FORECAST OF COMPRESSION INDEX OF FINE GRAINED SOIL BY UTILIZING ADAPTIVE NEURO FUZZY INFERENCE SYSTE

Page 1

e-ISSN: 2582-5208 International Research Journal of Modernization in Engineering Technology and Science Volume:02/Issue:09/September-2020

Impact Factor- 5.354

www.irjmets.com

FORECAST OF COMPRESSION INDEX OF FINE GRAINED SOIL BY UTILIZING ADAPTIVE NEURO FUZZY INFERENCE SYSTEM Mr. CH. Naga bharath*1, A. Adiseshu*2, B. Sai kumar*3, M. Anil kumar*4, A.V. Anil babu*5, B.V. Chandra Sekhar*6 *1M.Tech,

Assistant Professor, Civil Engineering, Gudlavalleru Engineering College, Gudlavalleru, India.

*2,3,4,5,6Student,

Civil Engineering, Gudlavalleru Engineering College, Gudlavalleru, India.

ABSTRACT The compression index is one of the important compressibility parameters to determine the settlement calculation for fine-grained soil layers. These parameters can be determined via carrying out lab oedometer test on undisturbed examples. Nonetheless, the test is very tedious and expensive. Subsequently, many exact recipes dependent on regression analysis have been presented to gauge the compressibility parameters using soil index properties. We are making an endeavor on Adaptive Neuro Fuzzy Inference System (ANFIS) model for prediction of compressibility parameters To manage the difficulties involved, in this task an endeavor has been made to show property of soil for example Compression index parameter in terms of Depth of foundation (Df), liquid limit (WL), and N value, Fine fraction (ff). The Neuro Fuzzy back propagation network is constructed to show the Compression index. For all the models chosen, Depth of foundation, liquid limit, N value and Fine fraction are taken as input parameters and compression index of soil as yield parameter from essential soil properties. For this reasons 1226 soils test information was gathered from the consultancy cell of Geotechnical Designing Lab, the test outcomes relating a wide scope of soil properties. Among the all out 979 datasets are utilized for training, 183 datasets for checking and staying 62 datasets for testing. Distinctive fluffy models were attempted with Df, N,WL, ff as input boundaries and Cc as output boundary. The architectures developed by varying membership functions for four input parameters are [5 4 4 5] (Gbellmf), [6 4 3 4] (Trapmf), [4 4 3 4] (Gbellmf), [3 3 3 4] (Gbellmf , trapmf), [3 3 4 4] (Gbellmf, trapmf). The best Neuro Fuzzy model is identified by analyzing the performance of different models developed. Adaptive Neuro Fuzzy Inference System model with [3 3 4 4] (gbellmf) architecture was found to be quite reasonable in predicting desired output Pictorial presentation of results gives a better idea than quantitative assessment. In this line another attempt has been made for checking the predicted values of compression index of soils with the observed values graphically. A graph is plotted between the predicted values and observed values of output for training and testing process which shows predicted values are close to observed values for all output parameters. Keywords: Adaptive Neuro Fuzzy Inference System, Compression index, Depth of foundation, Liquid limit, N, Fine fraction, Gbellmf, Trapmf.

I.

INTRODUCTION

Compression index of a dirt is the most unpredictable property to understand considering the large number of elements known to influence it. A great deal of maturity and ability might be required with respect to the specialist in deciphering the consequences of the laboratory tests for application to the conditions in the field. So as to adapt to the above complexities, customary types of building displaying approaches are reasonably streamlined. An elective methodology, which has given some guarantee in the field of geotechnical building, is Adaptive Neuro Fuzzy Inference Framework (ANFIS).In these examination Compression index (cc) for soils are predicted utilizing Adaptive Neuro Fuzzy Inference Framework (ANFIS). ANFIS model is created utilizing NN device in MATLAB programming (7.5.0). In the paper an endeavor has been made to demonstrate Compression index as far as Profundity of establishment (Df) Fine Division (FF), Fluid Cutoff (WL), and N esteem. An Adaptive Neuro Fuzzy

www.irjmets.com

@International Research Journal of Modernization in Engineering, Technology and Science

[493]


e-ISSN: 2582-5208 International Research Journal of Modernization in Engineering Technology and Science Volume:02/Issue:09/September-2020

Impact Factor- 5.354

www.irjmets.com

Inference Framework is utilized to display the compression index . The best fuzzy model is distinguished by investigating the exhibition of various models examined.

