Research On Airport Noise Monitoring Data Recovery Based On Deep Learning

With the rapid development of China's civil aviation construction,while driving the growth of national GDP,the problem of noise pollution has become more and more prominent,and noise prevention and control work has become extremely important.In the rapid development of Internet of Things technology,large-scale remote monitoring systems are widely used in various industries.The monitoring system usually consists of a large number of sensors.By collecting and analyzing real-time data from the sensor network,it not only reflects the real-time operating status of the monitoring system,but also helps airport staff to make the best decisions when abnormal faults occur.A plurality of noise monitoring nodes are set in the noise range to capture the noise data in time,and then the noise data is analyzed to provide a decision basis for noise prevention.This requires each monitoring device to continuously acquire and transmit noise data reflecting the real condition of the monitoring node.However,noise monitoring equipment will inevitably suffer damage,aging or other abnormal conditions that prevent it from working properly.Often,sensor hardware may not be repaired in a short period of time due to the geographical distribution of the monitoring nodes and the complexity of the equipment.Therefore,how to repair the missing data in software way when the node is abnormal has attracted wide attention of researchers.This paper makes use of the advantages of deep learning in feature extraction and data processing,improves and optimizes the algorithm in the missing data recovery,and validates the validity on the measured noise dataset of the Capital Airport.The main work is the following three points:(1)The problem of missing data recovery in the monitoring system is a common phenomenon in the monitoring system.This paper summarizes the problem of missing data recovery in the large-scale monitoring system based on the problem of missing data recovery,and details the problem of the missing data,classic data recovery methods and data mining based data recovery methods.(2)In view of the weakness of traditional data recovery methods in feature extraction,this paper proposes a data recovery method based on deep denoising auto-encoder,which uses the advantages of deep learning model in feature extraction to optimize in the part of model preprocessing.By reconstructing the input data and compressing the representation of the input data,a new expression of the deep feature of the sample set is obtained,which not only fully reflects the characteristics of the input data,but also greatly reduces the redundancy between related information.Combined with the actual experimental conditions,the structural design and parameter setting are improved.Finally,the superiority of the feature extraction preprocessing based on the deep denoising auto-encoder model is proved by experiments.(3)The data recovery method for shallow models often fails to achieve satisfactory results in prediction,and a data recovery method based on deep belief network is proposed.This approach essentially considers the spatial and temporal correlation of data in the modeling process.The method of similarity measurement is used to select highly relevant monitoring nodes to reduce the influence of interference information in the data and further improve the prediction accuracy of the model.In order to quickly and accurately determine the number of hidden layers and their nodes in the deep belief network,the minimum reconstruction error criterion and the eigendimension estimation method are applied to select the appropriate deep belief network hyperparameter.By comparing with the classical data recovery methods on the airport noise monitoring dataset,the superiority of the data recovery method based on deep belief network is proved.