Keshavarz, Z., Torkian, H. (2018). Application of ANN and ANFIS Models in Determining Compressive Strength of Concrete. Soft Computing in Civil Engineering, 2(1), 62-70. doi: 10.22115/scce.2018.51114

Zahra Keshavarz; Hojjatollah Torkian. "Application of ANN and ANFIS Models in Determining Compressive Strength of Concrete". Soft Computing in Civil Engineering, 2, 1, 2018, 62-70. doi: 10.22115/scce.2018.51114

Keshavarz, Z., Torkian, H. (2018). 'Application of ANN and ANFIS Models in Determining Compressive Strength of Concrete', Soft Computing in Civil Engineering, 2(1), pp. 62-70. doi: 10.22115/scce.2018.51114

Keshavarz, Z., Torkian, H. Application of ANN and ANFIS Models in Determining Compressive Strength of Concrete. Soft Computing in Civil Engineering, 2018; 2(1): 62-70. doi: 10.22115/scce.2018.51114

Application of ANN and ANFIS Models in Determining Compressive Strength of Concrete

^{1}Graduate Student, Department of Civil Engineering, Islamshahr Branch, Islamic Azad University Islamshahr, Iran

^{2}Faculty Member, Department of Civil Engineering, Islamshahr Branch, Islamic Azad University Islamshahr, Iran

Receive Date: 15 September 2017,
Revise Date: 24 September 2017,
Accept Date: 18 October 2017

Abstract

Concrete compressive strength is recognized as one of the most important mechanical properties of concrete and one of the most significant mechanical properties in determining the quality of the produced concrete. Since the traditional procedures of determining the compressive strength of concrete require time and cost, scholars have always been looking for new methods to replace them with existing traditional methods. In this paper, soft computing methods are investigated for determining the compressive strength of concrete. To be specific, 150 different concrete specimens with various mix design parameters have been built in the laboratory, and the compressive strength of them have been measured after 28 days of curing in the water. Five different concrete mix parameters, (i.e. cement, water to cement ratio, gravel, sand, and microsilica) were considered as input variables. In addition, two soft computing techniques have been used in this study which are Artificial Neural Network (ANN) and Adaptive Neuro Fuzzy Inference (ANFIS) System. Results have shown that both of ANN and ANFIS models are successful models for predicting the compressive strength of concrete. Also, results have shown that ANFIS is more capable than ANN in predicting the compressive strength of concrete.

Nowadays, the use of soft computing techniques by different researchers are increased. Soft computing techniques, also known as data driven models, are models based on the computational modeling and work based on input-output data. Using these methods would result in saving a significantamount of time and cost, besides the accuracy of these models. Among all, Artificial Neural Network (ANN) and Adaptive Neuro Fuzzy Inference System (ANFIS) are used in multiple occasions by scientists. The ANN technique is composed of multiple nodes which imitates biological neurons in the human brain. The neurons connect to each other through links. The ANFIS is a kind of Artificial Neural Network that is based on Takagi-Sugeno fuzzy inference system.

Scientists have utilized the ANFIS and ANN models in different area of science successfully. Krishna et al. have successfully used ANN and ANFIS in studying the fluidized bed with internals [1]. Hamdia et al. have effectively used these two models in predicting the fracture toughness of PNCs [2]. Naderpour et al. developed a new approach to obtain the FRP-confined compressive strength of concrete using a large number of experimental data by applying artificial neural networks [3]. Khademi et al. have used the ANN and ANFIS models efficiently in determining the displacement in concrete reinforcement buildings [4]. Gupta et al. have skillfully used these models in estimating the performance of plate fin heat exchanger by exploration [5]. Talebizadeh and Moridnejad have performed the uncertainty analysis on the forecast of lake level fluctuations [6]. Khademi et al. have estimated the 28-days compressive strength of concrete through ANN and ANFIS models [7]. Zemgyi Ma et al. have used these models to classify the heating value of burning municipal solid waste in circulating fluidized bed incinerators [8]. Güneyisi et al. have successfully predicted the flexural over strength factor for steel beams using artificial neural network [9]. Nasrollahi has skillfully predicted the optimum shape of large-span trusses according to AISC-LRFD using Ranked Particles Optimization [10]. The aforementioned statements were examples of application of successfully used different soft computing methods for predicting various civil engineering characteristics. Keshavarz provided a summary of the literatures in which the civil engineering characteristics were predicted through soft computing models [11].

