基于改进PCA与有序多分类Logistic的充填管道磨损风险评估

Risk Assessment of Filling Pipeline Wearing Based on Improved PCA and Ordered Multi-class Logistic

为准确预测矿山充填管道磨损风险,建立改进PCA与有序多分类Logistic回归组合的充填管道磨损风险评估模型。结合实际经验,选取12项指标(9项定量指标和3项定性指标)建立评估模型。依据改进PCA算法,筛除影响力指数小的充填料浆密度和充填料浆腐蚀性2项指标,将优选出的主要指标代入有序多分类Logistic回归模型,依照相应概率大小进行风险性等级判定,最后预测矿山充填管道磨损风险等级概率。该方法摒弃了关联性较低的指标,得到可靠的充填管道磨损风险概率分布,为类似矿山科学预测管道磨损风险及采取有效防护措施提供了理论依据。

The filling mining method is mainstream mining method used in major mines today. The safe implementation of filling technology depends on the construction of a good filling pipeline transportation system.Considering the complexity of the filled pipeline system and the large number and variety of influencing factors,in order to accurately predict the wear risk of the filling pipeline of the mine,a wear risk assessment of the filling pipeline based on the improved PCA and the ordered multi-class Logistic regression combination model was built.On the basis of practical experience,a total of 12 items including the volume fraction of the filling slurry,the filling doubled line,the corrosiveness of the filling slurry,and the material of the pipe were selected. Reasonable risk levels were divided according to the characteristics of each indicator and the corresponding evaluation model was constructed.The filling production data of four mines such as Jinchuan Longshou mine and Dahongshan copper mine were taken as samples.The improved PCA algorithm was used to analyze the correlation of each index,and the three principal components with a cumulative contribution rate of 91.125% and factor load of the principal component of each indicator were obtained. In the end,the two indicators with low correlation and small influence index,the density of the filling slurry(27.75)and the corrosiveness of the filling slurry(27.60)were deleted.The preferred main indicators were substituted into the ordered multi-class Logistic regression model,the continuous indicator values were linearly fitted to the discrete risk levels,and the regression coefficients,standard errors and significance levels of each index were calculated, and the equations of probability fluctuations were solved. Finally,the probability of four mines corresponding to different wear risks Ⅰ(not easy to wear),Ⅱ(relatively easy to wear),Ⅲ(easy to wear),Ⅳ(very easy to wear) were obtained. They were:Jinchuan Longshou mine is 0.247,0.440,0.154,0.153;Dahongshan copper mine is 0.179,0.240,0.323,0.258;Hedong gold mine is 0.181,0.227,0.345,0.247;Xincheng gold mine is 0.181, 0.227,0.345,0.247.From a theoretical point of view,the risk level corresponding to the probability of the largest value was used as the final judgment level of the wear risk of the mine filling pipeline,and Jinchuan Longshou mine is Ⅱ,Dahongshan copper mine is Ⅲ,Hedong gold mine is Ⅲ,Xincheng gold mine is Ⅲ. The mine also makes corresponding level protection and maintenance measures based on this. From the actual production application,it is said that the probability of the Ⅳ risk level should be paid special attention,and the normal operation of the filling pipeline should be guaranteed to the greatest extent within the scope permitted by technology and funds. The mathematical method combining improved PCA and ordered multi-class Logistic regression avoids the collinearity between the various indicators,reduces the interference of the weak indicators on the evaluation results,and obtains the accurate risk of wear risk of filling the pipeline.It provides a theoretical basis for scientific prediction of pipeline wear risk and the implementation of effective and economical protection measures for similar mines.The mine can construct an appropriate filling pipeline protection system according to its own development.