Application of Taguchi method in reducing the number of experiments and optimizing the factors of artificial neural network related to the phenomenon of design of stable size of RipRap around bridge piers

Document Type : Research Article

Authors

1 Dept. of Civil Engineering, University of Mohaghegh Ardabili. Civil Engineering-Water Resource Management and Engineering Graduated M.Sc.

2 Associate Professor, Dept. of Civil Engineering, Faculty of Engineering, University of Mohaghegh Ardabili, Ardabil, Iran

3 Assistant Professor of Hydraulic Engineering in Shahid Rajaee Teacher Training University

Abstract

Application of Taguchi method in reducing the number of experiments and optimizing the factors of artificial neural network related to the phenomenon of design of stable size of Riprap around bridge piers
Extended Abstract
Introduction
Hydraulic phenomena are generally studied in the laboratory, so it seems necessary to use a method that reduces the number of experiments and saves time and time with good accuracy, and Taguchi method is one of these methods. In this regard (Dalir et al., 2021), Taguchi method and response level in the laboratory model of the pond were used to evaluate the efficiency of trapping and sedimentation and introduced the effective parameters. (Ranjbar-Zahedi et al., 2021) To reduce local scouring around bridge piers and to minimize the number of experiments, they used the 27 proposed Taguchi experiments to determine the optimal size and location of the structure (Atarodi et al., 2020) for Design of decomposition geometric parameters using Taguchi and Taguchi-GRA methods. In this study, in addition to the Taguchi property in reducing the number of studies in the study of the stable size of the riprap around the bridge piers (in the experimental part the number of experiments and in the neural network the number of responses), the optimization and prediction of Taguchi (a property that is less studied Taken.) Be used. Then, for validation, the results of Taguchi prediction were compared with the results of artificial neural network.
Methodology
Taguchi method was studied as one of the experimental design methods based on reducing the number of experiments and proposing different but limited compounds for studies. Also, the analysis of the results has been examined using the mean mean graph and the signal to noise ratio (S/N) and the best combination of parameters has been introduced. Also, using the ANOVA table, the effect of the levels of each parameter and finally the effective parameters have been determined. In order to evaluate the results, the results of Taguchi method with the results of ANN artificial neural network were evaluated using the predictive property. It should be noted that Taguchi method has been used to adjust the adjustable parameters of the neural network for the phenomenon of stone crushing failure. In designing the stable size of the riprap around the bridge piers to protect against scour, 145 research laboratory data (Karimaei Tabarestani and Zarrati., 2013) have been used and according to these data, four parameters including flow rate in six levels, rock size, ratio The length to the width of the bridge pier and the angle of the pitch relative to the flow direction were examined at three levels. Also, the adjustable components of the neural network, including the four components of the number of neurons in the first and second hidden layers, the training function and the transmission function in each layer were examined at three levels. Finally, the results of the neural network were compared with the results of the Taguchi prediction.
Results and Discussion
In designing a stable size of riprap around bridge piers to protect against scouring, the Taguchi method with a reduction of 87 and 89%, respectively, compared to the results (Karimaei Tabarestani and Zarrati., 2013) and the complete factorial method, reduces the number of tests and saves time and money. have been. In the analysis of the results using Taguchi, it was found that the maximum flow depth for crimping stability will occur when the flow rate is 0.06 m3 / s, d50 is 0.00205 m, L is 35 m and θ = 20 is 20 Degree and in this regard the most effective parameter Q was introduced. Also, the best ANN artificial neural network based on the optimal combination introduced by Taguchi based on S/N diagram analysis, with three layers and correlation coefficient (R) equal to 0.971, will occur when the first and second hidden layers each contain 7 The neuron is a training function of trainlm and the transmission function of each layer is tansig. Also, according to ANOVA analysis, the most effective factor with a high rate of 95.07% participation is the transfer function. Finally, the Taguchi-assisted neural network with a detection coefficient of 0.94 performed better than the Taguchi method with a detection coefficient of 0.79 in predicting the results of designing a stable size of riprap around bridge piers to protect against scouring.
Conclusion
The results of the present study show that by using the Taguchi orthogonal array table and analyzing its results in the experimental section, the optimal combination of parameters can be determined with only a small number of experiments and the optimal solution can be predicted. Also, Taguchi method is a more suitable alternative to trial and error method for adjusting neural network parameters, and the results obtained from designing neural network parameters with Taguchi method are of good accuracy.
Keywords
Taguchi Method - Artificial Neural Network - Bridge Piers- Riprap - Scour.

