Manufacturing platform · Calendering

Mechanical compression simulation with downstream consequences.

DEM resolves particle rearrangement, contact deformation, porosity reduction and spring-back. Bonded-particle extensions predict fracture. Time-dependent AI surrogates reproduce the full calendering trajectory in seconds rather than hours of DEM.

Book a calendering review →See the full platform

A coated electrode passing between polished steel calendering rollers

Physics of the nip

Contact forces, spring-back and particle fracture, resolved.

Mechanical compression is primarily simulated using the Discrete Element Method (DEM), which resolves particle rearrangement, contact deformation, porosity reduction and spring-back under uniaxial loading. The result is a 3D microstructure that carries the full history of the electrode (slurry, drying, compression) into downstream models.

Extended DEM formulations with bonded-particle models also predict particle fracture during aggressive calendering, a critical failure mode for brittle active materials such as high-nickel cathodes. The resulting microstructures give direct access to effective transport properties (tortuosity factor, electronic conductivity, active surface area) that feed the downstream electrochemical models.

The calendering method stack

From hours of DEM to seconds of inference.

01 / 04

DEM compression

Particle rearrangement, contact deformation, porosity reduction and spring-back under uniaxial loading. The reference physics for calendering simulation.

02 / 04

Bonded-particle fracture

Extended DEM formulations predict particle fracture during aggressive calendering. A critical failure mode for brittle high-nickel and silicon-bearing active materials.

03 / 04

Time-dependent DL surrogates

Trained on DEM-generated compression time-series. Deep-learning surrogates reproduce the full calendering trajectory including relaxation in seconds rather than hours, with mean squared errors below 2%.

04 / 04

Bayesian optimization

ML regression combined with Bayesian optimization navigates the calendering parameter space (compression degree, line speed, formulation) against multi-objective targets such as energy density versus power capability.

Calendering by the numbers

< 2%

MSE on full trajectory reproduction

seconds

surrogate inference vs. hours of DEM

microstructure piped to electrochemical solvers

in-line

mixed-reality monitoring demonstrated on roll

Beyond simulation

Real-time monitoring meets predictive control.

Mixed-reality-assisted approaches, coupling computer-vision detection of calender roll parameters with real-time ML inference, have been demonstrated for in-line predictive monitoring. The same surrogate that accelerates design-space exploration in the office runs at line speed in the plant, fed by camera measurements rather than offline inputs.

This closes the loop between calendering settings and downstream electrochemical performance. A setpoint change at the calender becomes a predicted change in capacity, rate capability and cycle life within seconds, not weeks of physical commissioning.

Common questions

What calendering teams ask.

Yes. The compression simulation operates on the full stack with appropriate boundary conditions at the foil interface and between layers. Spring-back, layer-to-layer adhesion and asymmetric compression effects all fall out of the simulation, not from post-hoc corrections.

For chemistries inside the training envelope, the surrogate predicts directly. For new formulations, a small number of DEM ground-truth runs fine-tune the surrogate. Typically a few dozen runs are enough to re-calibrate it, after which it produces calendering trajectories in seconds for the full parameter space.

Yes. The 3D microstructure exiting the calendering module is the direct input to the downstream electrochemical FEM solver. Tortuosity, porosity, active surface area and contact networks flow through automatically, so a calendering parameter change gets evaluated against capacity and rate capability targets, not just porosity.

Calendering module

Calibrate the nip to a capacity target, not a porosity one.

Share a target microstructure and a downstream performance objective. We'll return a calendering parameter window and a DEM-trained surrogate scope.

Book a calendering call →See cell assembly