Research On Trend Knowledge Discovery Of Coal Mine Safety Based On Times Series Data Mining

Coal is the basic energy for Chinese economic and social development.Safety production is the prerequisite to ensure the stable supply of energy.Although the level of coal mine safety in China has improved in recent years,there is still a certain gap compared with developed countries in Europe and the United States,and the situation of coal mine safety is still grim.After years of coal mine information construction,a large number of safety data have been accumulated in enterprise information system.It is not only the actual needs of coal enterprises,but also the important research content of coal mine safety management to give full play to the value of data,assist the safety management work and further improve the safety situation of coal mines.However,at present,there is no systematic study on the trend law contained in the coal mine safety time series data in the research of macro and micro coal mine safety time series data mining and the data mining model of coal mine safety information system.Due to the lack of relevant research,it is difficult to effectively play the role of coal mine safety time series data,including analyzing the complex system structure of coal mine safety system,identifying the trend change relationship between hazard sources,and promoting the realization of intelligent risk identification and prevention.It can not provide valuable guidance and reference for effectively assisting coal mine safety management and continuously improving the data mining function of coal mine safety information system.Therefore,based on the above practical background and insufficient research,this paper proposes the research topic of coal mine safety trend knowledge discovery based on time series data mining,take coal mine safety time series data as mining object and trend knowledge discovery as goal.Based on the review of the general theory of data mining and knowledge discovery,the classification of accident causes and hazard sources,the data structure of coal enterprises,the trend analysis and description of time series,the piecewise linear representation(PLR)of time series,and the measurement of trend similarity,the key scientific problems to be solved in the research are put forward,which includes the composition of coal mine safety time series data,the connotation of coal mine safety trend knowledge and the scientific process of its discovery,the selection and construction of effective methods of trend knowledge discovery,among which the effective methods of trend knowledge discovery include the trend description primitive system and piecewise linear representation method used for time series dimensionality reduction trend transformation,as well as the trend similarity measurement method for data mining.The specific research contents are as follows:1)Research on data structure of coal mine safety time series.In view of the composition of coal mine safety time series data,based on the theoretical basis of the classification of accident causing factors and hazard sources,and based on the existing data of coal enterprises,this part constructs the coal mine safety time series data system including personnel,equipment and facilities,environment,organization and internal management,relevant external and other categories,and analyzes its main characteristics,so as to lay a good data foundation for the coal mine safety trend knowledge discovery.2)Research on time series trend description primitive system.In view of the trend element system used to describe the trend knowledge of coal mine safety,this part puts forward the corresponding relationship between the trend element and the change direction and mean level of segmented subsequence.On this basis,a nine element trend element system is defined,which provides an effective method for accurately describing the local trend and the overall trend of coal mine safety time series.3)Research on the connotation and discovery process of coal mine safety trend knowledge.First of all,in view of the connotation of coal mine safety trend knowledge,based on the existing time series trend analysis and description of the relevant research,combined with the knowledge of coal mine safety field,this part proposes the connotation of coal mine safety trend knowledge,That is to say,the trend sequence pattern and the trend similarity relationship between the sequence are formed by the orderly connection of several trend primitives,which are contained in the data of coal mine safety time series and have practical value for coal mine safety management.Its includes the frequent sequence pattern of single sequence and the trend similarity relationship of multiple sequences and their common patterns.It also analyzes its main characteristics from the aspects of complex diversity,dependence,dynamic expansion,intuitive understandability and sparsity,analyzes its main functions from the aspects of effective use of coal mine safety time series data,analysis of complex system structure of hazard sources,prediction reference and indication between similar trend sequences,intelligent risk identification and prevention of hazard sources,etc.Then,aiming at the scientific process of coal mine safety trend knowledge discovery,this paper puts forward a process model of knowledge discovery of coal mine safety trend,which uses segmented linear representation and nine element trend description system to preprocess dimensionality reduction trend transformation,sequential pattern discovery method and trend similarity measurement method to mine data,and evaluates data mining results with true reliability and useful value for coal mine safety management.So as to provide the process framework reference and guidance for coal mine safety trend knowledge discovery.4)Research