GNAI Visual Synopsis: A visual representation of an AI program processing sound waves from brake noise testing into a digital interface displaying various subjective rating predictions.
One-Sentence Summary
Antonio Rubio from Applus IDIADA explores the creation of an AI model to predict subjective brake noise ratings using objective data. Read The Full Article
Key Points
- 1. Applus IDIADA’s research focuses on developing an AI model capable of converting objective brake noise measurements into subjective ratings that normally rely on human perception.
- 2. The team used a robust data cleaning process to remove outliers and trained the model with over 8,000 noise events classified by a highly skilled driver, enhancing the model’s predictive accuracy.
- 3. The final model is a combination of an XGBoost classification model, predicting categorical ratings with a confidence level, and an XGBoost regression model, providing a numerical rating prediction—the models switch based on a confidence threshold.
Key Insight
By harnessing machine learning techniques, specifically the use of XGBoost algorithms, Applus IDIADA engineers have pioneered a method to objectively interpret and predict subjective assessments of brake noise, which could significantly streamline the vehicle testing process and enhance quality control.
Why This Matters
This research exemplifies how machine learning can bridge the gap between human subjective experience and objective data analysis, leading to more consistent vehicle quality assessments and potentially higher consumer satisfaction due to the improved precision of brake noise evaluations.
Notable Quote
“A combination of two trained models is proposed to get notable accuracy of the results: a classification model and a regression model.”
