**8. Process parameters modelling using ANN**

The input layer in this work consists of four neurons that correspond to each of the three control factors and one neuron in the output layer that corresponds

**145**

**Figure 11.**

*Predicted mechanical properties of triangular pin profile using ANN.*

*Experimental Investigations on AA 6061 Alloy Welded Joints by Friction Stir Welding*

to each response. To find the finest network architecture, differently hidden layer networks were created and tested with distinct numbers of hidden layers and neurons; various algorithms is used for practice; the hidden layer and output layer transfer functions are altered and the generalization ability of the various networks is noted. Finally, the ideal network is chosen to forecast the strengths. Twenty-five concealed neurons are trained for the ideal architecture of the network. The ANN network is run on MATLAB software. The data set with 27 models is randomly split into two classifications: The training dataset consists of 75% of information and

*DOI: http://dx.doi.org/10.5772/intechopen.89797*

**Figure 10.** *Predicted mechanical properties of conical pin profile using ANN.*

*Experimental Investigations on AA 6061 Alloy Welded Joints by Friction Stir Welding DOI: http://dx.doi.org/10.5772/intechopen.89797*

to each response. To find the finest network architecture, differently hidden layer networks were created and tested with distinct numbers of hidden layers and neurons; various algorithms is used for practice; the hidden layer and output layer transfer functions are altered and the generalization ability of the various networks is noted. Finally, the ideal network is chosen to forecast the strengths. Twenty-five concealed neurons are trained for the ideal architecture of the network. The ANN network is run on MATLAB software. The data set with 27 models is randomly split into two classifications: The training dataset consists of 75% of information and

**Figure 11.** *Predicted mechanical properties of triangular pin profile using ANN.*

*Aluminium Alloys and Composites*

optimized welding conditions.

**8. Process parameters modelling using ANN**

region that is metal consolidated and does not dilute the base material. The results of this research showed that in the event of AA6061 aluminum alloy, the joints made with an axial force of 2 KN had better tensile strengths. **Table 4** shows the

The input layer in this work consists of four neurons that correspond to each of the three control factors and one neuron in the output layer that corresponds

**144**

**Figure 10.**

*Predicted mechanical properties of conical pin profile using ANN.*

25% of information in test information set. 20 training models for ANN strength modelling are considered. The weights are frozen after the training and the model is checked for experimental findings.

**Figures 10(a)**, **11(a)** and **12(a)** show the experimental and ANN computed tensile strength values for AA6061 materials with a conical pin tool, triangular pin tool and threaded pin tool, and it is clear that the values predicted by ANN are very close to the experimental values. **Figures 10–12** show the ANN prediction values and experimental values for the different mechanical properties. The experimental investigation it is proved that the triangular pin profiles yield better results.

**147**

**Figure 13.**

*SEM images of different tool pin profiles of FSW joints of AA 6061.*

*Experimental Investigations on AA 6061 Alloy Welded Joints by Friction Stir Welding*

SEM assessments are completed utilizing a Scanning Electronic Microscope for 6061 aluminum alloy to dissect the weld testimony on weld chunk surface. The SEM photos of the joints at different procedure parameters are appeared

*DOI: http://dx.doi.org/10.5772/intechopen.89797*

**9. SEM analysis**

**Figure 12.** *Predicted mechanical properties of threaded pin profile using ANN.*

*Experimental Investigations on AA 6061 Alloy Welded Joints by Friction Stir Welding DOI: http://dx.doi.org/10.5772/intechopen.89797*
