This study examines the influence of dataset choice on neural network training, comparing the dataset using the S–N curves most commonly obtained in the literature (R = 10, −1, and 0.1) with two others that exchange alternate S–N curve (R = −1) data for another S–N curve, which can be chosen either within the variable load region with predominance of tension or the variable load region with predominance of compression. In addition, the present investigation defines a stopping criterion for modular network training. A total of 10 datasets drawn from the literature were analyzed to achieve the desired results.
MandellJFSamborskyDD. DOE/MSU composite material fatigue database: test methods, materials and analysis, Albuquerque: SAND97-3002, Sandia National Laboratories, 1997; 140–140.
2.
SutherlandHJ. On the fatigue analysis of wind turbines, Albuquerque: SAND99-0089, Sandia National Laboratories, 1999; 133–133.
3.
VassilopoulosAPManshadiBDKellerT. Influence of the constant life diagram formulation on the fatigue life prediction of composite materials. Int J Fatigue2009; 32(5): 659–669.
4.
FreireRCSJrAquinoEMFDóriaNetoAD. Comparative study between ANN models and conventional equations in the analysis of fatigue of GRP. Int J Fatigue2009; 31: 831–839.
5.
LeeJAAlmondDPHarrisB. The use of neural network for the prediction of fatigue lives of composite materials. Compos Part A: Appl Sci Manufac1999; 30(10): 1159–1169.
6.
Al-AssafYEl KadiH. Fatigue life prediction of unidirectional glass fiber/epoxy composite laminate using neural network. Compos Struct2001; 53(1): 65–71.
7.
El KadiHAl-AssafY. Prediction of the fatigue life prediction of unidirectional glass fiber/epoxy composite laminate using different neural network paradigms. Compos Struct2002a; 55: 239–246.
8.
El KadiHAl-AssafY. Energy based fatigue life prediction of fiberglass/epoxy composites using modular neural network. Compos Struct2002b; 57: 85–89.
9.
FreireRCSJrAquinoEMFDóriaNetoAD. Building life diagrams of fatigue using artificial neural networks. Int J Fatigue2005; 27: 746–751.
10.
Freire RCS Jr, Aquino EMF and DóriaNeto AD. Applying ANN modeling in fatigue failure of GFRP. In: 19th International Congress of Mechanical Engineering, Vol. 11, 2007, Brasilia, DF: Brasil, pp. 1–10.
11.
DOE/MSU. Composite material fatigue database, Bozeman: Montana State University-Bozeman, 2003; 272–272.
12.
Harris B, Gathercole N, Lee JA, et al. Life-prediction for constant-stress fatigue in carbon-fibre composites. Phil Trans R Soc Lond A 1997; 355(1727):1259–1294.
13.
GathercoleNReiterHAdamT. Life prediction for fatigue of T800/5245 carbon-fibre composites: I constant-amplitude loading. Fatigue1994; 6: 523–532.
14.
KawaiMKoizumiM. Nonlinear constant fatigue life diagrams for carbon/epoxy laminates at room temperature. Compos: Part A2007; 38(11): 2342–2353.
15.
BeheshtyMHHarrisBAdamT. An empirical fatigue-life model for high-performance fibre composites with and without impact damage. Compos – Part A: Appl Sci Manufact1999; 30: 971–987.
Harris B. A parametric constant life model for prediction of the fatigue lives of fibre reinforced plastics. In: Fatigue in Composites, Ed. Woodhead Publishing, CRC Press, Chapter 3, 2003; 546–568.
18.
Freire RCS Jr, Sena Bremmer BV de. Establishing a stop criterion inartificial neural network training for composite materials. COBEM, Gramado, RS: Brasil, 2009.