In order to get to the objective of this study, various concrete specimens were made in the laboratory and cured in water for 28 days. In overall, 150 different concrete specimens were constructed in the laboratory. The concrete specimens were made in cylindrical shape with the diameter of 150 mm and height of 300 mm. The input parameters influencing the compressive strength of concrete were cement, gravel, sand, microsilica, and water to cement ratio. The characteristics of these data is shown in Table 1.

Where is the experimental strength of i^{th} sample, is the averaged experimental strength, is the determined compressive strength of ith sample, and is the averaged determined compressive strength.

3. Artificial neural network

Artificial neural networks is famous as data processing systems which include the artificial neurons. These artificial neurons are a vast number of simple, highly incorporated processing elements inspired by the complex structure of the brain. For the purpose of improving their performance, these neurons have the ability to learn from the experiences. Artificial neural network can be identified as a network comprise of several processors which are called neurons (also known as units). Each neurons include a numerical data which could be known as weights [12, 13].

In order to build a useful neural network, 3 steps should be considered: (1) Selection of appropriate architecture for the artificial neural network, (2) using sufficient training data for creating the network, and (3) Using different test data groups to test the accuracy of the network. Figure 2 shows the architecture of artificial neural network model [13, 14].

Fig. 2. Architecture of Artificial Neural Network (adapted from [10]).

The connection strength is estimated through weighted connections. These weights are trained to make the output variables as close as possible to target values. ANN is consist of three different steps of train, validation, and test. The training step goal is to minimize the error function. In the validation step, the artificial neural network is used to construct the model and it works independently from the training step. The test step is used to anticipate the accuracy of the machine algorithm [14].

Neural networks are categorized into layers. The input layer and output layers are consist of input data and the resulting output data. Between the input and output layers, the hidden layer exists which includes neurons and are connected by the weights, which are explained earlier. There might be any number of hidden layers, and the two-layer network may map any number of non-linear relationship. Each layer is a vector containing any number of R of neurons, and the output of the layer is a vector of length R containing the output from each neuron in that layer. This output vector is next passed as the input vector for the next hidden layer, and this specific process continues for all hidden layers till the final output of the network is reached [15, 16].

The concrete specimens that are studied in this research have the compressive strength of between 200 Kg/m^{3} to 350 Kg/m^{3}. The total of 150 specimens have been studied in this research. In order to have the better understanding of the concrete specimens used in this study, the number of specimens in each intervals are presented in Table 2.

Table 2.

Number of Specimens in Each Interval.

Interval of Concrete Compressive Strength (Kg/m^{3})

Number of Specimens in each Interval

150-200

39

200 to 250

28

250 to 300

32

300 to 350

51

In this study, these total of 150 data are divided into three categorization of train, validation, and test, for the ANN modeling. This categorization is shown in Table 3.

Table 3.

Categorization of Studied Data in ANN Modeling.

Step

Percentage

Number of Specimens

Train

60%

90

Validation

10%

15

Test

30%

45

Total

100%

150

Different algorithm were ran in this study to find the best fitted model, and among all, the Levenberg- Marquardt (LM) was chosen. The results of the test step for the ANN modeling of these specimens is shown in Figure 3. This Figure demonstrates the relationship between the measured and estimated compressive strength of concrete.

Fig. 3. Comparison of Measured and Anticipated Data for the Test Step of ANN Model.

According to this Figure, the R^{2} coefficient for ANN modeling of the test data is equal to 0.942. This number confirms that the ANN model is a suitable model in predicting the compressive strength of concrete.

4. Adaptive Neuro-Fuzzy Inference System

Adaptive Neuro-Fuzzy Inference System also known as ANFIS incorporates the self- learning ability of neural networks with the linguistic expression function of fuzzy inference. In order to train Takagi- Sugeno type fuzzy inference system which would result in search for the optimal elements, ANFIS combines the least squares and back propagation gradient. In other words, ANFIS is a multilayer feed-forward network in which each node performs a particular function on receiving signals and has a set of parameters pertaining to this node. Similar to ANN, ANFIS is capable of mapping unseen inputs to their outputs by learning the rules from previously observed data. The architecture of ANFIS is shown in Figure 4 [4, 17, 18].

