Artificial neural network predictive modelling of laser microgrooving for commercial pure titanium (CP Ti) Grade 2 / Sivaraos ...[et al.]

• Main Authors: x, Sivaraos (Author), Zuhair, A.K (Author), Salleh, M.S (Author), Ali, M.A.M (Author), x, Kadirgama (Author)
• Format: Book
• Published: Universiti Teknologi MARA, 2021-04.
• Subjects: Article; PeerReviewed
• Online Access: Link Metadata
• Notes: https://ir.uitm.edu.my/id/eprint/47728/1/47728.pdf
• Keywords: Computer software; Neural networks (Computer science)

Grooving is the process of making a narrow channel on a surface of flat or cylindrical workpiece. Groove is precisely made to parts used in automotive, biomedical, and electronics industries. In automotive industries, groove plays an important role especially on mechanical parts to precisely locate seal (oring) to prevent gas/oil leakage between dynamic mating parts. On the other hand, artificial neural network (ANN) has been widely used in developing predictive models of various manufacturing processes to save huge amount of production time and money for industries. Unfortunately, very limited research has been investigated on micro groove quality employing ANN predictive models. Therefore, this research work presents on how the Artificial Neural Network (ANN) predictive model has been established, optimised and utilised to predict the laser micro-grooving quality of commercial pure titanium grade 2 material. A 3KW CO2 laser cutting machine was employed considering laser power, gas pressure, cutting speed, depth of cut and focal distance as the design parameters for modelling. On the other hand, three significant responses namely groove depth, groove width and groove corner radius were investigated. Experimental results were fed to establish the ANN predictive model, which then its parameters were optimized to gain high level prediction accuracy. The predicted results of ANN model presented the mean absolute percentage error for groove depth, groove width and groove corner radius at about 7.29%, 10.93% and 11.96% respectively. The obtained predictive results were found quite promising with the average of mean absolute percentage error (MAPE) for quality predictions which falls between 10 to 15%, concluding the validity of the developed ANN predictive model.