Surrogate Model based NSGA-II for Suspension Optimization for Two-Wheelers

Every time a motorcycle hits a bump, the rider feels it. The machine's suspension system is the barrier between the rough road and the human body, designed to absorb the shock of the terrain so the rider remains comfortable and in control. Engineers have long known that finding the perfect balance for these springs and shock absorbers is incredibly difficult. The system must be soft enough to smooth out a jarring pothole but stiff enough to keep the bike stable when turning or braking. If the suspension is too soft, the bike bounces uncontrollably; if it is too stiff, every vibration travels straight to the rider's spine. Traditionally, finding this sweet spot has been a slow, expensive process of trial and error, requiring thousands of computer simulations that take hours to run just to test a single idea.

A researcher at Yatri Motorcycles in Kathmandu, Nepal, has developed a new way to solve this problem that is both faster and more precise. Instead of running those slow, heavy simulations for every single test, they built a mathematical shortcut. They created a simplified digital model of a motorcycle's front and rear suspension, treating the bike as a "half-car" system that captures how the front and back wheels move together. They then used a technique called surrogate modeling, which is essentially training a smart computer program to learn the behavior of the full, complex system based on a smaller set of data. Once this program learned the rules, it could predict how the bike would ride with near-perfect accuracy in a fraction of a second, allowing the researcher to test thousands of combinations instantly.

The researcher used this fast, predictive model to run a sophisticated search algorithm known as NSGA-II. This algorithm acts like a tireless explorer, sifting through millions of possible settings for the front and rear springs and dampers to find the best possible trade-offs. The goal was to minimize two specific things at once: the vertical bouncing of the rider (heave) and the rocking motion of the bike (pitch). These two goals often fight against each other; a setting that stops the bouncing might make the rocking worse, and vice versa. By using their fast surrogate model, the researcher was able to map out a complete landscape of these trade-offs, identifying a specific set of parameters that offered the best overall comfort without compromising safety.

The results of this approach were striking. In a traditional setup, testing the thousands of combinations the researcher explored would have required over five hours of continuous computer time. By using their trained shortcut model, the same work was completed in less than a minute. The new method reduced the total computational time by more than four times compared to the standard approach. More importantly, the specific suspension settings the algorithm selected actually worked. When the researcher took the winning configuration and tested it in the original, slow, high-fidelity simulation, it proved to be a genuine improvement. The optimized setup reduced the vertical acceleration felt by the rider by 16 percent and slightly improved the stability of the bike's rocking motion.

The study also confirmed that these improvements did not come at the cost of safety. The new settings kept the bike within its physical limits, ensuring the suspension did not bottom out or stretch too far, and maintained the correct height of the bike under a heavy load. The researcher validated their findings by comparing the predictions of their fast model against the results of the slow, traditional simulations, finding that the shortcut was accurate enough to trust for real-world design. This work demonstrates that by combining smart data training with powerful search algorithms, engineers can fine-tune complex mechanical systems with a speed and precision that was previously impossible, leading to a smoother, more comfortable ride for two-wheeler enthusiasts.