Fig. 4. Architecture of ANFIS Model (Adapted from [4]).

The ANFIS model includes 5 different layers. Each layer includes different nodes described by the node function. Layer one is the layer in which all the nodes are adaptive nodes with a node function. Layer two is the layer in which node multiplies incoming signals and the output is the product of all the incoming signals. Layer 3 calculates the ratio of the i^{th} rules firing strength to the sum of all rule’s firing strength of the nodes. In layer 4, each node calculates the contribution of the i^{th} rule to the overall output. In layer 5, the signal node calculates the final output as the summation of all input signals [4, 17-21].

The ANFIS model is divided into three steps of train, check, and train. The portion of data used in each step in ANFIS modeling is shown in Table 4.

Table 4.

Categorization of Studied Data in ANFIS Modeling.

Step

Percentage

Number of Specimens

Train

60%

90

Check

10%

15

Test

30%

45

Total

100%

150

The results of the test step for the ANFIS modeling of these specimens is shown in Figure 5. This Figure shows the correlation between the measured and estimated compressive strength of concrete.

Fig. 5. Comparison of Measured and Anticipated Data for the Test Step of ANFIS Model.

According to Figure 5, the R^{2} coefficient for ANFIS modeling of the test data is equal to 0.923. This number confirms that the ANFIS model is a suitable model in predicting the compressive strength of concrete.

5. Comparison of ANN and ANFIs

In order to better illustrate the efficiency of ANFIS and ANN models, there is the need to compare their results with each other. One of the well-known factors that could be used in such prediction comparisons is the coefficient of determination, defined in EQ (1). The coefficient of determination is simply a statistic that gives some information about the goodness of fit of a model. In other words, it is a statistical measure of how well the model predicts the actual data points. The higher coefficient of determination indicates that the model better fits the data. The R^{2} coefficient for both of the models is shown in Table 5.

Table 5.

R^{2} coefficient for ANN and ANFIS Models

Model

R^{2 }Value

ANN

0.942

ANFIS

0.923

Based on what explained before, the higher values of coefficient of determination would indicate the better capability of the model in predicting the specific studied characteristics. According to this table, the R^{2 }value for the ANN model is equal to 0.942 which shows the capability of this model in estimating the compressive strength of concrete. In addition, the R^{2 }value is determined as 0.923 for the ANIFIS model which demonstrates the fact that this model is a skillful model in predicting the compressive strength of concrete. To overall, the R^{2 }value for both the ANN and ANFIS models are high, and therefore, both of the models are skillful ones in predicting the compressive strength of concrete. In addition, comparison of these two models, shows that the R^{2 }value of ANN model is higher than the one for the ANFIS model. As a result, the ANN model is a more capable model than ANFIS in predicting the compressive strength of concrete.

6. Conclusion

This study is a comprehensive study on performance of soft computing models in predicting the concrete compressive strength. In this study, two soft computing models of ANN and ANFIS were used to estimate this characteristic of concrete. Result show that ANN and ANFIS models are both successful models in determining the compressive strength of concrete. Also, the comparison of ANN and ANFIS models showed that the ANN model is more accurate than ANFIS in predicting the compressive strength of concrete.

References

[1] Krishna, M. S. V., Begum, K. M. S., &Anantharaman, N. (2017). Hydrodynamic studies in fluidized bed with internals and modeling using ANN and ANFIS. Powder Technology, 307, 37-45.

[2] Hamdia, K. M., Lahmer, T., Nguyen-Thoi, T., &Rabczuk, T. (2015). Predicting the fracture toughness of PNCs: A stochastic approach based on ANN and ANFIS. Computational Materials Science, 102, 304-313.

[3] Naderpour, H., Kheyroddin, A., &Ghodrati Amiri, G. (2010). Prediction of FRP-confined compressive strength of concrete using artificial neural networks. Composite Structures, 92 (12), 2817-2829.

[4] Khademi, F., Akbari, M., &Nikoo, M. (2017). Displacement Determination of Concrete Reinforcement Building using Data-Driven models. International Journal of Sustainable Built Environment.

