You are on page 1of 33

DR.

ANDRES AMADOR GARCIA GRANADA (Orcid ID : 0000-0002-6271-2594)


Accepted Article
Article type : Article

Title: Numerical and Experimental Study of Blow & Blow for Perfume

Bottles to Predict Glass Thickness and Blank Mold Influence

Authors: Adrià Biosca1,2, Salvador Borrós3, Vicenç Pedret Clemente1 , Matthew R. Hyre4 and

Andrés-Amador García Granada2 *

1
Ramon Clemente, El Masnou, Spain
2
Grup d’Enginyeria en Producte Industrial (GEPI), Institut Químic de Sarrià - Universitat Ramon

Llull, Barcelona, Spain


3
Grup d’Enginyeria de Materials (GEMAT) Institut Químic de Sarrià - Universitat Ramon Llull,

Barcelona, Spain
4
Mathematics, University of Northwestern,, St. Paul, MN, USA

• Corresponding autor: andres.garcia@iqs.url.edu

Abstract:

Glass forming to produce perfume bottles with specific thickness distribution profiles is based on trial and

error and requires several tests in production line. These tests are expensive and time-consuming, which

increases time to market. The use of a numerical model aims to reduce the number of prototypes by performing

virtual tests of the mold equipment and the process conditions. This article presents results of numerical

simulations of the blow and blow forming process to predict glass thickness distribution. Correlation of the

simulation results of the glass temperature with experimental infrared measurements on the glass skin and

This article has been accepted for publication and undergone full peer review but has not
been through the copyediting, typesetting, pagination and proofreading process, which may
lead to differences between this version and the Version of Record. Please cite this article as
doi: 10.1111/ijag.13208
This article is protected by copyright. All rights reserved.
experimental validation of glass forming simulations and influence of the blank mold cavity in the thickness

distributions of perfume bottles are provided. Finally validation of the results of axisymmetrical and three-
Accepted Article
dimensional models for axisymmetric bottles defining a useful technique to use the right blank mold for desired

thickness distribution while reducing the trial and error testing.

Keywords: “numerical and experimental study”, “blow and blow forming process”; “glass thickness

prediction”; “perfume bottle”, “axisymmetrical and 3D models” (Web: simulation, glass, viscosity and extra

blow and blow forming process)

I. Introduction

Despite the tendency for decades to reduce the use of glass for packaging, when it comes to the perfume

industry, perfume bottles are almost all made of glass as glass preserves the properties of the product stored

inside and adds value to packaging thanks to its transparency, brightness and weight. The quality requirements

for glass perfume bottles are very demanding: containers tend to have difficult shapes (usually non-

axisymmetric) with complex glass thickness distributions and there is a low tolerance to container defects.

The industrial process for producing glass containers is called blow and blow (BB) glass forming process.

This is a complex process and a full understanding has not yet been reached. It is not possible to see the glass

flow within the molds during the blowing stages. Obtaining accurate values of glass and forming tools

properties is very difficult due to high temperatures, corrosive environment and fast-paced forming process.

The design of the mold cavities and the definition of machine operation times and process temperatures are

key issues to develop and mass produce optimal containers. The manufacture of glass containers is still based on

empirical knowledge and trial and error methodology. The use of heuristic problem-solving techniques does not

guarantee an optimal solution when only a specific amount of time and resources are available. A sub-optimal

production process can cause container defects and a decrease of the production rate.

This article is protected by copyright. All rights reserved.


For the development of a new glass perfume bottle several stages are required:

(i) The shape and specifications of the container are initially defined.
Accepted Article
(ii) All the mold equipment pieces are designed and prototypes of each piece are CNC machined in cast iron.

(iii) Machine operation times and process temperatures are defined and the mold prototypes are tested on an

individual section (IS) machine to manufacture the first units of the new glass container.

(iv) The quality of the perfume bottles formed is controlled during the tests and adjustments are made to

solve the production problems and container defects.

(v) If necessary, production tests are repeated with new designs of mold cavities and production process

configurations until the desired container quality is reached.

The development of perfume bottles with very demanding quality requirements and specific thickness

distribution profiles generally requires several iterations and production tests. The experimental iterations are

expensive, they increase time to market and they involve huge downtimes in manufacturing. The use of

numerical simulations will allow a better understanding of the production process and improve the quality of the

manufactured containers, while reducing the amount of expensive experimental iterations.

The aim of this article is to demonstrate the feasibility of numerical simulations as a tool to support the

design of mold cavities and the definition of the machine operation times for new developments of perfume

bottles. A numerical and experimental study of the blow and blow forming process has been conducted in the

Ramon Clemente (RC) (1) glassmaking plant. The results of the glass forming simulations of the perfume

bottles are presented and compared with their corresponding section profiles obtained from the industrial

manufacturing process. A numerical model capable of predicting final thickness distributions as a function of

blank mold cavities and production process conditions using axisymmetrical and three-dimensional models is

presented.

II. Previous Work on Glass Forming Simulations

The first attempts with numerical simulations in glass forming were published during the 1980s, in 1984 (2)

and especially in 1986 (3), including both the blank side and the blow side of the press and blow (PB) forming

process. The evolution of numerical techniques along with the improvement of the computational capabilities

helped the implementation of the numerical approach over years. This, together with experimentation, made it

possible to understand glass flowing and heat transfer phenomena inside the molds during the forming process.

This article is protected by copyright. All rights reserved.