II.

ANFIS ARCHITECTURE

The ANFIS is a fuzzy Sugeno model placed in the structure of adaptive frameworks to encourage learning and variation. Such structure makes the ANFIS modeling more systematic and less dependent on master information. To introduce the ANFIS architecture, two fuzzy on the off chance that rules dependent on a first request Sugeno model are thought of: Rule 1: In the event that (x is A1) and (y is B1) at that point (f1 = p1x + q1y + r1) Rule 2: In the event that (x is A2) and (y is B2) at that point (f2=p2x + q2y + r2) where x and y are the inputs, A1 and B1 are the fuzzy sets, f1 are the outputs inside the fuzzy district indicated by the fuzzy standard, P1, q1 and r1 are the plan boundaries that are resolved during the training cycle. Fig. 1, represents the thinking instrument for this Sugeno model where it is the premise of the ANFIS model.

Fig.1: A two-input first-order Sugeno fuzzy

Fig.2: ANFIS architecture

model with two rules The ANFIS architecture to execute these two principles is appeared in Fig.2, in which a circle shows a fixed node, while a square demonstrates an adaptive node. Adaptive neuro fuzzy inference framework essentially has 5 layer architectures and every one of the function is clarified in detail later. In the main layer, all the nodes are adaptive nodes. The outputs of layer 1 are the fuzzy membership evaluation of the inputs. In the subsequent layer, the nodes are fixed nodes. They are labeled with π, showing that they proceed as a straightforward multiplier. In the third layer, the nodes are likewise fixed nodes labeled by N, to demonstrate that they assume a standardization part to the terminating qualities from the past layer. In the fourth layer, the nodes are adaptive. The output of every node in this layer is basically the result of the standardized terminating quality and a first request polynomial (for a first request Sugeno model). In the fifth layer, there is just one single fixed node labeled with Σ. This node plays out the summation of every approaching sign. Fig .3 shows the variety in the Sugeno model that is proportionate to a two-input first-request Sugeno fuzzy model with nine principles, where each input is expected to have three related MFs. Fig .4 represents how the two dimensional input space is divided into nine covering fuzzy districts, every one of which is administered by a fuzzy if-then standard. As such, the reason part of a standard characterizes a fuzzy district, while the subsequent part specifies the output inside the area.

www.irjmets.com

@International Research Journal of Modernization in Engineering, Technology and Science

[494]


e-ISSN: 2582-5208 International Research Journal of Modernization in Engineering Technology and Science Volume:02/Issue:09/September-2020

Impact Factor- 5.354

www.irjmets.com

Fig.3: Two-input first-order Sugeno fuzzy model with nine rules

Fig.4: The input space that are partitioned into nine fuzzy regions

III.

DATA USED FOR TRAINING AND TESTING

It has a sum of 1226 information with 4 inputs and 1 output. The information is separated into 3 segment which are training, checking and application. Training information has 979 examples, checking takes around 183 examples and the rest which is 62 examples are utilized for testing purposes. The run of the mill standardized information utilized for training stage is introduced in Table 1 and in Table 2 presents the ordinary standardized information utilized for testing stage. As this venture completely utilize the MATLAB programming, it is prudent to know its information the board framework. To work with mfiles and information, all records are put in an envelope remembered for the MATLAB working way. The working way is an assortment of organizers where MATLAB look through documents when these are called. Any document situated external this way will be undetectable to MATLAB. Table-1: Indicating typical input and output values of training data Liquid

Density

N value

Fine fraction

Compression index (%)

Limit (%)

(gm/cc)

