An investigation of hybrid models FEA coupled with AHP-ELECTRE, RSM-GA, and ANN-GA into the process parameter optimization of high-quality deep-drawn cylindrical copper cups

Proceedings of the Institution of Mechanical Engineers, Part C:

2022

Self Cite

Predicting the results of multi-stage deep drawing in determining the thickness distribution and thinning of the workpiece are measures that will help reduce production costs by saving materials and production time. Machine Learning (ML) technique is a promising method for predicting the thickness distribution (TD) of drawn copper metal. Macro and micro products are currently used in sensitive electronic and structural applications. In this study, ML defines the relationship between the scaled thickness and the respective number of stages governed by the scaling law. It includes the development of Artificial Neural Network (ANN) tools based on machine learning to model the relationship between thickness by scaling law and stages based on multi-stage deep-drawn cups. TD is a measure of consistency. The TD is measured from the initial blank of 1500 mm long, 750 mm wide, and 6 mm thick Cu strip, reduced by 50% in nine successive steps until a final thickness of 0.1875 mm. Two ANN models used for TD prediction are Bayesian regularization (BR) and Levenberg-Marquardt (LM) algorithms. A trained machine learning model can successfully predict and verify the unseen data of 1.5 and 0.38 mm TD. ANN is used to predict the Finite Element Analysis (FEA) results and confirm them through the experimental results. The developed model can predict the TD of the multi-stage cup with the die design parameters. The difference between the TD predicted value and the measured value is based on the simulation results of multi-stage cups using the finite element method. When the predicted and measured TD, the difference in cup drawing depth is 0.5%–2.0%. The results show that the LM model is suitable for predicting the TD of formed copper cups following the scaling law.

Section: Confirmation Testsupporting

confidence: 78%

Experimental validation of machine learning models for prediction of the thickness distribution of directionally rolled copper strips under scaling law

Proceedings of the Institution of Mechanical Engineers, Part C:

2022

Self Cite

“…A carefully selected heat treatment at 150°C can relieve stress without recrystallization. Mechanical properties [17][18][19][20][21] and anisotropy [17][18][19][20][21] are obtained by uniaxial tensile test according to ASTM E8-the lower dimensional standard, and the forming limit diagram of this test passes the Nakajima test. 17 The schematic diagram of the employed rolling process is shown in Figure 1(a) and (b) depicts the route sequence.…”

Section: Methodsmentioning

confidence: 99%

Artificial Neural Network prediction of forming limit diagram for directionally-rolled, size scaled copper strips

Harshavardhana

Proceedings of the Institution of Mechanical Engineers, Part C:

2023

Estimating the forming limit diagram (FLD) is tedious and cost-intensive. Methods driven by data and artificial intelligence are used to determine the relationship between scaled thickness and the forming rates of various cups drawn out of ETP copper sheets. Machine learning (ML) techniques have a good chance of predicting the FLD of copper alloys, and they are being used increasingly in sensitive electronic and structural applications. The current research aims to create ML-based artificial neural network (ANN) tools to model the relationship between scaled thickness and forming rates as a function of formability. The FLD is measured for copper strips of initial dimensions of 1500 mm long, 750 mm wide, and 6 mm thick, whose thickness was reduced successively by 50% in nine incremental steps. Thus, 3, 1.5, 0.75, 0.38, and 0.19 mm sheets were obtained and used to determine FLD through the Nakajima approach. An FEA model of the drawing was made in Altair Inspire Form, and the simulation results were used to train a two-step ML. A Bayesian regularization (BR) and Levenberg-Marquardt algorithm (LM) were used in the first step to predict strains’ maximum and minimum points. In the second step, the minor strains predicted in the first step are used as inputs. Using the same feature set, the BR and LM algorithms predict the major strain, showing a linear trend until the middle and then a nonlinear trend. The trained ML model was used to predict unknown intermediate values for estimating the over-learning and over-fitting problems here for 2 and 0.25 mm thick sheets and are validated experimentally. The variation between the FLDs of predicted and experimentally verified data falls between 2% and 5%. Such small changes in the FLD values show that the proposed ML model could be used to predict the FLDs of copper strips.

“…Hill's anisotropy parameters for the specimen orientation and strip anisotropy. 28,29 The true plastic stress is the true plastic strain curve of the material according to Swift's hardening rule. 28,29 The Altair Inspire form with Quads has an element size of 50 mm with adaptive technique and Automatic Tool position, radios explicit solver and post-processing -Altair Analysis Explorer.…”

Section: Numerical Simulation-based Experimental Trailsmentioning

confidence: 99%

“…28,29 The true plastic stress is the true plastic strain curve of the material according to Swift's hardening rule. 28,29 The Altair Inspire form with Quads has an element size of 50 mm with adaptive technique and Automatic Tool position, radios explicit solver and post-processing -Altair Analysis Explorer. The punch was made to move in the vertically downward motion to engage the blank top surface at a speed of 1 mm/s, from an initial height of 5 mm from the blank's top surface and designed to move to the designated depth of extrusion.…”

Section: Numerical Simulation-based Experimental Trailsmentioning

confidence: 99%

Hybrid optimisation of input process parameters of deep-drawn cylindrical cups from directional rolled copper strips

Proceedings of the Institution of Mechanical Engineers, Part C:

2022

Deep drawing is a method of producing sheet metal. It is extensively used in the automobile, packaging and home appliance industries. Incorrect parameter selection might induce defects in stamped components. Trial and error can improve product accuracy, but at the expense of execution and calculation. An optimisation approach and a response surface model are used to determine the optimal parametric parameter design. This study proposes a full and efficient optimisation approach, starting with modelling by RSM (Response Surface Methodology) and concluding with optimal process parameter identification. Based on this, a neural network was developed to improve cylindrical cup deep drawing. Three characteristics of fabricating the cups are clearance, punch radius and coefficient of friction. This parameter demonstrates the resultant tool force, spring back, maximum forming limiting curve and maximum thinning rate of cylinder cups. Initially, the three inputs are varied to measure the outcome using simulation software by Altair inspire form. Then, the equations for each output are designed using quadratic-based linear model fitting. The generated objective function is then used to establish the ideal process parameters for the cylindrical cups. The proposed method is compared to real-time physical experiments. The optimised parameter microstructure at the high-stress zone and the product quality are satisfactory without defects.