In 2002 (4) numerical simulations were presented of the entire glass forming process, from the gob formation to

the final container blowing. The evolution has been extensively reviewed in the past (5,6).
Accepted Article
The glass industry has recently used glass forming simulations to study several topics:

(a) Optimization of mold temperature and cooling control using VertiFlow cooling systems (7–10).

(b) Optimization of the thickness distribution in order to reduce the weight of the containers and improve

their mechanical properties (11). The use of the narrow neck press and blow (NNPB) forming process for

lightweighting is a common practice in glass containers for the food industry. Current food containers have

experienced a weight reduction of more than a 50% compared to those of two decades ago (12).

(c) Study the causes of the known container defects and find solutions to improve production rates of

glassmaking (13).

(d) Support experimental studies to measure the thermal and mechanical properties at the glass-mold contact

interface (e.g. heat transfer (14–16)) and how the use of lubricants is related to heat transfer (17) and friction

coefficients (18).

(e) Reverse engineering algorithms were presented to optimize the design of the blank mold cavity using a

Bezier curve for BB containers (19) and perfume bottles (20).

Numerical simulations are commonly used to study the NNPB forming process. Despite the differences in the

gob loading and parison forming stages, the production process remains very close to the BB forming process.

Although several authors have presented numerical results of the glass thickness distribution (21–24), there is

a lack of publications where the numerical results are compared with the profiles of manufactured container to

appreciate the reliability of the numerical results and/or pursue the validation of the numerical model. Moreover,

published work on BB forming process simulations for glass perfume bottles is rarely seen (20,25).

III. Blow and Blow Process Description

The industrial glass forming process takes place inside two different molds. When the forming process in

both molds is carried out by blowing the glass against the walls of the mold cavity, this is known as the blow

and blow (BB) forming process. As an alternative, instead of blown, glass can be pressed with a plunger into the

first mold, then it is called press and blow (PB) or narrow neck press and blow (NNPB). These two other

This article is protected by copyright. All rights reserved.


processes are used to produce jars and lightweight containers (26–28) respectively, which are generally used for

the food and drink industry.


Accepted Article
The production process begins with the fusion of the raw materials in the furnace. The melt is then refined at

the working end and conditioned in the forehearth. Afterwards the molten glass stream flows through the feeder

and is cut into discrete gobs of glass. These gobs are then distributed by the delivery equipment to each IS

machine, where the BB forming process takes place (see Fig. 1).

Once the glass gob is cut and delivered, it is loaded through the funnel into the blank mold, which is placed

upside-down. First the neck of the container is formed: a small plunger is inserted from the bottom and the settle

blow is performed by blowing pressurized air from the baffle, which has been placed on top of the funnel after

the loading stage. Afterwards, the funnel is removed and the baffle is placed on top of the blank mold, defining

the blank cavity. The plunger is then removed and pressurized air is blown from the neckring, allowing the glass

to expand throughout the blank cavity to create the parison.

At this point the parison is still upside-down. The baffle is then removed and the blank mold is opened

allowing the parison, still held by the neck, to be inverted and transferred into the blow mold by the swing of a

mechanical arm. At that moment, the blow mold closes, the neckring opens and the parison remains hanging on

top of the blow mold by the neck. The blow head is placed on top of the mold, blowing the glass against the

walls of the blow cavity to create the final shape of the container.

Finally, the blow mold is opened and the tong takes out the container to a conveyor belt. The manufactured

containers are sent through an annealing lehr, where the cooling rate is controlled to release glass internal

stresses. After the quality control inspection, the containers are packed in boxes and pallets.

This article is protected by copyright. All rights reserved.


IV. Mathematical and Numerical Models
Accepted Article (1) Governing Equations

Defining the glass domain under the forming process involves the solution of differential equations that

describe the conservation of mass, momentum and energy of the flow. Glass is assumed to be an incompressible

fluid.

The momentum equation for an incompressible fluid is reduced to

∇· =0 (1)

where is the velocity vector.

For an incompressible fluid the conservation of momentum yields

= −∇ − ∇ · + (2)

where is the density, the operator D⁄D is the material time derivative, is the pressure, is the viscous

stresses tensor and is the volume force. The viscous stresses can be combined with the pressure into a total

stress tensor as shown below

= + (3)

where is the unit tensor.

The energy equation for an incompressible fluid is given by

=− ∇· + − : ∇v (4)

where is the heat capacity, is the temperature, is the heat flux, is the volumetric external heat source

and : ∇v is the viscous heating rate.

Glass has been modeled as a generalized Newtonian fluid. The constitutive equations for the glass domain

defined as a generalized Newtonian fluid are

= −2 , (5)

where is the viscosity, is the shear rate and is the deformation tensor, defined by

This article is protected by copyright. All rights reserved.


= + (6)
Accepted Article
where is the dyadic product or the velocity gradient tensor and is its transpose. Finally, the shear rate

is defined by


= 2 : (7)

(2) Boundary Conditions

The boundary conditions for fluid dynamics in the flow domain Ω can be written as either velocity

components (Dirichlet BC) or surface traction components (Newmann BC)

= , for x ∈ Ω (8a)

· = ̅ , for x ∈ Ω (8b)

where is the outward vector normal to the boundary, and ̅ are specified functions.

Thermal boundary conditions at the glass mold interface are the most important boundary conditions of the

glass forming problem. These can be defined specifying the temperature (Dirichlet BC) or the heat flux

(Newmann BC) at the boundaries

= , for x ∈ Ω (9a)

= , + + for x ∈ Ω (9b)

where d/d is the derivative normal to the boundary, and are specified functions and and are the

convective and radiative components of the heat flux.

