Environmentally sensitive blasting design based on risk analysis by using artificial neural networks


Ozer U., Karadogan A., Ozyurt M. C. , Sahinoglu U. K. , Sertabipoglu Z.

ARABIAN JOURNAL OF GEOSCIENCES, cilt.12, 2019 (SCI İndekslerine Giren Dergi) identifier identifier

  • Cilt numarası: 12 Konu: 2
  • Basım Tarihi: 2019
  • Doi Numarası: 10.1007/s12517-018-4218-7
  • Dergi Adı: ARABIAN JOURNAL OF GEOSCIENCES

Özet

The aim of this study is to develop an artificial neural network (ANN) which can design an environmentally sensitive blasting project and predict peak particle velocity (PPV) for an urban foundation excavation project with risk elements having different vibration-carrying capacities. In the study area, there are risk factors with different vibration capacities such as revetment systems and ongoing and completed reinforced concrete structures. It is mandatory to use the PPV limit values specified in the Turkish norm when assessing damage to the completed buildings. However, the vibration-carrying capacities of all structures in Turkish norm are accepted as the same. This situation may pose a risk to the buildings under construction. This risk has been avoided by using Jimeno et al. approach, where PPV limit values vary according to the type of buildings and the concrete setting times. The evaluation of different risk factors according to different damage criteria has made blasting excavation activities a complicated problem. In order to solve this problem, an ANN was used which knows the damage criteria that should be based on the element of risk and the geological and rock properties of the site. At the same time, the ANN can predict the blasting designs to be applied according to the element of risk, concrete setting times, and the distance to the risk point and can estimate the PPV to be occurred. Site-specific vibration propagation equation has been obtained as a result of the test shots. Using this equation, the maximum charge amounts per delay were calculated in different regions of the field, and different designs were proposed accordingly. ANN was trained with the samples representing the test shots, and the proposed designs and the performance were evaluated. The outputs of the ANN model, which can learn the problem and provide high accuracy estimates, were applied at 37 shots. PPV values measured at 37 shots were below the damage limits. This shows that the network is capable of the geological and rock properties of the site, and outputs that can represent vibration-carrying capacities of elements of risk. As a result, it is understood that ANN was found to be an effective tool in solving complex problems such as in this study.

The aim of this study is to develop an artificial neural network (ANN) which can design an environmentally sensitive blasting project and predict peak particle velocity (PPV) for an urban foundation excavation project with risk elements having different vibration-carrying capacities. In the study area, there are risk factors with different vibration capacities such as revetment systems and ongoing and completed reinforced concrete structures. It is mandatory to use the PPV limit values specified in the Turkish norm when assessing damage to the completed buildings. However, the vibration-carrying capacities of all structures in Turkish norm are accepted as the same. This situation may pose a risk to the buildings under construction. This risk has been avoided by using Jimeno et al. approach, where PPV limit values vary according to the type of buildings and the concrete setting times. The evaluation of different risk factors according to different damage criteria has made blasting excavation activities a complicated problem. In order to solve this problem, an ANN was used which knows the damage criteria that should be based on the element of risk and the geological and rock properties of the site. At the same time, the ANN can predict the blasting designs to be applied according to the element of risk, concrete setting times, and the distance to the risk point and can estimate the PPV to be occurred. Site-specific vibration propagation equation has been obtained as a result of the test shots. Using this equation, the maximum charge amounts per delay were calculated in different regions of the field, and different designs were proposed accordingly. ANN was trained with the samples representing the test shots, and the proposed designs and the performance were evaluated. The outputs of the ANN model, which can learn the problem and provide high accuracy estimates, were applied at 37 shots. PPV values measured at 37 shots were below the damage limits. This shows that the network is capable of the geological and rock properties of the site, and outputs that can represent vibration-carrying capacities of elements of risk. As a result, it is understood that ANN was found to be an effective tool in solving complex problems such as in this study.