the estimation of tbm penetration rate using artificial neural network optimized with particle swarm optimization and firefly algorithms, case study: tabriz metro line 2

The penetration rate (pr) is a critical parameter in tunnelling, as it directly determines project timelines, cost, and overall efficiency. developing accurate predictive models for the penetration rate (pr) is crucial for optimizing tunnelling performance and enabling more effective project planning. to meet this need, this study employs advanced metaheuristic optimization algorithms to augment an artificial neural network (ann) for improved penetration rate (pr) prediction. specifically, particle swarm optimization (pso) and the firefly algorithm (fa) were employed to refine the model's accuracy. the research utilized data from the tabriz metro line 2 project. the data integrated key influencing factors, which were categorized as follows: geological parameters, including soil friction angle, cohesion, unit weight, shear modulus, and water table depth; and machine parameters, including torque, thrust force, and rotational speed. the explicit goal of the model's optimization was to minimize the normalized mean squared error (nmse) for its predictions against the actual measured values. the results demonstrate that both pso and fa significantly enhanced the predictive performance of the baseline ann model. however, the firefly algorithm proved superior, achieving a higher coefficient of determination (r² = 0.836 for test data, compared to 0.780 for the pso-optimized model) and a lower nmse. this key outcome is attributed to the fa's robust search capabilities, confirming its effectiveness in identifying optimal model parameters for complex, nonlinear relationships in tunnelling. the findings provide a reliable, data-driven framework for predicting tbm performance, offering substantial practical value for project planning and execution in geotechnical engineering.

the estimation of tbm penetration rate using artificial neural network optimized with particle swarm optimization and firefly algorithms, case study: tabriz metro line 2

the penetration rate (pr) is a critical parameter in tunnelling, as it directly determines project timelines, cost, and overall efficiency. developing accurate predictive models for the penetration rate (pr) is crucial for optimizing tunnelling performance and enabling more effective project planning. to meet this need, this study employs advanced metaheuristic optimization algorithms to augment an artificial neural network (ann) for improved penetration rate (pr) prediction. specifically, particle swarm optimization (pso) and the firefly algorithm (fa) were employed to refine the model's accuracy. the research utilized data from the tabriz metro line 2 project. the data integrated key influencing factors, which were categorized as follows: geological parameters, including soil friction angle, cohesion, unit weight, shear modulus, and water table depth; and machine parameters, including torque, thrust force, and rotational speed. the explicit goal of the model's optimization was to minimize the normalized mean squared error (nmse) for its predictions against the actual measured values. the results demonstrate that both pso and fa significantly enhanced the predictive performance of the baseline ann model. however, the firefly algorithm proved superior, achieving a higher coefficient of determination (r² = 0.836 for test data, compared to 0.780 for the pso-optimized model) and a lower nmse. this key outcome is attributed to the fa's robust search capabilities, confirming its effectiveness in identifying optimal model parameters for complex, nonlinear relationships in tunnelling. the findings provide a reliable, data-driven framework for predicting tbm performance, offering substantial practical value for project planning and execution in geotechnical engineering.