(3) Free Surfaces and Contact Detection

Free surfaces are used to track the displacements of the boundaries of the glass domain during the blowing

stages. Before contact with the mold cavity, dynamic and kinematic conditions define the displacement of the

free surface.

This article is protected by copyright. All rights reserved.


For the dynamic condition the normal force requires to be a prescribed value. In general, the prescribed value

will be zero, however, in the blowing stages it will be defined as the blowing pressure.
Accepted Article
The kinematic equation for time-dependent problems is defined as

· = , · (10)

where is the normal vector to the free surface and , describes the position of the free surface.

Contact detection between the moving surfaces and the walls of the mold is implemented through a penalty

technique (29). At each time step, each node on the glass domain is tested to determine whether it is in contact

with the mold walls. Once contact is detected, the discrete momentum equation of the nodes is then modified by

a penalty term to avoid penetrations. In the normal direction to the contact surface the following condition is

applied

=− − · (11)

where is the normal force vector, is the penalty coefficient and and are the velocities of the glass

nodes and the mold, respectively.

Using the same principle in the tangential direction, a friction coefficient between glass and mold can be

defined as

=− − , (12)

where is the tangential force vector and is the slip coefficient. When is equal to a sticky contact

is defined. If is smaller than then the slipping of glass on the surface of the mold is allowed.

(4) Numerical Solution

To solve the conservation equations, the constitutive equations and the boundary conditions presented before

the commercial FEM code ANSYS Polyflow has been used. The numerical simulations describe the glass flow

during the BB glass forming process. An axisymmetrical and a three-dimensional model have been implemented

to account for complex perfume bottle shapes.

This article is protected by copyright. All rights reserved.


To handle large displacements of the free surfaces during blowing stages a Lagrangian remeshing technique

has been used. Volume conservation is verified so as not to modify the initial volume of the glass domain due to
Accepted Article
remeshing. A penalty technique is defined for contact detection of the free surfaces with the walls of the mold.

Planes of symmetry have been used whenever possible to reduce the computational cost of the simulations:

only a section and ¼ of the domain have been simulated in the axisymmetrical and in the three-dimensional

models, respectively.

V. Relevant Material Properties

(1) Viscosity

Viscosity is the most important property in glassmaking, not only because it defines the resistance of glass to

flow, but also because it has a strong dependence on temperature. The viscosity together with the temperature

distribution strongly influences the glass flow in the forming process, especially during the blowing stages.

Glass is an amorphous solid, so it has no crystalline structure or melting point therefore. This means that the

viscosity of the glass, or its resistance to flow, varies continuously with temperature: from 103 dPa·s around

1,200 ºC where glass behaves like a fluid to 109 dPa·s around 650 ºC where the viscosity is so high that glass

loses the ability to flow. The working range for the glass forming process can be defined as a function of

temperature for a defined glass composition (30).

Glass viscosity is a very sensitive material property, not only to the changes in temperature and glass

composition, but also to variations in the furnace operation, raw material suppliers or weather conditions. Glass

viscosity is periodically characterized in external laboratories to control viscosity changes over time and

maintain a specific working range (see Fig. 2).

The strong temperature dependence on glass viscosity is commonly expressed using the VFT (Vogel-

Fulcher-Tamman) equation (31)

log = + (13)

where , and are empirical coefficients that must be defined for each glass composition.

This article is protected by copyright. All rights reserved.


The non-Newtonian behavior of glass due to shear thinning has been characterized in the past (32). Although,

these phenomena only begin to become important as the glass temperature descends near its transition
Accepted Article
temperature and glass is subjected to high stresses.

The viscoelastic effects may be relevant at the NNPB pressing stage, where they can be modeled to predict

defects as the glass fills the neckring (6,13). Viscoelastic effects and structural relaxation models have been

taken into account in the numerical simulations of the TV panel pressing process to predict the glass properties

at the very end of the forming process (33). Although, these phenomena only begin to become important as the

glass temperature descends near its transition temperature and glass is subjected to high stresses.

At the end of the BB forming process, when the final blow finishes and the blow mold is opened, viscosity

values are high enough to withstand gravity and low deformation effects. At the same time, the values of

viscosity of the glass container are still under the annealing point, so the internal stresses can be quickly relaxed.

Thus,, glass can be considered a Newtonian fluid for the modeling of the BB forming process. Therefore, the

elastic effects are not taken into account.

(2) Effective Thermal Conductivity

As indicated in Eq. (9b), the heat flux within the glass domain and its boundaries is defined by conduction

and radiation components.

During forming operations inside the molds, the semitransparent internal radiation of the glass has been

approximated using an effective conductivity that accounts for the extra amount of heat due to the internal

radiation. The radiative component has been approximated as a diffusive process and has been added to the

Fourier’s law conductivity term (34–37).

Radiation effects are not negligible and the parison is not optically thick, but using an effective conductivity

based on temperature and thickness (38) compensate for radiative effects during forming. This approximation is

valid as the heat flux to the mold that is due to semitransparent radiation is, on average, an order of magnitude

less than that which is due to contact conductance (39). However, the amount of error introduced for the heat

This article is protected by copyright. All rights reserved.


transfer in the glass-air interface when the molds are open may be larger (i.e. during the reheating stage before,

during and after the parison inversion)


Accepted Article
(3) Heat Transfer at Glass-Mold Interface

The most complex phenomenon in the simulations of the glass container forming process is the thermal

contact between the glass and the molds. In the glass-mold interface, the main mode of heat transfer is contact

conductance. There is a contact resistance that causes convective heat transfer to deviate from perfect contact.

