The Outlier Detection Of Airport Noise Monitoring Nodes Based On Time Series

As the airport noise pollution problem is more and more serious, and the complaints caused by the noise that has large impact on the environment are increasing every year, the airport noise issue has become one of the obstacles affecting the sustainable development of china’s civil aviation industry. Anomalies contained in the airport noise time series data are of great significance to the timely improvement of the equipments of aircraft and airports, and solutions can be proposed to avoid problems getting more serious. Through finding out the anomalies in the airport noise data which are collected by monitoring nodes distributed in some areas of the airport, the corresponding measures could be taken. There are many reasons for incorrect data acquisition or monitoring nodes invalidity, thus we need detailed analysis for causes of these abnormal cases.In order to find out the reason for the anomalies of monitoring nodes, we need to study how to choose the abnormal monitoring nodes. The outlier detection method of airport noise time series on the single monitoring node is used to reduce the dimension of airport noise time series and symbolize the time series. Then the new measurement method is used to measure the processed time series. According to the result of the final measurement, the outlier detection process is conducted using knearest neighbor outlier factor. The method decreases the quantity of airport noise data to a large extent and improves the computation speed, besides, it reduces the impact of the weakened form information using the improved measurement method. Using the above procedure, the abnormal monitoring nodes could be detected.According to the detected abnormal monitoring nodes, we study the reasons for their abnormalities. To achieve this goal, it is necessary to predict the data of the monitoring nodes on which the abnormal data have been monitored. We propose a node neural network ensemble integrating combination prediction model to improve the prediction accuracy, which trains the corresponding neural network prediction model using the data collected by the monitoring nodes that are strongly relevant to the abnormal ones. The obtained prediction model can be used as a standard of determining the abnormal reasons.Finally, we find out the reasons for the occurence of abnormal nodes. A noise-detection algorithm for correlation noise monitoring nodes is proposed, which is used to compute the similarity degree between the predicted value and the actual value. According to the obtained result, it can be judged that whether the nodes are abnormal or not. If the candidate nodes are judged to be abnormal nodes, then they are the damaged nodes, otherwise, it can be inferred that they are caused by airport or aircraft.