基于LM-BP算法的隧道爆破振动灾害预测的应用

Application of Tunnel Blasting Vibration Disaster Prediction Based on LM-BP Neural Network

为了对隧道爆破振动灾害的危险状态进行有效地预测,实验采用基于Levenberg-Marquardt(LM)算法改进的BP算法,建立以实测隧道爆破掏槽眼装药量、爆心距和爆破振速为主要爆破影响因素的神经网络模型,对振速进行预测分析,预测结果与实测数据吻合良好;继而引用GB 6722—2014《爆破安全规程》所规定的临界安全振速反向预测掏槽装药量,通过反向预测计算得出满足安全振速要求的临界掏槽装药量。预测结果表明:LM-BP算法相比传统的经验模型在振速预测上表现更好,通过反向的预测运算,能有效预知临界装药参数,对爆破振动安全预测及控制有积极的意义。

The BP neural network model,with charging amount of blasting cut,blasting center distance and blasting velocity as main factors,is established based on Leyenberg-Marquardt（ LM） calculation method; and the blasting vibration velocity is predicted and analyzed. The charging amount of blasting cut is calculated by means of critical blasting vibration velocity in related criteria. The calculation results show that LM-BP neural network method is superior to traditional method in terms of prediction of blasting vibration velocity; the blasting cut charging amount calculated by means of critical blasting vibration velocity inverse calculation method is rational and effective.