This resistance is the result of a thin gap of air and combustion products of the lubricant at the glass-mold

interface. The thickness of this gas layer depends on the surface finish of the mold and varies according to the

process conditions: the separation between glass and mold increases due to the thermal contraction of the glass

and decreases due to the blowing pressure.

Several authors have performed extensive experimental studies of these phenomena to characterize the heat

flux at the glass-mold contact interface as a function of the glass and mold conditions (14–16) and (36–45).

Although these studies have shown that heat transfer may be a function of the operating conditions,

temperatures, material properties, pressure, contact time, etc. there is not yet an accurate model to predict these

phenomena on the glass forming simulations. Until a full understanding is reached, an accepted approach is to

define heat transfer with a coefficient. The heat transfer coefficient is usually written as a function of time and

adjusted to fit the experimental data (46). (see Fig. 3).

VI. Results and Discussion

(1) Problem Definition

To validate the numerical glass forming model under different forming process conditions, two perfume

bottles from Ramon Clemente (RC) with differentiated shapes, specifications, machine operation times and

process temperatures have been selected (see Table 1). The IS forming process cycle refers to the total amount

of time needed to produce a single perfume bottle in a section. These two examples are representative of the

wide range of RC most common containers and forming process conditions.

This article is protected by copyright. All rights reserved.


The choice of these two perfume bottles has been made for different reasons:
Accepted Article
(a) From the point of view of the design of the mold cavity different approaches were followed. In the case of

SIERRA 60, it was intended to obtain a bottle with a uniform glass thickness distribution throughout the section

profile. The main priority was to avoid thin and fragile areas, especially at the edges and vertices of the bottle.

Instead, ALFA 200 followed the opposite principle. The trend of recent years in perfume bottles is to obtain an

overweighed aspect (47). This is used to enhance the aesthetic properties of glass and increase the perceived

value of the final product. The main objective was to accumulate a large amount of glass at the bottom of the

bottle to obtain a thick base using a limited amount of glass.

(b) The conditions of the production process depend very much on the container shape and the glass weight.

The production machine rates, the process temperatures, the gob specifications and the machine operation times

must be adjusted accordingly to form the optimal gob and parison for each specific container; which implies

adjustments to all the conditions of the forming process. This was taken into account when defining the glass

forming problem.

(c) The numerical models of the two perfume bottles have different “scales” of time and volume (see Table

1). The definition of the boundary conditions at the glass-mold interface as a function of contact time (i.e. heat

transfer coefficient), together with the use of actual machine operation times is necessary to obtain accurate

results. Depending on the domains sizes and the total simulation time, different values of time step and mesh

size were defined to have reasonable computation times without compromising accuracy.

Covering all these aspects was important to ensure that the numerical model can handle glass forming

simulations in a wide range of scenarios.

(2) Methodology

In order to carry out the simulations of the BB forming process for a glass perfume bottle, it is necessary

previously to gather information about the glass, the bottle and its production process. Numerical glass forming

models require accurate data of temperature-dependent material properties and additional parameters that are

often difficult to obtain. Most of this information must come from the production line of the glassmaker

manufacturer: process conditions, glass temperatures and properties, machine operation times, mold cavity

shapes, etc. To make experimental measurements of glass in the forming process is a complex and time-

consuming task. And in the event that it is possible, it requires quite expensive and fragile measuring devices.

This article is protected by copyright. All rights reserved.


However, these are key problems in modeling, not only to adequately feed the numerical model but also to be

able to criticize the obtained results and validate them. In short, to have confidence with the glass forming
Accepted Article
numerical model and the predicted glass thickness distributions.

In order to perform the numerical simulations of the BB glass forming process, the following information

must be specified in the numerical model: geometry of the molds and glass domains, glass material properties,

boundary conditions, production process parameters and temperatures of the glass and molds:

(a) The geometrical boundaries of the blank and blow mold domains are defined from the cavities of the mold

equipment. The glass domain needs to be defined too; this procedure will be explained later.

(b) Material properties for the glass domain are defined from experimental data. Tests were performed in an

external laboratory to characterize the viscosity of the RC glass Coefficients of the VFT viscosity equation

(Eq. 13) expressed in log η (dPa·s) are A = -0.026, B = 5200 and To = 200 (see May2016 plot in Fig. 2 with

regression R2=0.9702). Density and the specific heat of the glass are assumed constant and are defined from the

chemical composition of the RC glass. Effective conductivity has been defined as a function of glass

temperature using a UDF function, dependence of the thickness is not taken into account in the numerical

model. Effective conductivity values expressed in W/m·ºK are in a range from 4.25 at 1,027ºC to 2.4 at 727ºC

and have been defined from previous work (4,14,15).

(c) Boundary conditions of the heat transfer coefficient at the glass-mold interface is defined as a function of

the contact time. Figure 3 shows the values used in the simulations. The heat transfer coefficient has been

defined from previous work (4,14,15) and others.

(d) Counter blow and final blow pressures are obtained from the production process. From the IS drum

rotation degrees, the machine operation times can be obtained. Then the duration of the glass forming stages

modeled in the simulations can be defined (see Table 2). The production machine rate is specified for the

production process in a single gob six IS machine. The simulation ends after the complete expansion in the blow

mold.

