Optimization of cutting parameters in the drilling process of aluminum 5083-H111 material with ann

Aluminum alloys have become frequently preferred alongside many new composite and alloy materials due to their lightness, durability and physical properties. These alloys can be used in the automotive, aircraft industry, maritime and construction sectors. With the use of aluminum alloys in these sectors, the need for drilling has increased considerably in terms of the use of fasteners encountered in the manufacturing phase of these sectors. Aluminum 5083 H111 alloy is mainly preferred in the maritime and aerospace industries because it is light and highly corrosion-resistant. In the manufacturing processes in these sectors, it is inevitable to apply thousands of hole drilling processes. In addition, in these areas where care is taken to produce with high precision and quality, the quality of fasteners and holes becomes more important than anything else. Keeping the quality and efficiency in these processes at the maximum level can be achieved by optimizing the parameters in the drilling process. A safe working environment and sustainability are of great importance in engineering studies. For the product or system to be manufactured to be reasonable in terms of length, quality and cost, all forces acting on cutting tools and workbenches should be examined in detail and measured well and sensitively. When the results obtained as a result of theoretical calculations and the data obtained as a result of the application are compared, it is seen that most of the time, they do not overlap. For this reason, experimental analysis and measurement are of vital importance. Among the machining methods, drilling processes include 33-40% of the methods. Hole quality in drilling processes varies depending on parameters such as circularity, axial misalignment, surface roughness, hole size and burr formation. The working conditions and the quality of the assembly processes depend on the hole quality. The higher the quality of the hole axis, the smoother the precision assembly process can be. Misalignment affects the stress distribution and causes the material to be subjected to excessive tensile forces. As a result of the roughness in the holes, abrasive deterioration occurs in the material. In designs with more than one sample, mandatory tolerances occur for the surface of the hole, its axes and the positions of these axes. In particular, tolerances such as hole axis, perpendicularity of the axis to the surface, eccentricity, roughness of the hole axis, and circularity are critical in sensitive industries such as aerospace. In this study, AA5083 H111 alloy was studied and 3 parameters were determined in the experiments. Each parameter was divided into 3 different levels. Experiments were carried out with these combinations and 27 holes were drilled. In the experiments, 3 different cutting speeds (80,100,120 m/min), 3 different feed rates per tooth (0.06, 0.09, 0.12 mm/tooth), 3 different cooling types (dry, air cooled and liquid cooled) were studied with 3 parameters. TiAlN coated HSSE-Co5 drill was used. Experiments were made using Taksan TMC-700 V brand and model CNC vertical machining center, ESIT AX3 load cell, NI cDAQ-9188 data acquisition unit and FlexLogger software. The resistance force exerted by the cutting tool in the Z axis of each drilled hole was considered. Artificial intelligence predictions were carried out through the NNTools module of the MATLAB package program. The estimation data obtained from the ANN were included in the text through the program's graphical outputs, the parameters' effects on the process were examined, and the estimations and the actual test results were interpreted with the help of graphics. In this study, HSS (high speed tool steel) was studied with a drill. The reason for this is that it is more economical than carbide drills. The working set consists of TiAlN coated HSSE-Co5 drills. All experiments were done with a single drill. Force-Displacement graph was drawn for the experiment performed in each cooling type. Exponential smoothing method was used since it was seen that there were signal noises in the data. In addition, this study will also use artificial intelligence, apart from hole drilling experiments. The reliability of the experiments will be questioned by using the artificial intelligence-supported prediction program called ANN (Artificial Neural Networks). At the same time, it will be possible to see how efficiently the artificial intelligence works by comparing the prediction results with the experimental results. In this way, in future studies, analyzes can be made with specific parameters with the help of artificial intelligence, and it will be possible to predict how much cutting force can be encountered in a specific error percentage range. Therefore, it will be advantageous regarding material, time and manufacturing costs. The artificial neural networks (ANN) method is an approximate method that contains input and output data sets and defines the complex relationships between these sets. The ANN method is frequently used in the industry to create a model with the experimental parameters and results and to model the estimation structure to eliminate the necessity of repeating high-cost experiments with the data obtained from the experiments. Identify complex process properties of problems using empirical equations. ANN has a structure consisting of an input layer, an output layer and one or more hidden layers. The input and output layers are sets of neurons that define these layers. Although there is no specific amount for hidden layers; generally, 1 or 2 are used. The output layer receives all the responses from the hidden layer and creates an output vector. While ANN models with a single output layer give better results, the efficiency of the results is affected since models with multiple output layers are more complex. The analysis will be performed with the help of the MATLAB software Neural Net Fitting application. With the help of artificial intelligence, the cutting forces will be estimated and compared with the experimental results. In the study using MLP (Multi-Layer perception) method, 3 parameters were entered as input and cutting force was requested as output. While the results of the experiments performed during drilling with all three cooling types were analyzed, it was understood that machine learning and analysis were more efficient by using the Levenberg Marquardt backpropagation method, which is one of the methods available in the application, So it was studied with this method. 15% of the data is reserved for testing, 15% for validation, and 70% for use in the training function. Working with 3 inputs, 1 output and 3 hidden layers, 6 neurons are assigned to the first hidden layer, 8 neurons to the second hidden layer, and 10 neurons to the third hidden layer. Learning was tried 3 times with layers with 2-4-6, 4-6-8, 8-6-10, 6-8-10, 3-5-7, 8-10, 6-8, 6-10, 1-3-7 neuron arrays, respectively. However, no error rate below 10% could be obtained, and the least error rate and machine learning were obtained from the 6-8- 10 arrayed alternative. These sequence selections are based on guesswork and trial and error based on previous studies. For each of these layers, logsig and tansig functions were tried with different combinations, and it was seen that the tansig function was more efficient for all layers. Artificial intelligence predictions approached actual experiments with an average error rate of 3.63% in 3 different experiments. As a result of the experiments and estimations, it has been seen that the optimum parameters in the cutting tool and material conditions are 100 m/min cutting speed, 0.06 mm/tooth, feed per tooth and liquid cooling.