Case Study: Mazda achieves faster SKYACTIV engine development and halved model complexity with MathWorks

Mazda Speeds Next-Generation Engine Development of SKYACTIV TECHNOLOGY

Mazda faced growing difficulty finding optimal calibration settings for its SKYACTIV engines—especially the SKYACTIV‑D diesel—because traditional spreadsheet-and-test‑cell methods were slow and unreliable in high‑dimensional search spaces. To reduce soot and NOx while meeting stringent European and Japanese emissions standards, Mazda needed ECU‑embeddable statistical models (initially ~38–40 parameters) that were both accurate and light enough to run on production ECUs. Mazda worked with MathWorks using MATLAB, Simulink, and the Model‑Based Calibration Toolbox to address these challenges.

MathWorks’ tools enabled Mazda to design DOE‑based test plans, build statistical response models, generate embeddable models (including a reduced maximum cylinder pressure model), and produce optimal calibrations via the toolbox’s CAGE tool and a MATLAB optimization interface. As a result, Mazda halved model complexity (reducing ~38–40‑parameter models to ~20), cut calibration workload and test‑cell time, and improved model accuracy (smoke model RMSE reduced by 80%), contributing to compliant SKYACTIV‑D production engines installed in vehicles such as the Mazda CX‑5.

Mazda

Shingo Harada

Assistant Manager