Constitutive neural network training data for DeltaFix tool

The data files Data1 to DATA2 reflect the feature extraction and feature selection steps of the explanatory and response variables of a constitutive neural network model. Data 1 is preliminary extracted from finite element analysis in ANSUS software. The FEA was conducted for a set of 2-D workpieces with boundary conditions of Clamping and machining forces. The aim of this test is to determine the deformation level of the workpiece under different configurations of fixture layout and machining forces. A sequential data preprocessing was applied to transfer Data 1 to Data 6 which will be used to train the constitutive neural network model. The steps include sensitivity analysis of the discretization step, outliers’ elimination, Kendall correlation test, stepwise regression significance test, Confidence Based Repetition of Observations (CBRO), standardization. The CNN model is part of DeltaFix tool development. DeltaFix is developed to work on NX 10.0 CAD environment, based on C++ and NXOpen libraries. The tool aim to solve the problem of fixture synthesis where an optimization is carried out to obtain the robust fixture layout for a workpiece with known clamping and machining forces. The CNN model is responsible for part of the evaluation process in the tool during fixture layout optimization task (The CNN predicts the deformation level of a workpiece based at a specific fixture layout).