Machine learning helps reveal key factors affecting tire wear particulate matter emissions
Environment International, 2024年12月19日
本研究采用机器学习方法识别影响轮胎磨损颗粒物排放的关键因素，发现避免剧烈驾驶行为、降低胎面温度及减少小花纹数量可有效减少排放。
Abstract
The construction and interpretation of a machine learning based model tire wear emissions provides new insights into the refined assessment control non-exhaust emissions. Tire particles (TWPs) are generated with every rotation tire. However, obtaining TWPs under real driving conditions revealing key factors affecting challenging. In this study, we obtained dataset by simulating process using simulator custom-designed test conditions. This study shows that PM 2.5 accounts for about 65 % 10 . response relationship between TWP (both 2.5-10 ) (the radial force, lateral tangential speed, torque, contact area, total contour length tread temperature) was (ML) method. random forest (RF) developed displayed good prediction performance an R 2 0.84 0.78 on set, respectively. Model-related (similarity network graph) model-unrelated (partial dependence plots centered-individual conditional expectation plots) explainability methods were used to break black box ML. Model results show feature parameters-emission relationships different. Avoiding strenuous behaviors (TTF < 400 N, TLF N), reducing temperature (T 45℃), minimizing number small patterns feasible ways reduce TWPs.