There is no doubt that traffic noise level is considered a harmful environmental pollution that has a serious impact on human quality of life. This paper shines a light on the traffic noise level in the Arabian Gulf region. More specifically, it predicts the traffic noise level on a ring road in Kuwait by using an adaptive neuro-fuzzy inference system (ANFIS). Field measurements data were collected from 20 different measurement points twice a day. It resulted in 480 measurements of ten variables: traffic noise level, light and heavy vehicle count, average speed of both, road width, building height, pavement condition, and air and roadway temperature. To assist in collecting the data, a vision-based vehicle detection system was developed using machine learning. The system successfully managed to reach an accuracy of 90%, whereas the ANFIS traffic noise prediction model achieved a RMSE of 0.0022. The model was then tested on a different road as a validation step, where it gave a RMSE of 0.06. Afterward, two sensitivity analysis techniques were utilized to rank the nine input variables from the highest relative importance to the lowest: the $R^2$-based metric and single-input single-output. Based on the results, the most important variable was light vehicle count, and the least effective variable was heavy vehicle count. The air and road temperatures were ranked the fourth and the seventh respectively. Subsequently, four different scenarios were designed to predict the traffic noise level in 2025. The first three scenarios were based on the sensitivity analysis results. Scenario I assumes a reduction in the speed limits on the ring road from 120 km/h to 100 km/h. Scenario II assumes the building height would be high, which will give the same effect as adding a noise barrier. Scenario III assumes there would be a truck curfew in the evening. Finally, Scenario IV assumes there would be no noise control system at all. The results were equal to 76.01, 80.66, 83.36, and 84.56 dBA respectively. It can clearly be seen that a traffic noise control system can reduce traffic noise effectively.

毫无疑问，交通噪音水平被认为是一种有害的环境污染这严重影响了人类的生活质量。这篇论文揭示了阿拉伯湾地区的交通噪音水平。更具体地说，它通过使用自适应神经模糊推理系统 (ANFIS) 来预测科威特环路上的交通噪声水平。每天两次从 20 个不同的测量点收集现场测量数据。它对十个变量进行了 480 次测量：交通噪声水平、轻型和重型车辆数量、两者的平均速度、道路宽度、建筑物高度、路面状况以及空气和道路温度。为了协助收集数据，使用机器学习开发了一种基于视觉的车辆检测系统。该系统成功地达到了 90% 的准确度，而 ANFIS 交通噪声预测模型的 RMSE 为 0.0022。然后将该模型作为验证步骤在不同的道路上进行测试，其 RMSE 为 0.06。然后，利用两种敏感性分析技术将九个输入变量从最高相对重要性到最低排名：${\mathrm{电阻}}^2$基于度量和单输入单输出。根据结果​​，最重要的变量是轻型车辆数量，而最不有效的变量是重型车辆数量。气温和路面温度分别排在第四和第七位。随后，设计了四种不同的情景来预测 2025 年的交通噪声水平。前三种情景是基于敏感性分析结果。情景 I 假设环路上的限速从 120 公里/小时降至 100 公里/小时。情景二假设建筑物高度很高，这将产生与添加隔音屏障相同的效果。情景 III 假设晚上会有卡车宵禁。最后，场景 IV 假设根本没有噪声控制系统。结果分别等于 76.01、80.66、83.36 和 84.56 dBA。