(e) Glass temperature measurements have been made during the forming processes for the two bottles using

an infrared thermal camera. The temperature profiles on the glass skin obtained in the different stages of the

forming process where glass is visible and can be measured using infrared will be used to verify the numerical

This article is protected by copyright. All rights reserved.


results. Figure 4 shows representative infrared captures of the SIERRA 60 bottle forming process, glass

temperature decreases from 1,030ºC to around 650ºC and different color legends have been used.
Accepted Article
Measurements during the stages of gob forming and loading are used to define the initial temperature of the

glass gob (48), which is assumed to be constant. The temperature measurements of the parison once the blank

mold is opened and until the blow mold is closed are used to follow the parison reheating. The final container

along with the parison temperature measurements can be used to roughly estimate the heat transfer from the

glass to the blank and blow molds, respectively.

Blank and blow mold temperatures are defined from bibliographic data.

(f) Finally, bottles are cut to obtain cross sections to be able to compare the thickness distribution of the

numerical simulations with the thickness distribution of the glass bottles manufactured in the production line.

(3) Numerical Simulation Description

Detailed simulation results of the axisymmetrical model of the SIERRA 60 bottle are presented at different

simulation steps of the BB forming process (see Fig. 5). The following explanation can be extended to the four

simulation cases presented.

The stages of gob forming, gob loading and settle blow are not modeled. Thus, as an initial condition, the

glass gob is considered to be inside the blank mold and the neckring cavity is filled with glass. The free surface

at the top of the loaded gob has been defined using the infrared thermal camera after the settle blow. The volume

of the glass domain is in accordance with the weight of the glass bottle. This defines the initial geometry of the

glass domain (Fig. 5A). When the simulation starts, heat transfer begins immediately at the glass-mold contact

interfaces of the walls of the blank mold cavity and the neckring. Due to the contact heat transfer, a temperature

distribution in the glass domain is defined before the counter blow stage (Fig.5B), defining in turn a viscosity

distribution that will define the glass flow. In the counter blow stage, pressure is applied and glass expands

through the blank cavity to create the parison (Fig.5C, 5D and 5E). In the blowing stage the glass domain is

remeshed several times to avoid distortion of the mesh elements due to the large deformations of the free

surfaces. After expansion, glass continues to cool in contact with the walls of the mold cavity (Fig.5F) until the

blank mold is opened. Then, the parison skin is allowed to start reheating (Fig.5G). Inversion effects are not

considered, gravity is inverted when the parison is transferred to the blow side (Fig.5H). During the parison

This article is protected by copyright. All rights reserved.


reheating, the internal temperature gradients in the glass domain redistribute the temperatures until the parison is

ready to be blown inside the blow mold (Fig.5I). Parison stretching does not have a significant effect on
Accepted Article
perfume bottles. Perfume parisons are relatively short, thin and light compared to parisons in other industries

(e.g. beer or wine bottles). The parison run distances are generally very small to be able to hide the baffle seam.

Thus, the stretching stage is not taken into account. The final blow expands the parison throughout the blow

mold cavity (Fig.5J, 5K and 5L) to define the glass thickness distribution of the final perfume bottle. The

simulation time ends after the expansion of the glass has finished. The remaining blowing time would be used to

cool the glass bottle until the blow mold is opened, but the thickness distribution is already defined.

Deformations of the bottle due to the thermal contraction of the glass or to the stress relaxation phenomenon are

not considered either.

(4) Validation of the Numerical Models

To validate the BB glass forming model under different forming process conditions, simulations of two

Ramon Clemente (RC) perfume bottles have been performed. In addition one of the two containers was tested

with two blank mold cavity designs to show their influence on the obtained glass thickness distribution. In all

the numerical simulations, the same material properties and boundary conditions were used. Mold design and

process parameters (e.g. machine operation times, process conditions and glass temperatures) for each container

were provided by the manufacturer.

It is important to realize how different both bottles and their forming processes are:

(a) The designs of the blank mold cavities seek to obtain very different glass thickness distributions, as can be

seen by comparing the parison overcapacity ratio in both blank molds.

(b) ALFA 200 is more than three times heavier than SIERRA 60, as a result, all process temperatures and

machine operation times are drastically modified.

(c) The dependence on the material properties and heat transfer at the glass-mold interface with the

temperature of the glass and the contact time are essential to obtain reliable results.

(A) SIERRA 60 Bottle

Numerical simulations of the BB forming process with for the SIERRA 60 bottle have been performed.

This article is protected by copyright. All rights reserved.


The comparison of the bottles produced with the predicted results of the numerical model shows a good

agreement for the glass thickness distribution of the perfume bottle (see Fig. 5). The glass thickness distribution
Accepted Article
predicted by the simulations correlates with the cross section profile of the glass bottle. The characteristic shape

of the glass distribution reproduces well in the results, especially in the bottom of the container.

This example demonstrates the ability of the blow and blow numerical model to predict the glass thickness

distribution of perfume bottles.

It can be stated how manufactured bottles, even axisymmetrical, do not always show a perfect symmetrical

distribution of the glass thickness due to deviations of the production process (Fig. 6D).

The reheating of the parison skin can be observed from the parison temperature measurements obtained after

the blank mold has been opened and until the blow mold is closed. These temperature measurements show a

trend similar to the numerical results of the temperature field in the glass domain. Although there is a very good

reliability in the results of the glass thickness distributions, more efforts are needed to minimize the error of the

temperature measurements (which is currently around 15%).

(B) Influence of the Blank Mold, ALFA 200 Bottle

Usually several production tests are required to test new iterations of the mold cavity designs during the

development of a new perfume bottle until the desired glass thickness distribution is obtained. For this reason,

the ability of the numerical model to predict the influence of the blank mold cavity in the glass thickness

distributions has been tested.

