Applying improved artificial neural network models to evaluate drilling rate index

The drilling rate index (DRI) is the most important input parameter of a commonly used performance prediction model for drilling and rock excavation. In this paper, the hybrid artificial neural network (ANN) with back propagation (BP) algorithm, simulated annealing algorithm (SAA), firefly algorithm...

Full description

Saved in:
Bibliographic Details
Published in:Tunnelling and underground space technology Vol. 70; pp. 114 - 124
Main Authors: Fattahi, Hadi, Bazdar, Habibollah
Format: Journal Article
Language:English
Published: Oxford Elsevier Ltd 01.11.2017
Elsevier BV
Subjects:
Artificial neural network
Artificial neural networks
Back propagation
Back propagation networks
Compressive strength
Computer simulation
Correlation coefficients
Drilling
Drilling rate index
Errors
Firefly algorithm
Heuristic methods
Invasive weed optimization algorithm
Mathematical models
Neural networks
Optimization algorithms
Performance indices
Performance prediction
Scleroscope hardness
Shuffled frog leaping algorithm
Simulated annealing
Simulated annealing algorithm
Tensile strength
Simulated annealing algorithm
Drilling rate index
Artificial neural network
Invasive weed optimization algorithm
Shuffled frog leaping algorithm
Firefly algorithm
ISSN:0886-7798, 1878-4364
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The drilling rate index (DRI) is the most important input parameter of a commonly used performance prediction model for drilling and rock excavation. In this paper, the hybrid artificial neural network (ANN) with back propagation (BP) algorithm, simulated annealing algorithm (SAA), firefly algorithm (FA), invasive weed optimization algorithm (IWO) and shuffled frog leaping algorithm (SFLA) were used to build a prediction model for the indirect estimation of DRI. The estimation abilities offered using five ANN models (ANN-BP, ANN-SAA, ANN-FA, ANN-IWO and ANN-SFLA) were presented by using available data given in open source literature. In these models, strengths (Uniaxial Compressive Strength (UCS) and Brazilian Tensile Strength (BTS)) and indexes properties (Shore Scleroscope Hardness (SSH), diametral point load strength index (Is(50) →) and axial point load strength index (Is(50)↓)) were utilized as the input parameters, while the DRI was the output parameter. Various statistical performance indexes were utilized to compare the performance of those estimation models. The comparative results revealed that hybrid of SAA and ANN yield robust model which outperform other models in term of higher squared correlation coefficient (R2), variance account for (VAF) and lower mean square error (MSE), root mean squared error (RMSE) and mean absolute percentage error (MAPE).
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0886-7798
1878-4364
DOI:10.1016/j.tust.2017.07.017