presented to the network for learning. This forms one-iteration.At the end of iteration, test patterns are presented to ANN andthe prediction performance of ANN is evaluated. Further training of ANN is continued till the desired prediction performance is reached.The concept of steepest-descent method is used in BPA toreach a global minimum. The number of layers are decidedinitially. The number of nodes in the hidden layers are decided.It uses all the 3 layers (input, hidden and output). Flow-chartfor BPA is shown in Figure 1.
B. Steps Involved In Training Bpa
Forward PropagationThe hidden layer connections of the network are initializedwith weights.The inputs and outputs of a pattern are presented to thenetwork.The output of each node in the successive layers iscalculated by using equation (1).O
(output of a node)
(1)For each pattern, error is calculated using equation (2).E(p) = (1/2)
(2)Reverse PropagationFor the nodes, the error in the output layer is calculated usingequation (3).
=o(1-o)(d-o) (3)The weights between output layer and hidden layer areupdated by using equation (4).W
The error for the nodes in the hidden layer is calculated byusing equation (5).
(updated weights between hidden & outputlayer)
The weights between hidden and input layer are updated byusing equation (6).W
The above steps complete one weight updation.The above steps are followed for the second pattern for subsequent weight updation. When all the training patterns are presented, a cycle of iteration or epoch is completed. Theerrors of all the training patterns are calculated and displayedon the monitor as the MSE.E
(7)IV. RESULTS AND DISCUSSIONS
Static analysis calculates the effects of steady loadingconditions on a structure. Static analysis, however, includessteady inertia loads (such as gravity and rotational velocity),and time-varying loads that can be approximated as staticequivalent loads. Static analysis is used to determine thedisplacements, stresses, strains, and forces in structures or components caused by loads that do not induce significantinertia and damping effects. Steady loading and response
conditions are assumed; that is, the loads and the structure’s
response are assumed to vary slowly with respect to time.
Length X 0.2 mLength Y 0.5 mLength Z 0.5 m
Volume 5.2943e-003 m³Mass 41.561 kgCentroid X -8.688e-002 mCentroid Y 4.5371e-005 mCentroid Z -3.5088e-005 mMoment of Inertia Ip1 1.4413 kg·m²Moment of Inertia Ip2 0.80226 kg·m²Moment of Inertia Ip3 0.80168 kg·m²
Nodes 22941Elements 12652
Structural Steel > Constants
Density 7850 kg m^-3Coefficient of Thermal Expansion 1.2e-005 C^-1Specific Heat 434 J kg^-1 C^-1Thermal Conductivity 60.5 W m^-1 C^-1Resistivity 1.7e-007 ohm m
Figure 2 shows the amount of von
Mises stress in (Pa) presented at various nodes mentioned in the x-axis. Figure 3 presents strain on the alloy wheel.The node number is given as one of the inputs at the inputlayer of the ANN topology. Based on the requirements of theoutputs, number of input parameters can be increased. Table 2 presents the numbers of nodes used in the input layer, hiddenlayer and output layer during training.ANSYS 14 software is used for analyzing the alloy wheel.The numerical values of stress are used to train the artificialneural network (ANN) topology by using Back propagationalgorithm (BPA). The training of patterns used for the ANNalgorithms are chosen from the strain data generated usingANSYS program. During the training process, node numbersare presented in the input layer of the ANN andcorrespondingly, strain values are presented in the output layer of the ANN. Depending upon the type of values present in the patterns, the learning capability of the ANN algorithms varies.
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 11, No. 6, June 201370http://sites.google.com/site/ijcsis/ISSN 1947-5500