Research On Data Stream Transmission And Clustering Techniques Of Distributed Mobile Gas In Underground Coal Mine

As the application of mine IOT(the Internet of Things) in coal mine, and the application of lots of mobile gas detectors(e.g. intelligent miner’s lamp), a large amount of monitoring data of mobile gas are generated, which put a challenge to the traditional processing method that monitoring data based on fixed sensors. On the basis of researching on the features of monitoring data of mobile data, combining with the schema of mine IOT, the thesis studies on algorithms of acquisition, transmission, clustering, and data processing of gas data from four levels as “sensing, transmission, perception, and application”, which reduce adjusting workload of lots of mobile sensors greatly by making some reasonable adjustments on the data of mobile monitoring sensors through taking advantage of the data of neighboring fixed monitoring sites.The thesis studies on the features of gas monitoring data, including three features as “a huge quantity of data, compressive, and existed abnormal data” at frist. Based on the schema of IOT, the schema of monitoring system is designed, which is suitable for adjusting mobile gas detectors. Then, the thesis analyzes on two network models, single-hop and multi-hop network, which are suitable for data transmission of gas flow, and studies on network compressed sensing algorithm. According to time synchronization technology of mobile gas detectors within the same cluster, a improved RBS algorithm is given. Considering that random matrix will increase the cost of system in practical applications, Walsh-Hadamard matrix is researched, the reconstructing effects of which is almost the same with random matrix in simulation. By introducing the concept of total variation for gas data, the effective reconstructing of gas data can be realized by TVAL3 algorithm.Measurement matrix of network compressed sensing relies on the chosen routing technology: according to the complex underground environments and serious inference, dynamic self-organized routing technology based on multi-weight adjusting is proposed; according to that router nodes will be too unique in constructing paths of dynamic self-organized crouting technology, the idea of random routing is applied in self-organized routing technology, and by expanding routing and giving a certain probability, random dynamic self-organized routing technology is put forward, which solving the problem that there is a greater correlation in measurement matrix; according to that there is random uncertainty in measuring matrixes by random dynamic self-organized routing, pseudo-random self-organized routing technology based on pattern selection is proposed, which solving the problem of high storage in measuring matrixes.According to the character of time series of gas stream, the thesis studies on mathematic analysis method of time series similarity based on pattern beam and pattern set. Combining with short-time stationary of gas data, the feature selection method of short-time gas data based on sliding time window is proposed. On the basis of the fuzzy C-means clustering algorithm, short-time gas stream in the fuzzy C-means clustering algorithm is put forward to analyze the convergence effects of the data of gas stream based on binary statistic and multivariate statistic, which provides qualified data that available for analysis and calculation for data correction of gas sensors afterward.According to the features of “single-point, multi-position” and “multi-point, single-position” of measuring data of mobile gas, the thesis finally proposes the processing method of data of mobile gas flow in mining based on 2-tuple and n-tuple. By classifying the tested data from mobile and fixed gas detectors into two types of data as multi-dimensional data points of gas and tested data pairs of gas, adaptive weighting data fusion algorithm based on clustering and confidence fusion algorithm based on singer sensor multi-period measuring are proposed, which are applied in global decision of gas data, and to improve validity of calibration information of gas by opportunity communication.There are 67 images, 10 tables, 102 references in this thesis.