Research On Crack Damage Prediction Of Asphalt Pavement Based On Tire Noise

This study aims to propose and evaluate a machine learning algorithm based on a tire noise propagation mechanism for predicting crack damage in asphalt pavement.We adopted five machine learning algorithms including support vector classifier(SVC),random forest classifier(RFC),Adaboost,multi-linear perceptron(MLP)classifier,and stacking classifier are used to improve the accuracy of road pavement damage prediction by considering tire noise.To evaluate the accuracy of the model,different performance metrics were used,namely accuracy,precision,recall,and F1-score.To verify the performance of the model,we built a dataset based on road pavement damage data collected in September 2022 at Yuanjiang Road,Fuzhou,Fujian,China,using microphones,cameras,and GPS.In this study,three types of audio were recorded at different intervals but at the same speed,test 1 captured 6.10 minutes of audio,test 2 with 7.13 minutes of audio,and test 3 with 18.01 minutes of audio.This dataset focuses primarily on the start and end time of road damage,by taking into account the images of the road surface taken by the camera.After that,the recorded audio is labeled by using Audacity,and the road damage is classified for pre-processing.We trained and tested these five machine learning algorithms in a Python programming environment.Experimental results show that our proposed RFC method outperforms other models in predicting road pavement damage.The RFC method achieves the best performance on all metrics on the three test datasets(road damage audio 1,road damage audio 2,and road damage audio 3),where accuracy,precision,recall,and F1-score reaching 99%,98%,99%,and96%,it has obvious advantages over other models.Five machine learning algorithms were trained and tested in the Python programming language.Experimental findings show that the RFC method outperforms other models in predicting road pavement damage,and achieves better performance on all metrics on the three test datasets(road pavement damage in the audio 1 test,road pavement damage in the audio 2test,and road pavement damage of audio 3 test)with accuracy,precision,recall,and F1-score of 99%,98%,99%,and 96% respectively.it has obvious advantages over other models.In addition,this study confirms that road damage can be detected based on tire noise propagation datasets and using various classification prediction models.The prediction method based on tire noise proposed in this study is not only reliable and economical but also proved to be effective in practical applications.This study mainly helps to express the novelty of using traditional and ensemble models using different classification algorithms to analyze the performance of expected prediction models and improve accuracy results,as well as the novelty of using machine learning algorithms based on tire noise propagation.The results of this study can help predict and detect pavement damage and reduce the number of traffic accidents.The model with enhanced accuracy provides traffic authorities with enough information that they can use to prevent accidents,damage to pavements,and roads,and prepare for unforeseen events.This study shows how ML algorithms can be used to predict pavement degradation,which can help the world become more economically and environmentally sustainable,and solve the problems posed by damaged pavement.This alternative is also reliable,cost-effective,and effective.