Two applications namely BTHPeel and J2-Iso was developed in MATLAB environments so far. They are introduced below.
Extension to these apps will follow.
BTHPeel is a MATLAB application that can be used for optimization of cohesive parameters of thin laminate FE-model from peel tests. In the background it creates the beam peel model using the input parameters entered in this app interface as shown in the figure to the left. Required input data for the substrates to be peeled are peel arm’s Young’s modulus, Poisson’s ratio, density, yield stress, type of hardening (isotropic or kinematic), stress-plastic strain couple for defining bilinear hardening line, width and thickness. All the inputs are taken in the SI (mm) unit system. For interface model, cohesive stiffness and adhesive/interface thickness are the only inputs. The beam peel model is then simulated in ABAQUS in a given range of three cohesive parameters for optimization. The upper and lower bounds of these cohesive parameters, namely fracture energy of the interface, cohesive strength and the cohesive separation ratio of a trapezoidal cohesive law are given as input to the app. The app will further take input of the number of FE-simulations of beam peel to be executed in order to train an artificial neural network (ANN) and input of the number of genetic algorithm (GA) iterations for optimization. Based on these two numbers, the app will systematically execute a series of FE-simulation for a set of cohesive parameters within the lower and upper bounds and extract the results. This set of cohesive parameters and corresponding simulation outputs will be used to train the ANN. The ANN will be optimized in a GA against peel force and root rotation of 90 and 180 degree peel tests which are also inputs to the BTHPeel app. After several GA iterations as desired by the app user, an optimized set of cohesive parameters will be delivered. This application can be used with peel tests of any thin-flexible laminate with the given modelling assumptions.
J2-Iso is a MATLAB application that calibrates the plastic and damage material properties from a uniaxial tensile test. An isotropic elastic-plastic progressive damage constitutive model based on Hooke's law, J2 yield criteria, isotropic hardening, associated flow rule and ductile damage was used to describe the material in FE-simulation of the tensile test. J2-Iso takes the experimental tensile force-displacement response, specimen length, cross-section, Young’s modulus, initial yield stress, Poisson’s ratio and number of points to be calibrated, as input. The interface of the application is presented in the figure to the left. When executed, an input file for the simulation of a displacement-controlled tensile test is created and the hardening parameters are incrementally optimized as the specimen is loaded with increasing displacement values. At each displacement increment, J2-Iso optimizes the plastic strain and stress values until the desired level of accuracy is achieved between experimental and simulation force response at current displacement. The optimized plastic strain and stress values are appended to the input file and a new simulation executes for next displacement increment. This process continues until damage initiation is detected. The application detects the start of damage when the hardening perimeter value (stress) decreases at the next displacement increment. In the next step, it optimizes the damage variable by establishing an artificial neural network and updating that in an GA until the force displacement response of post-damage initiation simulation do not achieve certain accuracy at several key points compared to experimental response.
Process Optimizer: This is a tailored made app to show process optimization capability and hence exemplary
Download Volvo PO:
https://drive.google.com/open?id=1C84Mhe4CtWWfWruwytcqxZPYpmmYqjl6
​
A set of simulation was performed on a specific deep drawing application using Autoform Sigma. Out of many simulation variables, the friction coefficient of the punch and the blank holder and biaxial material anisotropy are some variables' data provided by the sheet metal suppliers. Due to variation on these properties, the final part may split (fail) under pre-defined process conditions. The split is expected when major and minor strain on a segment of the part exceed the forming limit curve (FLC). The blank holder force (BHF) can be adjusted to prevent such undesirables.
In a number of simulations, above mentioned four process variables, i.e. frinction coefficients, anisotropy and BHF were varied in a range and resultant major and minor strains were recorded. It was then possible to train an artificial neural network (ANN) relating these simulation inputs and outputs. The trained ANN could produce the major and minor strain with very good accuracy for a given set of four input process variables without need for further simulations.
Next step was to find for a given frictions and anisotropy, which BHF can avoid split. The default BHF of the process is first tried in ANN and major minor strains were check against FLC. If split is probable, the BHF is reduced as reduction in BHF helps to avoid split (with risk for wrinkling). The new reduced BHF is again tried in ANN followed by similar checking against FLC. This process is continued until split is avoided. A safety factor can be added to the app to get a BHF suggestion resulting strains well below FLC.
​