基于神经网络与遗传算法的地应力识别优化研究

Optimization study on in-situ stresses identification based on neural network and genetic algorithm

针对井下存在地应力测点相对较少,量测结果离散,计算非线性等问题,研究将神经网络与遗传算法应用于地应力测量中。利用BP神经网络代表矿井水压致裂的压力值和地应力参数之间的非线性关系,通过实际工程样本对神经网络进行训练,从而保证了网络预测的准确性。结合万福矿区工程实例,利用遗传算法识别出该矿井井下的地应力,通过与理论计算结果对比,验证了该方法应用于地应力识别是可行的、可靠的和方便的,在类似工程中具有一定的实用价值。

BP neural network and genetic algorithm are used in identification of in-situ stresses for the problemsthat are stress measuring points few, measurement discreteness and calculation nonlinear. BP neural network isused to map the nonlinear relationship between borehole pressures and stress parameters, and is trained by usingsamples from practical engineering to ensure accuracy of predicted results. At last, applying genetic algorithm,a Wanfu engineering example is conducted to obtain the in-situ stresses. By comparing theoretical calculationswith identified results,it validates that the proposed method is feasible,reliable and convenient in determinationof in-situ stresses,and can used effectively in similar engineering.