全尾砂充填体正交-BP神经网络强度预测

某矿山采用全尾砂胶结充填，需加入减水剂以保证充填料浆流动性能，其充填体强度各影响因素之间存在着更为复杂的化学物理作用。为了解各因素对充填体强度的影响规律，准确预测其强度，建立了基于正交试验的BP神经网络全尾砂胶结充填体强度预测模型。预测过程中，为提高样本可信度，以料浆浓度、灰砂比、减水剂比例为影响因素，设计了3因素、4水平正交试验方案作为研究数据基础。采用灰色关联度理论分析了各因素对全尾砂充填体强度的影响规律，结果表明：对充填体早期强度（7 d）影响最大的是料浆质量浓度，其次为灰砂比、减水剂比例，对充填体后期强度（28 d）影响最大的是灰砂比，其次为料浆浓度、减水剂。以7 d、28 d强度为输出因子，运用BP神经网络对充填体强度进行预测，预测结果与试验结果最大误差为9.98%，平均误差为2.71%，精度较高，预测可靠性强，具有较好的工程应用价值。

Strength Prediction of Unclassified Tailing Backfilling Based on Orthogonal-BP Neural Network

With the application of the unclassified tailing cemented filling in a mine，the water reducing agent is added to ensure the flow property of filling slurry.There exists more complex chemical-physical effect among various influencing factors of filling body strength.In order to grasp the effect law of various factors on the strength of filling body，and predict its strength accurately，the backfilling strength of cementing filling body prediction model is set up based on the BP neural network of orthogonal experiment.In the forecasting process，three factors and four levels orthogonal experiment scheme is designed as the research data base，considering slurry concentration，cement-sand ratio，water reducing agent adding proportion as the influencing factors，to improve the sample scientific credibility.The grey relational theory is used to analyze the effect law of every factors on the backfilling filling body：the biggest impact on the early strength of filling body (7 d) is slurry concentration，followed by cement-sand ratio，water reducing agent proportion.The late strength of filling body (28 d) is most affected by cement-sand ratio，then followed by slurry concentration，water reducing agent proportion.For purpose of predicting the strength of the backfilling filling body exactly，it regards 7 d，28 d strength as the output factor，and adopts BP neural network，and as a result the maximum relative error is 9.98%，the average relative error is 2.71%，which meets the requirements of filling body strength prediction，and has a good application value to the project.