Optimizing Engine Testing: A Key Challenge_
Reducing vehicle development time while ensuring compliance with emissions regulations is a major challenge for the automotive industry. OSE has developed an innovative approach to optimize engine test plans.

A Strategic Optimization of Test Plans_
OSE worked with emissions measurement experts from a major automotive manufacturer to develop an algorithm capable of significantly reducing the duration of engine and vehicle tests, while preserving the richness of the information. These tests, essential for measuring pollutant emissions, are traditionally long and costly, requiring hundreds of hours on engine or chassis dynamometers with significant cooldown times between cycles.
The main goals were to:
Optimize the test plan to reduce duration without loss of information.
Feed an internal machine learning model with high-variance and sufficiently large data.
Save time during test campaigns.
Preserve crucial information in the training data.
The complexity lay in meeting strict technical constraints (cycle duration, soaking times, specific speed profiles) and ensuring exhaustive coverage of possible driving scenarios.

OSE's Role_
OSE brought its expertise in physical modeling, machine learning, and operational research. Our main missions were:
Building a pattern database: we extracted and analyzed patterns (segments between two zero-speed points) from existing vehicle cycles, integrating a variety of descriptors (temperature, engine speed, torque, velocity, slope).
Pattern analysis and classification: we used PCA (Principal Component Analysis) for dimensionality reduction and a clustering algorithm to group similar patterns. A key finding was the correlation between the model error and cumulative emissions, highlighting the importance of targeting high-emission patterns.
Development of a selection and reconstruction algorithm: we prioritized high-emission patterns while ensuring broad coverage of all pattern clusters and compliance with speed profile constraints. We used an iterative local search algorithm to assemble selected patterns into a final test plan that respects all operational constraints and maximizes the objective function.
This approach, unlike a brute-force solution, enabled the handling of a large pattern set (over 100,000) and the generation of optimized test plans within a reasonable time frame.
This collaboration combined the manufacturer's domain expertise with OSE's scientific capabilities to design a lean engine test plan that remains rich in meaningful data and fully compliant with operational constraints.

OSE Solution Benefits_
Thanks to our involvement, OSE was able to provide the automotive manufacturer with a robust methodology and operational guidance to optimize their test campaigns. Although still at a prototyping stage (lower TRL than an industrialized tool), the concrete benefits are:
Significant reduction in test time: our algorithm cut total testing time by about 20%, saving hundreds of hours.
Preservation of information and data diversity: the optimized test plan ensures the quality of data feeding the machine learning models.
