A robust process model with two-stage optimization methodology for liquid composite molding process

Seyednourani, Mahsa (2020) A robust process model with two-stage optimization methodology for liquid composite molding process. [Thesis]

[thumbnail of 10356928_Nourani_Mahsa_Seyed.pdf] PDF
10356928_Nourani_Mahsa_Seyed.pdf

Download (3MB)

Abstract

Liquid Composite Molding (LCM) is the family of the advanced composite manufacturing method in which dry preform is placed into the mold cavity followed by filling of the preform with the resin system. The final composite part is obtained after the curing cycle. The success of the final part for LCM highly relies on the success of the impregnation of the resin system through the dry preform. In order to have fully impregnated domain, the inlet and vent locations, namely gates, should be engineered in such a way that as the resin is introduced through the inlet gate, vent should be placed where the resin arrives last. Otherwise, the process fails with the formation voids/dry spots beyond tolerance values. Additionally, the fill time of the preform domain for complete impregnation should be reduced considering both finalization of the impregnation before gelation time of the resin and the achievement of high production rates. Thus, the LCM process can be improved through the minimization of the two aforementioned parameters: void content and fill time. To achieve that one has to predict the flow patterns within the preform. Mathematical modeling of the resin flow in LCM process is described reasonably well as flow through porous media using Darcy’s Law coupled with the continuity equation. Darcy’s Law, which relates the resin pressure gradient with the resin velocity, requires two material properties: permeability tensor of the preform and the resin viscosity. Permeability tensor is a preform property and indicates the ease of flow through the preform. Generally, for the LCM models the permeability value is assigned as a bulk property, with the assumption of uniformity of the fibrous domain. This generates simple, deterministic model but as any material property variations, geometrical variations and lay-up of the assembly generate variations in permeability values, there will be some differences between real and predicted flow patterns. Another source of these differences stems from open channels (gaps) created between the edges and corners of the mold and/or inserts and preform. These ‘race-tracking’ channels have significant effect on the flow patterns which might cause formation of voids. The other property affecting the flow patterns, viscosity, reflects the resistance of the flow of the resin system through the preform and shows variations with temperature and time. The differences between predicted and actual values caused by over-idealized modeling negatively influence any actual utility of model predictions, particularly optimization. Therefore, the accuracy of the model entails accurate permeability data with race-tracking possibilities and viscosity as a function of time and temperature. Note that at this point this is no longer deterministic model. In this study, a new LCM modeling and optimization approach is introduced which aims to optimize void content and fill time using more realistic permeability and viscosity parameters. This is achieved by a two-stage optimization approach. In the first stage, the inlet and vent location optimization are implemented with Genetic Algorithm (GA) adaptation. The GA adaptation including the permeability variations in terms of racetracking possibilities identifies the optimal inlet/s and vent/s locations working for all possible race-tracking possibilities with equal occurrence probabilities. Also, the GA adaptation enables further decrease in fill time with multiple inlet and multiple vent optimization practices. Then, in the second stage, using the gate locations from first stage the fill time and void percentage is further improved by placing a tailored highly permeable layer (distribution media, DM). This stage includes the lay-out design of the DM layer using a Discrete Optimization algorithm which dictates successful impregnation with minimum void percent and minimum fill time for all possible flow disturbances due to race-tracking issue. Then, this methodology is validated numerically by using Liquid Injection Molding Simulation (LIMS) for various complex geometries under different constraints. For the resin system the viscosity function is adapted from a commercially available epoxy system and permeability variation is defined as racetracking channels with very high permeability values at the edges. Additionally, with the use of two-stage optimization methodology, computational time is decreased due to simplifications in objective function definitions
Item Type: Thesis
Uncontrolled Keywords: Gate optimization. -- Genetic Algorithm. -- Distribution media optimization. -- Liquid Composite Molding Process (LCM).
Subjects: T Technology > TS Manufactures > TS0155-194 Production management. Operations management
Divisions: Faculty of Engineering and Natural Sciences > Academic programs > Manufacturing Systems Eng.
Faculty of Engineering and Natural Sciences
Depositing User: IC-Cataloging
Date Deposited: 04 Nov 2020 15:23
Last Modified: 26 Apr 2022 10:35
URI: https://research.sabanciuniv.edu/id/eprint/41225

Actions (login required)

View Item
View Item