Research On Intelligent Decision-making Of Pavement Maintenance Based On Performance Prediction

With the extension of the service life of highways in China and the increase of traffic flow,under the continuous influence of traffic load and natural environment,the pavement performance inevitably gradually declines,and the demand for road maintenance is also increasing.If the pavement performance cannot be accurately predicted,it is impossible to scientifically use maintenance measures to restore the pavement to its due service level.Therefore,it is of great theoretical significance and practical value to establish an accurate pavement performance prediction model,and on this basis,to carry out scientific and intelligent maintenance decisionmaking research with the goal of maximizing benefits,so as to provide technical reference and support for China’s future road pavement performance evaluation and maintenance.In order to improve the pavement performance,prolong the service life of the road,and provide scientific support for road maintenance decisions,this thesis fully studies the problems of pavement performance prediction and pavement maintenance decision-making,and designs a decision-making model based on performance prediction.The main research is as follows :1.Correlation analysis of influencing factors of pavement performance is conducted.In order to ensure the accuracy of pavement performance prediction and maintenance decision-making,the correlation between each influencing factor and pavement performance was analyzed by Spearman correlation coefficient method and random forest algorithm respectively,and the features with greater influence were selected as new data sets to prepare data for subsequent prediction and decision-making research.2.A Stacking prediction model that can accurately predict pavement performance is proposed.Aiming at solving the problem of insufficient accuracy of traditional pavement performance prediction models,this thesis establishes multiple pavement performance prediction models based on machine learning algorithms,conducts comparative analysis,and selects high-precision prediction models to fuse in a Stacking manner.The model has high prediction accuracy and can provide a basis for accurate calculation of benefits for decision-making research.3.An intelligent decision model based on performance prediction is proposed.Aiming at solving the problems that traditional decision-making methods are highly dependent on expert experiences and difficult to accurately calculate the benefits of different pavement maintenance measures,this thesis calculates the benefits through the Stacking prediction model,and introduces the particle swarm optimization algorithm to automatically search for the optimal solution to maximize the benefits.Compared with the traditional decision-making method,the model can obtain the optimal decision more accurately and intelligently.4.The web application of the decision model is developed.As there exists the lagging of maintenance decision research in practical application,the application deploys the decision model at the back end and provides an access interface so that it can be put into industrial application.The performance prediction model based on machine learning methods can accurately predict the decay of pavement performance.The decision model based on performance prediction can accurately calculate the benefits of maintenance measures,and can automatically obtain the optimal decision with the help of particle swarm optimization algorithm,which lays a foundation for intelligent road maintenance.