Book a demo

03 / 07 — CIE ProAvailable · free trial

Model based calibration software.

Response models from test data, so trade-offs are explored before the dyno.

Spend fewer dyno hours per calibration using automated Design of Experiments (DoE), Physics-Informed Neural Networks (PINN), Pareto multi-objective optimization, and constrained Bayesian modeling.

Start a free trial See pricing

In pilot with Tier-1 powertrain teams

CIE Prodoe_spark_cam.cie Live
↑ Cam phaseSpark advance →
Optimum spark 29.3° · cam 17° · torque 100.0% feasible
Torque iso-linesKnock limitConstrained optimum
DoE · Gaussian process · constrained Bayesian optimum
DoE engine
D-Optimal, Latin Hypercube, Sobol, Box-Behnken
Model architectures
PINN, Gaussian Process, AutoML (Optuna)
Optimization
Constrained Bayesian + NSGA-II
Export
A2L (ASAP2), CDF, HEX — INCA & CAMEO ready

Capabilities

What CIE Pro does.

  1. B1

    Physics-Informed Modeling

    Combine empirical dyno response data with thermodynamics and electric motor thermal limits to prevent impossible calibrations.

  2. B2

    Multi-Objective Pareto Fronts

    Visualize the exact frontier where emissions, fuel efficiency, drivability, and peak torque compete, making tradeoffs deliberate.

  3. B3

    Constrained Bayesian Optimization

    Optimize engine maps and inverter modulation while strictly enforcing peak cylinder pressure, turbo overspeed, and knock thresholds.

  4. B4

    Instant Dataset Export

    Export verified calibration maps directly to A2L, DCM, CDFX, and hex patch sets ready for ECU flashing.

MATHEMATICAL FORMULATION

Surrogate modeling combining thermodynamic laws with dyno sweeps.

Physics-Informed Neural Network (PINN) Loss Formulation
L_total = w_data • MSE(y_meas, y_pred) + w_phys • ||dE/dt - (Q_in - W_ind - Q_loss)||^2 + w_bound • ReLU(P_max - 180 bar)^2

CIE Pro enforces hard thermodynamic boundary conditions during gradient descent. By penalizing violations of the First Law of Thermodynamics and mechanical overpressure limits (180 bar cylinder pressure, 220,000 RPM turbocharger shaft speed), the neural surrogate cannot propose physically destructive calibrations.

Where the test effort goes

Model-based calibration replaces physical test sweeps with surrogate models trained on a smaller set of measurements. The effect shows up in three places — how much dyno time a programme needs, how fast the optimizer converges, and how much physical validation is left before sign-off:

Where it shows up With CIE Pro What makes it possible
Dyno time 30–50% less Space-filling DoE — D-Optimal, Latin Hypercube, Sobol, and Box-Behnken designs capture maximum system variance per test point.
Optimizer convergence 50–70% faster Constraint-aware optimizers — Constrained Bayesian Optimization and NSGA-II — reach the Pareto front without wandering into infeasible regions.
Physical validation runs 80% reduction Physics-Informed Neural Networks encode known engine physics as constraints, so models generalize accurately even on sparse data.
Extrapolation safety Bounded by PINN physics loss Peak cylinder pressure, knock, and turbo overspeed limits are enforced as hard constraints rather than checked after the fact.

Workflow

From input to decision.

  1. 01

    Design DoE

    Generate space-filling test matrices to capture the maximum system variance with minimal dyno runtime.

  2. 02

    Train Models

    Fit Gaussian processes and neural surrogate models with live cross-validation and constraint boundary checks.

  3. 03

    Explore & Flash

    Scrub multi-dimensional Pareto surfaces, pick the optimal tradeoff point, and export validated calibration datasets.

Why it exists

The response surface has mass and consequence.

Calibration is not guess-and-check interpolation. CIE Pro turns complex multi-variable interactions into visual, navigable mathematical surfaces with hard boundary guarantees, allowing powertrain teams to deliver rock-solid calibrations in a fraction of the time.

Constraint-aware optimizers reach a feasible optimum in fewer test points, and exported maps feed straight into HELIX with variant tracking and audit trails.

Questions

Asked by engineers, answered plainly.

Does model based calibration reduce dyno time?

It reduces the number of measured points needed to characterise a response, because a space-filling design does not re-measure regions the model already predicts well. How much that saves depends on the number of factors, the smoothness of the response and the accuracy the decision needs, so a single percentage quoted across every programme would not be meaningful.

What is design of experiments engine calibration, and how is Gaussian process engine calibration used?

Model-based calibration replaces physical test sweeps with mathematical surrogate models trained on a smaller set of measurements. CIE Pro uses Physics-Informed Neural Networks (PINNs) and Gaussian Process regression to build these models, encoding known engine physics as constraints so the model generalizes accurately even with sparse data.

Does CIE Pro support Bayesian optimisation ECU calibration?

CIE Pro supports Constrained Bayesian Optimization (CBO), NSGA-II multi-objective Pareto optimization, and AutoML-driven hyperparameter search via Optuna with 5-fold cross-validation. Design of Experiments modes include D-Optimal, Latin Hypercube Sampling, Sobol sequences, and Box-Behnken designs.

What calibration file formats can CIE Pro export?

CIE Pro exports calibration maps and data in A2L (ASAP2), CDF (Calibration Data Format), and HEX formats. These are directly compatible with ETAS INCA, AVL CAMEO, and other standard ECU calibration tools used in OEM and Tier-1 workflows.

Does CIE Pro integrate with HELIX for team calibration workflows?

Yes. CIE Pro is designed to feed into HELIX, VehicleLab's enterprise calibration data management platform. Calibration maps generated in CIE Pro can be ingested into HELIX with full variant tracking, audit trails, and CRDT-based real-time collaboration for distributed engineering teams.

Next step

Experience the future of calibration.

Fewer dyno hours, physics-constrained models, and a Pareto front you can actually navigate — then export straight into the tools your team already runs.

Start a free trial Talk to an engineer

By task

Where CIE Pro is used.

Go deeper

Guides and comparisons for CIE Pro.

Works with