Response surface, neural network, and experimental approach to optimize process parameters and characterization of Al-Mg alloy by friction stir welding

Abstract

In the present study, an interaction relationship has been developed by following a design matrix consisting of few combinations of tool rotational speed, traverse speed, and tool pin configurations to understand the evolution of structural and mechanical properties of friction stir welded (FSW) Al-Mg alloy. The welded Al-Mg alloy was characterized in terms of microstructure to analysis different zones of weld using optical and scanning electron microscope. The mechanical properties such as bulk & micro hardness, and tensile strength were evaluated using Brinell, Vickers hardness tester, and tensometer respectively. The weld textures and grain size of the three different zones obtained using electron beam scattered diffraction (EBSD). Further, confirmation test was conducted to ensure the reliability. Results reveal that process parameters and tool pin profile influence the properties of the friction stir welded joints. The straight square pin configuration was the optimal structure for friction stir welding of Al-Mg alloy at a rotational speed of 1000 rpm and a welding speed of 40 mm/min, produced defect free weld with highest tensile strength of 136.25 MPa, bulk-hardness of 80.41 BHN, micro-hardness of 82 VHN in comparison with rest of the combinations considered in the investigation under same conditions. A fine recrystallized equiaxed grain with a partial fiber texture was evolved in the nugget zone. The measured experimental values agreed with the predicted data well.

Glass Fiber-Epoxy Composites with Carbon Nanotube Fillers for Enhancing Properties in Structure Modeling and Analysis Using Artificial Intelligence Technique

“…The hidden layer has a tangent sigmoid transfer mode activated. The output layer employs a linear transfer function and the “Levenberg–Marquardt” learning algorithm Table displays the correlation coefficient for ANN models.…”

“…The output layer employs a linear transfer function and the “Levenberg–Marquardt” learning algorithm. 30 Table 7 displays the correlation coefficient for ANN models. Comparisons were made between the experimental data and the ANN-predicted wear data ( Figure 16 a).…”

“…However, choosing the correct neural network is crucial for the accuracy of the ANN model . An ANN model consists of three layers: the input layer, hidden layer, and output layer .…”

“…28,29 However, choosing the correct neural network is crucial for the accuracy of the ANN model. 30 An ANN model consists of three layers: the input layer, hidden layer, and output layer. 31 The hidden layer acts as a bridge between the input layer and the output layer, and its bias and weights are adjusted until the difference between the experimental and predicted values is minimized, resulting in a minimal error.…”

A Review of Optimization and Measurement Techniques of the Friction Stir Welding (FSW) Process

“…In [187], CCD based on the RSM technique was applied to optimize the FSW parameters (RS: 800-1200 rpm, TS: 20-60 mm/min, pin profiles: straight cylindrical, straight square, tapered cylindrical) on the performance (surface roughness, and tensile strength) of Al-Mg alloy joints. A higher surface roughness (10.705 µm) was observed at RS and TS equal to 800 rpm and 20 mm/min, and a lower surface roughness (4.9 µm) was recorded at the straight cylindrical pin profile, with RS and TS equal to 1200 rpm and 20 mm/min.…”