Keywords


Atarodi, A., Karami, H., Ardeshir, A. and Hosseini, Kh. (2020). Optimization of the Geometric Parameters of the Protective Spur Dike using Taguchi Method and GRA. Journal of Water and Soil Science. 1(24), 13-26. (In Persian)
Beeravelli, V.N., Chanamala, R., Rayavarapu, U.M. R. and Kancherla, P.R. (2018). An Artificial Neural Network and Taguchi Integrated Approach to the Optimization of Performance and Emissions of Direct Injection Diesel Engine. European Journal of Sustainable Development Research, 2(2), 16. https://doi.org/10.20897/ejosdr/85412.
Braddock, R.D., Kremmer, M.L. and. Sanzogni, L. (1998). Feed‐forward artificial neural network model for forecasting rainfall run‐off. Environmetrics: The official journal of the International Environmetrics Society, 9(4), 419-32.
Dalir, M., Ziaei, A. and Sheikh Rezazadeh Nikou, N. (2021). Investigation of Trapping, Sedimentation and Volumetric Fraction Efficiency of Vortex Settling Basin Using Taguchi Method. Iranian Journal of Soil and Water Research. 52(5), 1337-1350. (In Persian)
Demuth, H. and Beale, M. (1992). Neural Network Toolbox: User's Guide: for Use with Matlab. MathWorks Incorporated.
Kant, S. (2017). Application of Taguchi OA array and Artificial Neural Network for Optimizing and Modeling of Drilling Cutting Factors. International Journal of Theoretical and Applied Mechanics, 12(1), 1-2.
Karimaei Tabarestani, M. and Zarrati, A.R. (2013). Design of stable riprap around aligned and skewed rectangular bridge piers. Journal of Hydraulic Engineering, 139(8), 911-6.
Karimaei Tabarestani, M. and Zarrati, A.R. (2019). December. Local scour depth at a bridge pier protected by a collar in steady and unsteady flow. Water Management, 172(6), 301-311.
Karimaei Tabarestani, M. (2020). Effect of correlation between hydraulic parameters on reliability analysis of designed riprap around bridge pier, Journal of Water and Soil Science, 14 (4), 51-68. (In Persian)
Kia, M. (2010). Neural Network in Matlab. Kian Rayane Sabz Publications. (In Persian)
Menhaj, M.B. (2018). Fundamentals of Neural Networks. Volume 1: Computational Intelligence. Amirkabir University of Technology. (In Persian)
Ranjbar-Zahedani, M., Keshavarzi, A., Khabbaz, H., Ball, J.E. (2021). Optimizing flow diversion structure as an effective pier-scour countermeasure. Journal of Hydraulic Research. 59(6), 963-976. https://doi.org/10.1080/00221686.2020.1862321.
Rashno, E., Zarrati A.R. and Karimaei Tabarestani, M. (2020). Design of riprap for bridge pier groups. Canadian Journal of Civil Engineering, 47, 516-522.
Razavizadeh, S. and Dargahian, F. (2018). Optimization of Artificial Neural Network Structure in Prediction of Sediment Discharge Using Taguchi Method. Iran-Watershed Management Science & Engineering. 43(12), 89-97. (In Persian)
Rezazadeh, R., Barani, G.A. and Naseri, A. (2019). Application of artificial neural networks in estimation of scour depth around the bridge pier with sticky sediments. (Research Note). Journal.of Hydraulics. 14(1), 141-149. (In Persian)
Rostamabadi, M., Salehi Neyshabouri, A.A and Zarrati, A.R. (2013). Optimization of Geometric Parameters of Submerged Vane in Straight Alluvial Channel with Taguchi Method and GRA. Modares Civil Eng. Journal. 13, 79-93. (In Persian)
Rostamabadi, M. (2017). Determining the optimal value of height and position of calming blocks using studies designed by Taguchi and complete factorial methods. Journal of Hydraulics. 12(2), 35-44. (In Persian)
Roy, R. (1990). A primer on the Taguchi method. Society of Manufacturing Engineers, New York.
Soukhtanlou, E., Teymourtash, A.R and Mahpeykar, M.R. (2018). Proposal of experimental relations for determining the number of sides of polygonal hydraulic jumps. Modares Mechanical Engineering. 18(1), 273-280. (In Persian)
Taguchi, G., Chowdhury, S. and Wu, Y. (2005). Taguchi's quality engineering handbook. Wiley.
Yao, A.W. and Chi, S.C. (2004). Analysis and design of a Taguchi–Grey based electricity demand predictor for energy management systems. Energy Conversion and Management, 45(7-8), 1205-17.
Zanjirchi, S.M., Hatami, M.M., Kadkhodazadeh, H.R and Banifatemeh, S.A.M. (2015). Improving the efficiency of forecasting productivity, using a Taguchi experiment design approach (case study: food industries in Iran). Productivity Management (Beyond Management). 8(32), 69-87. (In Persian).
  • Receive Date: 10 June 2021
  • Revise Date: 14 August 2021
  • Accept Date: 21 August 2021
  • First Publish Date: 21 August 2021