45

1.54

9

87

0.12

54

1.502

13

87

0.135

45

1.649

19

92

0.125

40

1.656

33

88

0.122

40

1.675

34

82

0.11

www.irjmets.com

(%)

@International Research Journal of Modernization in Engineering, Technology and Science

[495]


e-ISSN: 2582-5208 International Research Journal of Modernization in Engineering Technology and Science Volume:02/Issue:09/September-2020

Impact Factor- 5.354

www.irjmets.com

36

1.69

45

88

0.105

40

1.695

42

88

0.107

36

1.705

57

86

0.105

50

1.657

7

86

0.120

Table-2: Indicating typical input and output values of checking data Liquid limit (%)

Density

N value

(gm/cc)

Compression index

(%)

(%)

47

1.183

17

85

0.333

71

1.52

19

85

0.549

35

1.33

24

48

0.225

24

1.32

25

43

0.126

20

1.36

21

19

0.09

81

1.29

29

69

0.639

35

1.26

25

91

0.225

34

1.35

23

57

0.216

25

1.32

28

61

0.135

18

1.39

21

11

0.072

70

1.29

19

92

0.54

37

1.3

20

80

0.243

81

1.28

18

94

0.639

37

1.28

6

86

0.243

54

1.35

23

59

0.396

84

1.31

18

63

0.666

IV.  

Fine fraction

RESULTS AND DISCUSSION

The comparision among observed and predicted compression index esteems are appeared in table 3. The chart plotted among observed and predicted compression index esteems are appeared in fig .5. Table-3: Observed Values Vs Predicted Values Liqui d Limit (%) 29 29 23 26 25 30

www.irjmets.com

Dry Density (gm/cc)

N value

Fine fraction (%)

Observed Compression Index (%)

Predicted Compression Index (%)

1.047 1.052 1.063 0.975 1.001 0.748

19 26 30 12 10 22

61.3 43.4 46 81 54.2 49.6

0.171 0.171 0.117 0.144 0.135 0.18

0.157 0.1756 0.1305 0.0981 0.1281 0.1268

@International Research Journal of Modernization in Engineering, Technology and Science

[496]


e-ISSN: 2582-5208 International Research Journal of Modernization in Engineering Technology and Science Volume:02/Issue:09/September-2020 Liqui d Limit (%) 29 47 68 59 92 64 39 38 35 35 30 34 35 31 35 34

Impact Factor- 5.354

www.irjmets.com

Dry Density (gm/cc)

N value

Fine fraction (%)

Observed Compression Index (%)

Predicted Compression Index (%)

0.753 0.906 0.812 0.785 0.872 0.695 0.852 0.963 0.758 0.803 0.741 0.652 0.752 0.771 0.854 0.778

8 26 26 6 16 11 10 14 6 8 8 8 2 5 5 2

62.2 62.6 96.5 96.8 99.7 91.6 77 72.5 86.6 92.3 75.6 94.6 93.4 82.6 86.3 77.9

0.171 0.333 0.522 0.441 0.738 0.486 0.261 0.252 0.225 0.225 0.18 0.216 0.225 0.189 0.225 0.216

0.1972 0.2938 0.5599 0.4453 0.7324 0.4514 0.2962 0.2527 0.2518 0.2329 0.1922 0.2279 0.2271 0.1678 0.2897 0.2289

Fig 5: Calculated vs. Predicted Table-4: Results of ANFIS Models Model

M

Rules

Epochs

1 2 3

[5 4 4 5] [6 4 3 4] [4 4 3 4]

400 288 192

1000 500 500

4

[3 3 3 4]

108

500

5

[3 3 4 4]

144

500

www.irjmets.com

Membership Functions (mf) Gbellmf Trapmf Gbellmf 1,2 – gbellmf 3,4 – Trapmf 1,2 - gbellmf 3,4 – Trapmf

TIME TAKEN (Hours) 15.9 12.8 11.0

0.656 0.607 0.668

9.8

0.864

10.4

0.954

@International Research Journal of Modernization in Engineering, Technology and Science

[497]


e-ISSN: 2582-5208 International Research Journal of Modernization in Engineering Technology and Science Volume:02/Issue:09/September-2020

V.

