical enquiry to Silicon LogiX can specify the process, materials, signals, protocols and faults to validate.
Technical references
- ReelSync Raspberry Pi dashboard and architecture.
- Virtual-line profile, material properties, alarms and telemetry contracts.
- IETF RFC 9846: TLS 1.3 and RFC 6455: WebSocket.
- Buildroot manual, LVGL documentation and Raspberry Pi 3 Model A+.
Canonical source: Web tension control: simulated faults with SLX ReelSync.
Top comments (0)