Abstract: Aiming at the issue of insufficient accuracy in the prediction model of drilling speed for spiral drilling rigs established by existing empirical formulas, a back propagation neural network(BPNN) drilling speed prediction model based on the moth-flame optimization(MFO)algorithm was proposed. Firstly, the basic principles of the MFO algorithm were studied, and the specific process for optimizing the BPNN using the MFO algorithm was constructed.Subsequently, drilling data obtained from a construction site in Wuxi, Jiangsu was collected and factors affecting drilling speed were analyzed.A series of methods such as wavelet threshold denoising, normalization, and grey correlation analysis were used to preprocess the collected data,resulting in a training and testing set.Then,the MFO algorithm was applied to train the weights and thresholds of the neural network, replacing the original gradient descent method, and an MFO-BPNN drilling speed prediction model was established.Finally, a detailed comparative analysis was conducted on the prediction results and evaluation indicators of the above prediction model, BPNN model,genetic algorithm optimization-back propagation neural network(GA-BPNN) model, andparticle swarm optimization-back propagation neural network(PSO-BPNN) model.The research results indicate that the reliability of the drilling speed prediction model established using the MFO-BPNN has reached 91.65%. Furthermore, its coefficients of determination (R2) is better than the other three prediction models.The three error indicators are also the lowest. This indicates that the prediction accuracy of the model is good, suitable for practical applications in pile foundation engineering,and provides a new idea for drilling speed prediction under the influence of complex factors.
Key words:  spiral drilling rigs; drilling speed prediction; moth-flame optimization(MFO) algorithm;  back propagation neural network(BPNN); genetic algorithm optimization-back propagation neural network(GA-BPNN); particle swarm optimization-back propagation neural network(PSO-BPNN); coefficients of determination (R2);  pile foundation engineering