Research On The Identification Method Of The Blockage Fault Of The Iron Ore Concentrate Slurry Transportation Pipeline

As an important basic industry of national economy,iron and steel industry plays an important role in the process of promoting industrialization.However,with the continuous development of this industry,the stable supply of iron ore raw materials has become a bottleneck restricting its development.To achieve the goal of transporting large iron ore to the destination with high efficiency,low cost and pollution-free,pipeline hydraulic transport has become an important direction of development,and has been widely used.Although the pipeline transport of slurry has many advantages,such as large amount of transportation,energy saving,environmental protection and low loss,there are also high risks: the solid particles in the pipeline may be silted and blocked.The blocking phenomenon reduces the transport efficiency and affects the stability of the working conditions.In serious cases,the whole pipeline may be scrapped.Through machine learning and modern signal processing,the effective pipeline critical deposition velocity prediction model and blockage fault identification model are studied in this paper.The purpose is to provide a scientific and reasonable identification method for the recognition of blockage fault for the iron concentrate slurry pipeline.This paper mainly carried out the following research work:(1)Aiming at the complex characteristics of hydraulic transportation of iron concentrate slurry pipeline,a comprehensive experimental platform for iron concentrate slurry transportation pipeline has been independently designed and built.Based on the experimental platform,the experimental research on the related properties of material and slurry,the conveying process of slurry and the determination of the critical deposition velocity have been carried out.According to the relevant experimental results,the defects of the traditional critical deposition velocity formula,the main influencing factors of critical deposition velocity,the force and movement of solid particles,and flow pattern characteristics of solid-liquid two phase flow under different conveying velocity were analyzed.(2)Aiming at the problem that the traditional method based on mechanism analysis and numerical simulation for pipeline critical deposition velocity has the shortcomings,such as complex modeling,difficult to predict and low accuracy,this paper proposed a combination forecasting method for critical deposition velocity of iron concentrate slurry based on improved shuffled frog leaping algorithm(SFLA),which provides a reasonable basis for the accurate selection of the flow velocity of the iron concentrate pipeline.This method optimizes the weights of the combined forecasting model by using the improved SFLA,which makes the model comprehensive used the effective information provided by each single forecasting model,and avoids the problem of premature convergence and local optimality in the process of optimization.The results show that the mean squared error,average relative error and mean absolute error are only 0.79%,2.05% and 2.39% respectively,and it is confirmed that the combined forecasting model has better predictive performance than the traditional prediction model and the critical deposition velocity formula.(3)Aiming at the nonlinear and non-stationary of the iron concentrate slurry pipeline acoustic blockage signal is difficult to extract,and the effects of noise and outliers in pattern recognition,this paper proposed an iron concentrate slurry pipeline blockage fault recognition method based on Local Mean Decomposition(LMD),information entropy and improved Extreme Learning Machine(ELM).By combining LMD and information entropy theory,the corresponding information entropy measure index is extracted,and then the fault recognition model based on fuzzy theory is established.The experimental results show that the effective depiction and quantitative interpretation of the internal characteristics of different levels in the signal are realized by the measure index of the combination of different information entropy.With the increase of adding noise ratio,the fault recognition rate of the proposed method is still between 83.33%~91.67%,which confirms that the proposed method has good anti noise performance and blocking fault recognition effect.(4)Aiming at the problem that single kernel function can not be fully interpreted model of decision function,there is some redundancy between the features,and unbalanced sample has great influence on the identification of minority samples,this paper proposed a iron concentrate slurry pipeline blockage fault recognition method based on minimal redundancy maximal relevance(MRMR)and multi-kernel extreme learning machine(MKELM).To map the optimal feature of the mixed domain to the kernel function space,this paper used MRMR to optimize the extracted mixed domain feature set,and then establish a multiple kernel recognition model based on global kernel function and local kernel function.The experimental results show that the accuracy of the proposed method reaches 96.67% under the condition of sample equilibrium distribution,and the probability of misdiagnosis for a few samples is only 20% when the sample distribution is unbalanced.It not only effectively improves the computational efficiency and recognition accuracy of the model,but also has better reliability and practicability.In this paper,the problem of blockage fault recognition of iron concentrate slurry pipeline is taken as the starting point of the research.The pipeline critical deposition velocity prediction model and the blocking fault recognition model have enriched the theory and method of recognition for blocking failure of iron concentrate slurry pipeline,and promoted the application and development of recognition technology of blockage fault.