A Review on Machine Learning Models in Injection Molding Machines

Table 7

ANN models in injection molding.

Algorithm used	AI model taxonomy	Depth layer sizes, training time, and testing time	Dataset	Framework, core language, and interface	Advantages	Disadvantages	Ref.
Adaptive neural networks, feed-forward backpropagation	Supervised learning	NA	Mold position sensor, nozzle pressure sensor, and screw position sensor data are used	BCI framework	Robustness and adaptivity for general purpose application for an injection molding machine	An accurate mathematical model has not been obtained	[36]
ANN backpropagation algorithms	Unsupervised	The network consisted of two hidden layers: two neurons from the 2nd hidden layer and two neurons from the 1st hidden layer	Training dataset consists of position about a motor of the velocity of injections profile and valve pin position profile	MATLAB Simulink	Molding precision is high because of smooth injection, better molding repeatability, and lesser power energy consumption	The training dataset is not always distributed upon the entire length for available input value to confirm the best approximate level about networks	[2]
Artificial neural network	Supervised learning	ANN with 3 hidden layers of neuron	The output data for ANN refer to the second-order fiber orientation tensor. The input data refer to a normalized level comparative to the starting charging sizes.	Keras library, Python 3.6	Reliably predicts the fiber orientation of the long fiber compression molded composite	NA	[78]
Artificial neural network (ANN)	Supervised learning	It consists of three layers	The dataset consists of 3600 simulation and 476 experiments from 36 various molds	3D files using Python and STL extension	With high accuracy, it recommends injection molding process condition on the real and correct time in fix material condition	The quantity of data is not sufficient. The system does not respond very quickly to new products.	[43]
A multilayer perceptron (MLP) neural network	Supervised learning	It consists of three layers	Dataset consists of temperature, pressure, and position	MLP framework	Outstanding performances in terms of the predicted accurate cost deduction	NA	[45]