CIE Pro Technical Brief

Full-factorial dynamometer mapping requires sweeping every combination of engine speed, load, cam phasing, and ignition timing. For modern electrified and turbocharged powertrains with multiple variable actuators, full-grid sweeps require excessive test cell hours.

Standard statistical curve fits frequently predict physically invalid operating states in sparse data regions. CIE Pro addresses this challenge by generating space-filling DoE plans that cover the operating envelope with fewer measurement points and embedding physical conservation laws directly into neural surrogate models.

A space-filling DoE covers the operating envelope with fewer points than a traditional rectangular grid because it avoids redundant measurements in regions where system behavior is linear or already accurately predicted by the surrogate model.

Supported DoE algorithms include D-Optimal, Latin Hypercube Sampling (LHS), Sobol quasi-random sequences, Central Composite, and Box-Behnken matrices. Dynamic DoE generates transient frequency sweeps with convex hull constraints to enforce operating boundaries and component thermal limits.

CIE Pro trains Physics-Informed Neural Networks (PINNs) and Gaussian Process regressions that incorporate thermodynamic and electromagnetic constraints. These constraints prevent impossible calibration states, such as negative fuel consumption or torque outputs exceeding physical magnetic saturation limits.

Automated hyperparameter optimization runs via Optuna with 5-fold cross-validation, ensuring models generalize reliably across the entire operating map.

Multi-objective calibration balances competing targets: peak torque, tailpipe emissions, fuel consumption, and drivability. CIE Pro uses constrained Bayesian optimization and NSGA-II to map the exact Pareto frontier.

Calibration engineers can scrub across multi-dimensional surfaces, evaluate trade-offs in real time, and lock optimal calibration setpoints with verified safety margin bounds.

Verified calibration maps export directly into standard ECU calibration formats: ASAM MCD-2 MC (A2L), ASAM CDF, and Intel HEX. These datasets import seamlessly into standard calibration environments including ETAS INCA and AVL CAMEO, or feed into HELIX for release management.

CIE Pro does not perform live online calibration. It builds surrogate models, optimizes response surfaces, and generates validated datasets; writing parameters to a live, running ECU on a test bench is handled by your measurement and calibration front-end.

We do not claim arbitrary percentage reductions in dyno time. Time savings depend on the number of control factors, actuator non-linearities, and required model confidence intervals.