[5] Gupta, A. K., Kumar, P., Sahoo, R. K., Sahu, A. K., & Sarangi, S. K. (2017). Performance measurement of plate fin heat exchanger by exploration: ANN, ANFIS, GA, and SA. Journal of Computational Design and Engineering, 4(1), 60-68.

[6] Talebizadeh, M., &Moridnejad, A. (2011). Uncertainty analysis for the forecast of lake level fluctuations using ensembles of ANN and ANFIS models. Expert Systems with Applications, 38(4), 4126-4135.

[7] Khademi, F., Akbari, M., Jamal, S. M., &Nikoo, M. (2017). Multiple linear regression, artificial neural network, and fuzzy logic prediction of 28 days compressive strength of concrete. Frontiers of Structural and Civil Engineering, 11(1), 90-99.

[8] You, H., Ma, Z., Tang, Y., Wang, Y., Yan, J., Ni, M., ... & Huang, Q. (2017). Comparison of ANN (MLP), ANFIS, SVM, and RF models for the online classification of heating value of burning municipal solid waste in circulating fluidized bed incinerators. Waste Management.

[9] Güneyisi, E. M., D'Aniello, M., Landolfo, R., & Mermerdaş, K. (2014). Prediction of the flexural overstrength factor for steel beams using artificial neural network. Steel and Composite Structures, 17(3), 215-236.

[10] Nasrollahi, Amir. Optimum shape of large-span trusses according to AISC-LRFD using Ranked Particles Optimization. Journal of Constructional Steel Research, 134, 92-101.

[11] Keshavarz, Z. (2017). Predicting the Civil Engineering Characteristics through Soft Computing Models. Civil Engineering Research Journal, 1 (3), 555563.

[12] Khademi, F., Akbari, M., & Jamal, S. M. (2016). Predictia Rezistentei La Compresiune A Betonului Prin Testare Upv (Ultrasonic Pulse Velocity) Si Modelare Cu Retele Neuronale Artificiale/Prediction of Concrete Compressive Strength Using Ultrasonic Pulse Velocity Test and Artificial Neural Network Modeling. Revista Romana de Materiale, 46(3), 343.

[13] Yaman, M. A., Elaty, M. A., &Taman, M. (2017). Predicting the ingredients of self compacting concrete using artificial neural network. Alexandria Engineering Journal.

[14] Khademi, F., Jamal, S. M., Deshpande, N., &Londhe, S. (2016). Predicting strength of recycled aggregate concrete using artificial neural network, adaptive neuro-fuzzy inference system and multiple linear regression. International Journal of Sustainable Built Environment, 5(2), 355-369.

[15] Sitton, J. D., Zeinali, Y., & Story, B. A. (2017). Rapid soil classification using artificial neural networks for use in constructing compressed earth blocks. Construction and Building Materials, 138, 214-221.

[16] Khademi, F., Akbari, M., & Jamal, S. M. (2015). Prediction of compressive strength of concrete by data-driven models. i-manager's Journal on Civil Engineering, 5(2), 16.

[17] Karthika, B. S., &Deka, P. C. (2015). Prediction of air temperature by hybridized model (Wavelet-ANFIS) using wavelet decomposed data. Aquatic Procedia, 4, 1155-1161.

[18] Akib, S., Mohammadhassani, M., &Jahangirzadeh, A. (2014). Application of ANFIS and LR in prediction of scour depth in bridges. Computers & Fluids, 91, 77-86.

[19] Deka, P. C., &Diwate, S. N. (2011). Modeling Compressive Strength of Ready Mix Concrete Using Soft Computing Techniques. International Journal of Earth Sciences and Engineering, 4(6), 793-796.

[20] Pathak, S. S., Sharma, S., Sood, H., &Khitoliya, R. K. (2012). Prediction of Compressive Strength of Self compacting Concrete with Flyash and Rice Husk Ash using Adaptive Neuro-fuzzy Inference System. Editorial Preface, 3(10).

[21] Vishnuvaradhan, S., Chandrasekhar, N., Vasudevan, M., & Jayakumar, T. (2013). Intelligent modeling using adaptive neuro fuzzy inference system (ANFIS) for predicting weld bead shape parameters during A-TIG welding of reduced activation ferritic-martensitic (RAFM) steel. Transactions of theIndianInstitute of Metals, 66(1), 57-63.