基于神经网络和遗传算法的充填料配比优化设计方法与应用

发布时间：2018-09-03 15:26

【摘要】：长期以来,金属矿山的尾矿几乎都是置于地表尾矿库堆存起来,由于尾矿携带有超标污染物质,直接对矿区生态环境造成破坏。随着环保要求的不断提高,尾矿在地表存放引起的污染问题日益突出,尾矿的排放与治理已成为困扰金属矿山发展的重大因素。充填采矿法具有能尽可能将尾矿回填井下,解决矿石尾矿库库容不足问题,减少和消除尾矿在地表存放对环境的污染等优点得以广泛使用。在充填采矿法设计过程中,充填料浆的配合比设计是决定充填质量的首要因素,在控制充填总成本的前提下,选择合理的料浆配比,可以有效地保证充填体强度,满足回采工艺要求。因此,确定合理的充填料浆配比是保证安全、高效经济回采的重要前提。选择合理的料浆配比,一般通过实验室全面配比实验,测得不同配比参数下充填体的性能,进而推荐满足采矿方法强度要求的配比参数,但全面实验工作量大且容易受人为操作等因素干扰,在实际生产应用中具有一定的局限性。本文设计了充填体强度正交试验方案,通过实验获得了大量不同配比情况下的充填体强度数据,应用BP神经网络和现有的充填体强度预测模型,对不同配合比下充填体强度进行预测,并在此基础上采用遗传算法进行充填料配比优化设计研究,从而高效的推荐出满足不同采矿方法强度要求且满足矿山输送条件的充填料配比,主要研究内容如下:(1)分析并比较了多个常用的充填体强度模型,选择建立基于BP神经网络的全尾砂充填体强度预测模型,预测精度较高,对全尾砂充填料浆配比设计起指导作用。选择使用吴祥辉的充填体强度模型用于对废石-尾砂充填体强度值的预测,使用实验数据对后者进行拟合回归分析,结果表明选择和建立的充填体强度模型预测精度较高,能够满足实际工程对全尾砂充填体强度预测的精度要求,对废石-分级尾砂充填料配比优化设计具有很好的指导作用。(2)根据遗传算法理论,基于充填体强度预测模型,对充填料配比进行优化设计,获得一个满足采矿方法强度要求的充填料配比参数最优解集,从而为矿山决策者提供了最优选择,同时推荐出了具有良好抗离析性,且能实现自流输送的成本较低的充填料配比方案的选择方法,对于不同预配要求,只需修改部分参数,即可得到最优配比参数,大大提高了充填料配比的设计效率且精度较高。(3)在以上研究基础上,使用面向对象的C#语言编写了充填料性能预测和配比优化智能决策系统,将复杂的计算机语言界面化,该系统具有界面友好,使用方便,操作简单等特点,为充填料性能预测及配合比优化设计提供了良好的决策支持,大大提高了充填料配比设计的效率。
[Abstract]:For a long time, almost all the tailings of metal mines have been stored in the surface tailings reservoir. Because the tailings carry pollutants in excess of the standard, it directly damages the ecological environment of the mining areas. With the increasing requirement of environmental protection, the pollution caused by tailings stored on the surface is becoming more and more serious. The discharge and treatment of tailings have become a major factor that puzzles the development of metal mines. The backfill mining method has the advantages of being able to backfill the tailings underground as far as possible, solving the problem of insufficient storage capacity of the tailings, reducing and eliminating the environmental pollution caused by the tailings stored on the surface of the earth's surface, and so on. In the design process of filling mining method, the proportion design of filling slurry is the primary factor to determine the filling quality. Under the premise of controlling the total cost of filling, choosing reasonable slurry ratio can effectively guarantee the strength of filling body. Meet the requirements of mining process. Therefore, it is an important prerequisite to ensure the safety, high efficiency and economic recovery to determine the reasonable ratio of filling slurry. Selecting a reasonable mixture ratio of slurry, the performance of the backfill under different proportioning parameters is generally measured through the laboratory comprehensive proportioning experiment, and then the matching parameters which meet the requirements of mining method strength are recommended. However, the total experimental workload is large and easily disturbed by human operation, which has certain limitations in practical production and application. In this paper, the orthogonal test scheme of backfill strength is designed, and a large number of data of backfill strength under different proportions are obtained through experiments. The BP neural network and the existing prediction model of backfill strength are used. On the basis of predicting the strength of backfill under different mix ratio, genetic algorithm is used to study the optimum design of filling material ratio. The main contents of the research are as follows: (1) analyzing and comparing several commonly used backfill strength models, which can meet the intensity requirements of different mining methods and meet the requirements of mine transportation. The main contents of this paper are as follows: (1) several commonly used backfill strength models are analyzed and compared. Based on BP neural network, the prediction model of the strength of the whole tailing filling body is established, and the prediction accuracy is high, which plays a guiding role in the design of the slurry ratio of the whole tailing filling. Wu Xianghui's backfill strength model was used to predict the strength value of waste rock and tailings filling body, and the experimental data were used to fit the regression analysis of the latter. The results showed that the prediction accuracy of the selected and established backfill strength model was high. It can meet the precision requirement of the strength prediction of the whole tailings filling body in actual engineering, and has a good guiding function for the optimum design of the ratio of waste rock and graded tailings filling material. (2) according to the genetic algorithm theory, based on the strength prediction model of the filling body, The optimum design of filling material ratio is carried out to obtain an optimal solution set of filling material ratio parameters which meets the requirements of mining method strength, which provides the best choice for mine decision makers, and recommends that it has good anti-segregation property. And it can realize the selection method of filling material proportion scheme with lower cost of self-flow transportation. For different pre-matching requirements, only some parameters can be modified, and the optimal proportion parameters can be obtained. The design efficiency and precision of filling mixture ratio are greatly improved. (3) on the basis of the above research, an intelligent decision system for performance prediction and proportioning optimization of filling material is developed by using the object-oriented C # language, which interfaces the complex computer language. The system has the advantages of friendly interface, convenient use and simple operation. It provides a good decision support for the performance prediction of filling material and the optimum design of mixture ratio, and greatly improves the efficiency of the design of filling mixture ratio.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD853.34

【参考文献】