Different blank mold iterations were designed during the development of the ALFA 200 bottle. The

containers manufactured with the first blank mold (cavity A) showed a poor glass distribution at the base of the

bottle, more glass at the bottom of the bottle was desired. Therefore, a second blank mold (cavity B) was

designed and tested to improve the aesthetics of the ALFA 200 perfume bottle. Using the second blank mold

design, the glass thickness distribution at the bottom of the bottle was improved, a thicker and flatter glass base

was obtained and the aesthetics of the bottle were improved (see Fig. 7).

Numerical simulations of the BB forming process with the initial and the final design of the blank mold

cavity (cavities A and B, respectively) for the ALFA 200 bottle have been performed. All the forming process

conditions remained the same except for the design of the blank mold, both in the production tests and the

numerical simulations.

This article is protected by copyright. All rights reserved.


The comparison of the produced bottles with the predicted results of the numerical model shows a good

agreement for the glass thickness distribution of both perfume bottles using both blank cavities (see Fig. 8). The
Accepted Article
glass thickness distribution predicted by the simulations correlates with the cross section profiles of the glass

bottles. The modifications made to the second blank mold design (see Fig. 8A and 8E) improve the glass

thickness distribution of the numerical results and the manufactured final container (see Fig. 8C, 8D, 8G and

8H). As can be noted in the numerical results and the sectioned containers, the glass thickness on the shoulder

has been reduced and the bottom distribution has been improved: it is flatter and the bottom corners are more

rounded. The characteristic shape of the glass distribution reproduces well in the numerical results using for

both blank molds, especially in the bottom of the containers and results in an important improvement of the

glass thickness distribution at the base of the bottle. The small differences in the shape of the blank mold design

substantially improve the glass thickness distribution of the containers, this can be stated with the manufactured

bottles and the numerical model results.

This example demonstrates the capability of the blow and blow numerical model to predict the influence of

the blank mold cavity in the thickness distributions of perfume bottles.

(C) Axisymmetrical and Three-Dimensional Models Comparison

Glass perfume bottles tend to have difficult non-axisymmetric shapes. Many of the new RC developments are

non-axisymmetric bottles. Thus it is important to also develop a three-dimensional model for the simulations of

the BB forming process capable of handling all types of perfume bottle shapes.

For this reason, the simulation of the blowing stages of the SIERRA 60 bottle has been repeated using the

same forming process conditions. But instead of using the axisymmetrical model already presented, a three-

dimensional model has been used.

The comparison of the predicted results of the axisymmetrical model with the three-dimensional model

shows identical results for the glass thickness distribution for the perfume SIERRA 60 bottle (see Fig. 9). The

same results of the glass thickness distribution have been obtained. Although, it is important to state how the

large amount of mesh elements used in the three-dimensional model greatly increases the computational time.

This example demonstrates the ability of the glass forming model to handle all types of perfume bottles,

using the axisymmetrical or the three-dimensional models.

This article is protected by copyright. All rights reserved.


In the future, additional work will be carried out using the three-dimensional model to predict the glass

thickness distribution of non-axisymmetrical bottles.


Accepted Article
(5) Sensitivity Study

(A) Effect of Initial Glass Temperature

As mentioned above, initial gob temperature has been determined from experimental measurements during

the gob forming and loading stages using an infrared thermal camera. By varying only the initial glass

temperature and maintaining all other process parameters, such as glass viscosity and blank and blow mold

temperatures, the model only shows slight differences in the glass thickness distribution results (see Fig. S1).

This figure shows that thickness distribution is similar for three different initial temperatures. The purpose of the

project is to define the right blank mold that leads to a desired concave, convex, flat or straight shape avoiding

thin parts. For this purpose the effect of initial temperature, within measurement errors, leads to the same shape

results.

The amount of time necessary to achieve full expansion of the glass during the counter and final blow stages

varies with the temperature of the glass, since the glass flow velocity is in turn related to the glass viscosity. But,

in all cases, the temperature distribution in the glass domain remains almost the same. Therefore, there is the

same viscosity distribution and the glass thickness distribution also remains nearly identical.

The authors believe that the model is much more sensitive to the heat transfer coefficient and to the effective

conductivity (i.e. the temperature gradients along the glass domain) than to the temperature itself. However,

more work is needed in the temperature measurements and numerical simulations to minimize the error (which

is currently around 15%).

(B) Effect of time step and mesh size

The dependence of numerical results on the mesh quality and time step has been studied due to the different

glass weight and duration of the IS forming process cycle of the two simulation models. The two bottles have

different “scales” of time and volume, it seems logical that it is not optimal to perform the BB glass forming

simulations of both bottles using the same time step and the same mesh size. Time step and mesh size must be in

accordance with the scale of the model.

This article is protected by copyright. All rights reserved.


It was also important to study how dependent the thermal problem of these two parameters is. The heat

transfer in the glass domain depends very much on both parameters. Simulations have been performed using
Accepted Article
different values of time step and mesh size, decreasing both parameters until obtaining the required precision.

Fine meshes are needed to obtain an accurate calculation of the heat transfer in the glass domain and in the

glass-mold interference, smaller time steps also lead to higher accuracy.

It can also be stated that remeshing techniques should be used with caution. As the numerical solution is

mesh size dependent, the use of finer meshes is recommended. Near the glass-mold contact interface there are

huge temperature gradients so a fine mesh is required; on the other hand, internal glass temperatures are more

even and bigger mesh elements may be used. Nevertheless, the use of remeshing techniques with different mesh

sizes to reduce the element count is not recommended. The use of a uniform mesh size for the entire glass

domain yields more accurate results (see Fig. S2).