Impact Factor- 5.354

www.irjmets.com

CONCLUSION

Among the all out 979 datasets are utilized for training, 183 datasets for checking and staying 62 datasets for testing. Distinctive fuzzy models were attempted with Df, N,WL, ff as input boundaries and Cc as output parameter. The important training and testing of information was conveyed with the proposed model. Based on this a [3 3 4 4] Gauss bell Architecture function is proposed. This is the principal model for predicting the Compression index of soils utilizing Adaptive Neuro Fuzzy Inference Framework.

VI.

REFERENCES

[1]

Hyun Il Park, Seung Rae Lee,( 2011), Evaluation of the compression index of soils using an artificial neural network. Computers and Geotechnics 38 (2011) 472–481.

[2]

Rani Sudha, Phani Kumar Vaddi, N.V.Vamsi Krishna Togati, 2013, Artificial Neural Networks (ANNS) For Prediction of Engineering Properties of Soils. International Journal of Innovative Technology and Exploring Engineering (IJITEE) ISSN: 2278-3075, Volume-3, Issue-1

[3]

Sridharan, A. and Nagaraj, H.B. 2000. Compressibility Behaviour of Fine-Grained Soils and Correlation with Index Properties. Canadian Geotechnical Journal, 37(3): 712–722.

[4]

Ghabousi J, Garett JR, Wu X, ―Knowledge based modeling of material behavior with neural networks”, ASCE J EngMech 1991; 117(1):132–53.

[5]

Kelvin.L.Priddy and Paul.E.Keller, EconomyEdition,2007).

[6]

M.A. Shahin, M.B. Jaksa, H.R. Maier, ―Artificial neural network applications in geotechnical engineering, Australian Geomechanics 36 (1) (2001) 49–62.

[7]

T.P. Thaker, K.S. Rao, “DEVELOPMENT OF STATISTICAL CORRELATIONS BETWEEN SHEAR WAVE VELOCITY AND PENETRATION RESISTANCE USING MASW TECHNIQUE”, Geo- Innovation Addressing Global Challenges, Toronto, Ontario, Canada,2011.

[8]

Ch. SubbaRao, “ESTIMATION OF SHEAR WAVE VELOCITY FROM SOIL INDICES”, Indian Geotechnical Journal, Springer, Volume 43, Issue 3, September 2013, (PP:267-273)

[9]

Pernot S, Lamarque CH, ―Application of neural networks to the modeling of some constitutive laws‖, Neural Networks 1999;12:371–92.

―Artificial

Neural

Networks‖,

Introduction

(Eastern

[10] Amardeep Singh And Sahid Noor, “Soil compression index prediction model for fine grained soils” International Journal of Innovations in Engineering and Technology, Vol 1Issue 4 Dec 2012. [11] Mosavi A. and M. Edalatifar. A Hybrid Neuro-Fuzzy Algorithm for Prediction of Reference Evapotranspiration. in International Conference on Global Research and Education. 2018.Springer. [12] Demuth H, Beale M, Hagan M (2007) Neural network toolbox user’s guide for use with matlab. Math works Inc, Natick Dipova N, Cangir B (2010). Teknik Dergi 21(3):5069–5086 (in Turkish). [13] D.W.C. Ho, and P.A. Zhang, “Design of Fuzzy Wavelet Neural Networks using the GA approach for function approximation and system identification”, J. Xu, IEEE Transactions on Fuzzy Systems, 2001, vol.9, pp. 200–211. [14] Kumar, V. P., and Rani, C. S., 2011, Prediction of Compression Index of Soils Using Artificial Neural Networks (ANNs), International Journal of Engineering Research and Applications (IJERA), Vol. 1, Issue 4, pp. 1554-1558.

www.irjmets.com

@International Research Journal of Modernization in Engineering, Technology and Science

[498]


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.