on data preprocessing method of coal mine safety time series.First of all,for the problem of piecewise linear representation(PLR)method in preprocessing,this part combines genetic algorithm(GA)with PLR,constructs a PLR method with self-adaptive characteristics,that is,a time series piecewise linear representation method based on GA(PLR_GA),and selects experimental data to verify the feasibility and effectiveness of this method,so as to provide an effective method with flexibility and applicability for the preprocessing of dimensionality reduction trend transformation of coal mine safety time series data.Then,on the basis of PLR_GA,combines with the nine element trend description system,this part puts forward a process model for preprocessing the trend transformation of dimensionality reduction for coal mine safety time series data.So as to provide a preprocessing process model for mining safety trend knowledge discovery.5)Research on trend knowledge discovery of single time sequence in coal mine safety.In order to verify the validity of the process model of mining safety trend knowledge discovery for mining safety single time series trend knowledge,this part selects the real time series data of coal mine safety,and uses the preprocessing process model of dimension reduction trend transformation for preprocessing,uses the sequential pattern discovery using equivalence classes(SPADE)to identify the frequent patterns in the preprocessed trend sequence data.Then,analyzes and identifies the frequent patterns that meet the needs of coal mine safety management and have use value by integrating the knowledge of coal mine safety field.At the same time,adjusts the compression ratio of preprocessing parameters,analyzes whether the same frequent mode exists under different compression ratio conditions,in order to verify the real reliability of the frequent mode.The research shows that:the model of the preprocessing process of the dimensionality reduction trend transformation of coal mine safety time series data can effectively retain the trend information of the original coal mine safety time series data.Using the process model of coal mine safety trend knowledge discovery can effectively discover the trend knowledge in single time series data of coal mine safety.6)Research on trend similarity measurement method.In order to provide an effective trend similarity measurement method for multi time series trend knowledge discovery of coal mine safety,on the basis of the nine element trend description system,this part defines the matching distance by comparing the similarities and differences between the trend primitives,uses the dynamic programming principle of dynamic time warping(DTW)method,constructs a dynamic pattern matching(DPM)method,and verifies its superiority by experimental data.Thus,it provides an effective trend similarity measurement method for multi time series trend knowledge discovery of coal mine safety.7)Research on multi time series trend knowledge discovery of coal mine safety.In order to verify the validity of mining safety trend knowledge discovery process model for discoving multi time series trend knowledge,this part selects the real time series data of coal mine safety,and use the preprocessing process model based on PLR_GA to carry on the dimensionality reduction trend transformation preprocessing,measures the similarity of the transformed trend series with DPM method,and uses hierarchical clustering method to identify and classify the trend categories,uses the SPADE algorithm to identify the common sequence patterns of different trend categories,through the differences of the common patterns of different trend categories to verify the real reliability of the trend similarity relationship between the data.The results show that the process model of coal mine safety trend knowledge discovery can discover the trend knowledge in multi time series data of coal mine safety.The innovations in the study are summarized as follows:1)Aiming at the problem of the composition of coal mine safety time series data,a coal mine safety time series data system including personnel data,equipment and facilities data,environment data,organization and internal management data,and external related data is constructed,which provides a reliable data selection basis for coal mine safety trend knowledge discovery.2)In view of the problem of trend element system to describe the trend knowledge of coal mine safety,this paper establishes the corresponding relationship between trend primitives and change direction and mean level of segment subsequences,and on this basis,constructs a nine element time series data trend description primitive system.3)To solve the problem of piecewise linear representation in data preprocessing,this paper proposes a piecewise linear representation method of time series based on GA(PLR_GA),combines with the nine element trend description primitive system,puts forward the method of transforming coal mine safety time series data into trend time series data further.4)To solve the problem of trend similarity measurement method,this paper defines the matching distance by comparing the similarities and differences between trend primitives,constructs a dynamic pattern matching method(DPM)based on the dynamic programming principle of dynamic time warping(DTW).5)In view of the scientific process of coal mine safety trend knowledge discovery,this paper puts forward a process model of knowledge discovery of coal mine safety trend used in trend knowledge discovery of coal mine safety,which uses PLR_GA method and nine element trend description primitive system for preprocessing of dimensionality reduction trend transformation,SPADE method and DPM method to mine data,and evaluates data mining results with true reliability and useful value for coal mine safety management.