VII. Conclusions

A numerical and experimental study of the blow and blow forming process has been carried out in the

Ramon Clemente glass manufacturing plant. The results of the glass forming simulations of the perfume bottles

are presented and compared with their corresponding section profiles obtained from the industrial manufacturing

process. The numerical model is able to predict final thickness distributions as a function of the blank mold

cavities and the production process conditions using axisymmetrical and three-dimensional models of two

selected perfume bottles. The two bottles are representative of the wide range of Ramon Clemente’s most

common perfume bottles and their forming process conditions. The numerical model can handle glass forming

simulations in a wide range of scenarios.

The numerical model predicts consistent results in the glass thickness distribution. The implementation of the

presented model will be a very useful simulation tool to acquire knowledge about the blow and blow glass

forming process and will help with the development of new glass perfume bottles. This simulation tool will be

used to guide the design of the mold equipment, specially the design of the blank mold cavity. The quality of the

manufactured containers will improve, while reducing the amount of expensive experimental iterations and time

to market.

This article is protected by copyright. All rights reserved.


In addition, the numerical models presented for the glass forming simulations are also a solid foundation to

continue working. The numerical model is undergoing continuous improvement with the study of more
Accepted Article
simulation cases based on bottles with different shapes and operation conditions. Further work is still needed to

obtain more accurate thermal data of the glass and the heat transfer in the interface with the molds. With the

help of field measurements in the production plant it will be possible to change the generic bibliographic data

used as input of the simulations to more specific production process data. For example, the conjugate heat

transfer between glass and mold has been successfully tested in the numerical model but has not yet been

implemented due to the lack of experimental or numerical data of the temperature distributions in the molds.

Acknowledgements

The authors gratefully acknowledge Matthew Hyre and Emhart Glass that provided useful experimental data

and guidance on ANSYS Polyflow and glass forming simulations; Fundació LaCaixa for its financial support

and ACCIÓ to partially fund this project with European Regional Development Funds under the Nuclis project

RD16-1-0015.

References

1. Ramon Clemente [Online]. Available: http://www.rclemente.net/.

2. Cormeau A, Cormeau I, Roose J. Numerical Simulation of Glass-blowing. Numer Anal Form Process. 1984;219–37.

3. César de Sá JMA. Numerical modelling of glass forming processes. Eng Comput. 1986;3(12):266–75.

4. Hyre MR. Numerical Simulation of Glass Forming and Conditioning. J Am Ceram Soc. 2002;85(5):1047–56.

5. Brown M. Computer simulation of the glass pressing process: A review. Int J Mater Prod Technol. 2008;33(4):335–48.

6. Choudhary MK, Venuturumilli R, Hyre MR. Mathematical Modeling of Flow and Heat Transfer Phenomena in Glass Melting ,

Delivery , and Forming Processes. Int J Appl Glas Sci. 2010;1(2):188–214.

7. Lankeu-Ngankeu PS. Improvements to Emhart Glass Vertiflow mold cooling applications in glass container production. Conf

Glas Probl. 2016;37(1):185–91.

8. Warude A. Analysis of glass mold to enhance rate of heat transfer. University of South Florida; 2004.

9. Hyre MR. Matching Mould Cooling Strategies to the Blank Side Forming Process. Glass International. 2007;(June):76–9.

10. Bewer T. Optimisation of the moulds cooling process. Glass International. 2007;(September):57–9.

11. Ferguson S. Stronger, lighter glass bottles; Bottero Spa is changing the way bottles are made and using simulation to do it. Design

Engineering. 2017.

12. Grayhurst P, Girling PJ. Packaging of Food in Glass Containers. In: Food and Beverage Packaging Technology. 2011. p. 137–56.

13. Hyre MR. Modelling defect creation during the forming process. Glass International. 2009;(November):55–7.

This article is protected by copyright. All rights reserved.


14. Hyre MR. Effect of Mold to Glass Heat Transfer on Glass Container Forming. In: Advances in Fusion and Processing of Glass III.

2004. p. 271–9.
Accepted Article
15. Hyre MR, Underwood BL. Experimental Measurement of Glass to Mold Heat Transfer for Computational Modeling of Container

Production. In: ASME International Mechanical Engineering Congress and Exposition. 2004. p. 319–24.

16. Moreau P, César de Sá JMA, Grégoire S, Lochegnies D. Integration of Heat Transfer Coefficient in Glass Forming Modeling With

Special Interface Element. AIP Conf Proc. 2007;908:750–64.

17. Moreau P, Lochegnies D, Grégoire S, César de Sá JMA. Analysis of lubrication in glass blowing: Heat transfer measurements and

impact on forming. Glas Technol. 2008;(February).

18. Moreau P, Montoya D, Lochegnies D, Vivier H. Friction analysis at the glass/tool interface and the lubrication influence during

hot forming cycles. In: Proceedings of the Institution of Mechanical Engineers, Part J: Journal of Engineering Tribology. 2013. p.

1253–60.

19. Groot JAWM, Giannopapa CG, Mattheij RMM. A Numerical Shape Optimisation Method for Blowing Glass Bottles. In: ASME

Pressure Vessels and Piping Conference. 2011. p. 497–506.

20. Lochegnies D, Moreau P, Guilbaut R. A reverse engineering approach to the design of the blank mould for the glass blow and

blow process. Glas Technol. 2005;46(2):116–20.

21. Giannopapa CG, Groot JAWM. Modeling the Blow-Blow Forming Process in Glass Container Manufacturing: A Comparison

Between Computations and Experiments. J Fluids Eng - Trans ASME. 2011;133(2):21103.

