384
A.-I. Berariu et al.
productivity, the process itself needs to be modeled with regards to the material properties, cutting mechanics, tool geometry, process kinematics and structural dynamics
in order to predict the forces, torque, power, form errors and vibrations during the
metal cutting operations.
One of the main issues that process engineers will encounter while trying to
optimize the MRR (Material Removal Rate) is the apparition of chatter. The current
understating of this phenomena leads to the conclusion that there are three types of
chatter: frictional chatter, mode coupling chatter and regenerative chatter. In their
research [1], the authors concluded that in order to control the milling stability it is
mandatory to know the cutting force coefficients, the model parameters in term of
cutting system, the dynamics characteristics of the workpiece, the damping effect of
the process, the tool runout and the gyroscopic effect. One of the most recent works
in the field [2] treats the frictional chatter separately. The team highlighted a sweet
spot where all vibrations are reduced suggesting a transient phase dependent on the
cutting velocities.
The authors of [3] managed to create a unified cutting force model that proved
to be precise where the geometric, kinematic and mechanic parameters are used to
predict the forces for turning, milling, boring and drilling operations.
The usage of Artificial Neural Network (ANN) is now largely available due to
the increasing efficiency in computation and has reached a point where it is feasible
to be applied also in the manufacturing field. In [4], the research team managed
to train an ANN to predict the cutting forces in all three directions F x , F y and F z
for turning operations. The database contained 96 samples and using this setup, the
team managed to validate the ANN model and the results showed a good correlation
between predicted forces and experimental forces.
In order to maximize the precision of the ANN model or to extend the dimensions
(add more variables like new materials, cutting parameters, etc.) the number of “training” samples needs to be increased exponentially. There are two ways to generate
samples, the first one is through experiments and the second is to use different mathematical models. The first one is usually more expensive than the second, but it offers
high precision. The second uses various analytical models or numerical models like
Finite Element Method (FEM) that are able to simulate the physics with sufficient
precision at a fraction of the cost.
The proposed method is centered on shoulder milling for proof of concept. The
shoulder milling operations include: shoulder/face milling, edging peripheral milling
and shoulder milling of thin deflecting walls. Also, the radial depth a c and axial
depth a p (Fig. 1) are considered constant for this method. The method is verified
using one cutting scenario where the Al6061_Machining material is used with the
aim of obtaining one sample that will become the input for a machine/deep learning
algorithm.
A.-I. Berariu et al.
productivity, the process itself needs to be modeled with regards to the material properties, cutting mechanics, tool geometry, process kinematics and structural dynamics
in order to predict the forces, torque, power, form errors and vibrations during the
metal cutting operations.
One of the main issues that process engineers will encounter while trying to
optimize the MRR (Material Removal Rate) is the apparition of chatter. The current
understating of this phenomena leads to the conclusion that there are three types of
chatter: frictional chatter, mode coupling chatter and regenerative chatter. In their
research [1], the authors concluded that in order to control the milling stability it is
mandatory to know the cutting force coefficients, the model parameters in term of
cutting system, the dynamics characteristics of the workpiece, the damping effect of
the process, the tool runout and the gyroscopic effect. One of the most recent works
in the field [2] treats the frictional chatter separately. The team highlighted a sweet
spot where all vibrations are reduced suggesting a transient phase dependent on the
cutting velocities.
The authors of [3] managed to create a unified cutting force model that proved
to be precise where the geometric, kinematic and mechanic parameters are used to
predict the forces for turning, milling, boring and drilling operations.
The usage of Artificial Neural Network (ANN) is now largely available due to
the increasing efficiency in computation and has reached a point where it is feasible
to be applied also in the manufacturing field. In [4], the research team managed
to train an ANN to predict the cutting forces in all three directions F x , F y and F z
for turning operations. The database contained 96 samples and using this setup, the
team managed to validate the ANN model and the results showed a good correlation
between predicted forces and experimental forces.
In order to maximize the precision of the ANN model or to extend the dimensions
(add more variables like new materials, cutting parameters, etc.) the number of “training” samples needs to be increased exponentially. There are two ways to generate
samples, the first one is through experiments and the second is to use different mathematical models. The first one is usually more expensive than the second, but it offers
high precision. The second uses various analytical models or numerical models like
Finite Element Method (FEM) that are able to simulate the physics with sufficient
precision at a fraction of the cost.
The proposed method is centered on shoulder milling for proof of concept. The
shoulder milling operations include: shoulder/face milling, edging peripheral milling
and shoulder milling of thin deflecting walls. Also, the radial depth a c and axial
depth a p (Fig. 1) are considered constant for this method. The method is verified
using one cutting scenario where the Al6061_Machining material is used with the
aim of obtaining one sample that will become the input for a machine/deep learning
algorithm.