A Review on Machine Learning Models in Injection Molding Machines

Table 10

ANN genetic algorithms 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.
ANN genetic algorithm	Supervised learning	Model structure with 1 hidden layer, where 7 neurons in the layer were chosen	Dataset by a combination of native (MT) and substitutionary (MS, i) data from other data domain	All implementations were done in Python. TensorFlow, Keras, and scikit-learn	Reduces the requirement of many training data	Experiments have not been designed with geometrically different parts	[29]
Artificial neural network genetic algorithm	Supervised algorithm	A 3-layered backpropagation network (BPN) with a supervised learning algorithm was employed for producing required output-input layers	Dataset consisted of parameters of mold design, part design, and process	Hybrid NN. GA systems are implemented on Visual Basic programming languages. Validation tests were performed in desktop computers for PII-three hundred processors and one hundred and twenty-eight Mbytes about memory as utilizing commercial mold flow simulation packages C MOLD	Time needed for the determination of starting process parameter for injection molding could be significantly reduced	The effectiveness of that system is not verified and validated	[108]
Genetic algorithm	Supervised learning	Three-layer network	Two datasets: one of them was from Petronas Penapisan (Melaka) and the second dataset was XOR common problem dataset	MATLAB	The performance of the proposed ANN-GA approach was compared with existing 5 approaches in terms of the prediction effectiveness	Optimum ANN parameters into a broader search space are not determined	[109]
Genetic algorithms	Supervised learning	NA	Dataset consists of parameters like cooling time, back pressure, screw rotation speed, injection pressure, injection rate, holding pressure, nozzle temperature, holding time, and open stroke	ENGEL injection molding machine (USA) was equipped with EC88/CC90 and A02 controllers	Different sized simplexes were the apt methods for online optimization of the situation	The percentage of reject was also susceptible to half pressure level and cool time	[110]
Genetic algorithm	Supervised learning	A three-layer network, the hidden layer contains six nodes.	The dataset was sixty from the GDC plant in Universiti Teknologi PETRONAS, and five hundred datasets from the Sdn Bhd in PETRONAS Penapisan and sixty-four experiment datasets from flank wear for the drilling process were used in developing the newly proposed architecture	MATLAB function	Process of designing ANN is less dependent on humans and is more sophisticated	The training data used are not sufficient	[111]
Genetic algorithm	Supervised learning	Two-layer network where a hidden layer contains the radial basis neuron and the output layer contains the linear neuron	A dataset used in this study consists of 102 sets.	Moldflow software	The proposed approach can effectively be helpful for the engineer’s determination of optimal gate, locations, and other designing elements for achieving competition profits about product cost and quality	During detection in the optimum location of the gate, many parts have not been considered	[44]
Genetic algorithm	Supervised learning	Three-layer network	Experimental data of Taguchi's parameter design method are utilized for effectively training and testing the BPNN model	Integrated numerical simulation software	Effectively helps engineers determine the final optimal process parameter setting	Multiresponse process parameter design problems	[112]