Feature: AI
Labview program, aſter which training was performed in Matlab on an artificial neural network (ANN). Tis enabled us to prepare vibration graphs to show tool wear, surface roughness and cutting forces according to different input values. Te flow chart of the process is shown in Figure 2. Modelling of surface roughness mechanism is a complex
process that relies on chip removal. For this reason, determining the roughness value analytically can be difficult. Equation 1, introduced by Boothroyd and Knight in 1989, determines the average surface roughness (Ra
):
(1) Wear is defined as material loss on the contact surfaces between
the cutting tool and the workpiece. In metal cutting processes, tool wear is defined as possible losses of processed products or used tools.
Figure 2: Process flow chart
The ANN architecture ANNs are information-processing systems developed from mathematical models of biological neural networks, which display a learned performance. ANNs are trained by pairing output values with corresponding input values. Training an ANN to obtain target outputs requires a great
during cutting under high cooling-liquid pressure is shown in Figure 1. Inconel 718 super alloy material was subjected to the turning process under various water pressure values. Cutting forces (progressive and tangential) were measured with a dynamometer, whilst displacement was measured with a vibration meter. Surface roughness and tool wear values were determined based on experiment parameters. Using experimental data, the cutting parameters (cutting speed, feed rate, cutting depth) and material properties – which together constitute the vibration-related factors – were collected in a database. Data was first taken into the
number of inputs and steps, as well as numerous output sets for the relevant inputs. Tese data sets are called “training” and “test” sets. Aſter the learning process, the test process is performed using the test inputs, to check the results of the designed network. Te mechanism that enables the arrangement of weights in the network (to generate the desired inputs from the ANN during training) is called the “learning algorithm”. Tere are many ANN architectures described in literature.
In our project we used the multi-layered Feed-Forward Backpropagation algorithm, which is suitable for engineering applications, per Neşeli et al. 2009. Figure 3 shows how surface roughness and tool wear
predictions are performed in an established system, by using an
Figure 4: The cutter used in the study Figure 3: ANN network structure
www.electronicsworld.co.uk September/October 2020 33
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68
