Research Of Tunnel Monitoring System By Neural Network Based On RT-WAR Algorithm

With the accelerated process of urbanization, road tunnel is becoming more and more important in urban traffic. Because tunnel was built in the underground, the environment inside became worse, and the management of the tunnel was more complex than normal. Therefore, to study the tunnel and produce efficient and rational management method is necessary.When the tunnel is running, it will produce a large amount of monitoring data. Digging for these monitoring data is of great value for developing monitoring strategies. This thesis will combine methods of data mining with intelligent management to put forward a set of specific models for tunnel monitoring system. Finally predicting the CO concentrations over time and traffic congestion to make the staff give the reasonable corresponding solutions in a timely manner, this will play an important role in reducing traffic injuries.A very classical algorithm in association rule is Apriori algorithm, which can dig out the correlation between the various projects. It is very helpful for making intelligent decision for tunnels. But this algorithm has two faults including repeatedly scanning the database and producing a large number of the candidate items. In order to solve these two problems, this thesis proposes method of reducing transaction. Its main idea is to reduce the times of database scanning and the number of candidate items, which improves the efficiency of the algorithm. Considering the features of monitoring data, this thesis introduces the concept of weight, and puts longitudinal weighting on data record, to make that data has different importance in different periods. Meanwhile, reinforce the importance of recent data on data mining and ensure the accuracy of data mining. Therefore, this thesis puts forward a new improved algorithm which is weighted association rule based on reducing transaction, short title is RT-WAR algorithm. Finally, the experiment is made to compare the Apriori algorithm with RT-WAR algorithm, the result shows that RT-WAR has been greatly increased in efficiency.The correlation between the data by association rules is useful for prediction of tunnel environment situation and traffic situation. This thesis forecasts when the CO concentrations exceed normal levels and traffic congestion, by neural networks for prediction. Early in the forecast, according to the results mining by association rule to determine the neural network input layer nodes and the initial weights. This approach is better than that of blindly giving input factors and random weights, more conducive to accelerating the training of neural networks, while increasing the accuracy of forecasts.This thesis has established a prediction model of the tunnel monitoring system, and put the historical data of actual operation of Wuhan shuiguo Lake crossing as sample data to make the experiment. Firstly, preprocess the data, then analyze the association rules, finally, use a neural network to predict the CO concentrations over time and traffic congestion. Matlab simulation experiments show that prediction model of the tunnel monitoring system is very effective.