22. Martins B, Machado M, Reis A, César de Sá JMA. Coupled thermomechanical model for press and blow forming processes of

glass containers. In: Congress on Numerical Methods in Engineering. 2017.

23. Ryzhakov PB, García J, Oñate E. Lagrangian finite element model for the 3D simulation of glass forming processes. Comput

Struct. 2016;177:126–40.

24. Feulvarch E, Moulin N, Saillard P, Lornage T, Bergheau JM. 3D simulation of glass forming process. J Mater Process Technol.

2005;164–165:1197–203.

25. Kloosterman G, Bouwman V. Automatic Remeshing and Rezoning for the Simulation of 3D Glass Bottle Forming with

Abaqus/CAE, Abaqus/Standard and Abaqus/Explicit. Int J Mater Form. 2009;2(SUPPL. 1):105–8.

26. Griffel H. Lightweighting in the Glass Container Industry. In: Conference on Glass Problems. 1987. p. 156–70.

27. Fenton A, Wiegand O, Mann KH. Light weighting containers. Glass International. 2018; 41(6):12-14.

28. Balan Mendoza Jaime S, Alves Ortiz S, Bassani Dantas T, Ferreira Damasceno C. A Comparison of the Performance of

Lightweight Glass Containers Manufactured by the P&B and B&B Processes. Packag Technol Sci. 2002;15(4):225–30.

29. ANSYS Polyflow User’s Guide. In ANSYS, Inc.; 2013. p. 395–413.

30. Fluegel A. Glass viscosity calculation based on a global statistical modelling approach. Glas Technol - Eur J Glas Sci Technol Part

A. 2007;48(1):13–30.

31. Fulcher GS. Analysis of Recent Measurements of the Viscosity of Glasses. J Am Ceram Soc. 1925;8(6):339–355.

32. Simmons JH, Ochoa R, Simmons KD, Mills JJ. Non-Newtonian Viscous Flow in Soda-Lime-Silica Glass at Forming and

Annealing Temperatures. J Non Cryst Solids. 1988;105:313–22.

33. Camp OO Den, Hegen D, Haagh G, Limpens M. TV Panel Production : Simulation of the Forming Process. In: Conference on

Glass Problems. 2003. p. 1–19.

34. Rawson H. Physics of glass manufacturing processes. Phys in Technol. 1974;91(5(2)):91–114.

This article is protected by copyright. All rights reserved.


35. Granger KC, Gardon R. Interaction of Radiation and Conduction in Glass. J Am Ceram Soc. 1969;52(10):548–53.

36. Trier W. Temperature Distribution and Heat Flow in Glass in Blank Molds of Container Machines. J Am Ceram Soc.
Accepted Article
1961;44(7):339–45.

37. Manthuruthil J, Sikri TR, Simmons GA. Simplified Mathematical Model Simulating Heat Transfer in Glass-Forming Molds. J Am

Ceram Soc. 1974;57(8):345–50.

38. McGraw DA. Transfer of Heat in Glass During Forming. J Am Ceram Soc. 1961;44(7):353–63.

39. Storck K, Karlsson M, Loyd D. Analysis of the blank mould - a transient heat transfer problem in glass forming. Trans Eng Sci.

1994;5:175–82.

40. Shetterly DM, Huff NT. Mold Surface Temperatures During Glass Container Forming. J Non Cryst Solids. 1980;38–39:867–72.

41. Huff NT, Shetterly DM, Hibbits LC. Glass to Metal Heat Flow During Glass Container Forming. J Non Cryst Solids. 1980;38–

39:873–8.

42. Pchelyakov SK, Guloyan YA. Heat Transfer at the Glass-Mold Interface. Steko i Keramika. 1985;(9):14–5.

43. Farag IH, Beliveau MJ, Curran RL. Heat Transfer During Glass Forming. Chem Eng Commun. 1987;52:21–32.

44. Grégoire S, César de Sá JMA, Moreau P, Lochegnies D. Modelling of heat transfer at glass/mould interface in press and blow

forming processes. Comput Struct. 2007;85:1194–205.

45. Fellows CJ, Shaw F. A Laboratory Investigation of Glass to Mould Heat Transfer During Pressing. Glas Technol. 1978;19:4–9.

46. Hyre MR. Container forming modeling: Sensitivity to boundary conditions, material properties, and underlying physics modelling.

In: Seminar presented at the “Jornada sobre Simulación en Procesos de Fabricación de Vidrio” at IQS, Barcelona, Spain. IQS,

Barcelona; 2017.

47. Redding M. Transforming glass for fragrances and skincare; glass bottles and jars always look elegant. Beauty Packaging. 2017;

48. Biosca A, Borrós S, Pedret V, García A-A. Glass Gob Modeling and Experimental Validation Using a Drop Test. In: MATEC

Web Conf (IC4M). 2018. p. 1–6.

Figure Caption List

Figure 1. Container forming IS machine stages of the blow and blow forming process: (A) gob loading, (B)

settle blow, (C) counter blow, (D) inversion, (E) reheat and stretch, (F) final blow and (G) take out.

Figure 2. Temperature dependence on viscosity plotted on a logarithmic scale. Different RC glass samples as

example of viscosity variation over years.

Figure 3. Heat transfer coefficient at the glass-mold interface used in the numerical simulations It is defined

as a function of the contact time.

Figure 4. Infrared thermal captures of the glass forming stages were glass can be seen: (A) and (B) gob

forming, (C) gob loading, (D) blank mold open, (E) parison reheating, (F) blow mold open, (G) and (H) take

out.

This article is protected by copyright. All rights reserved.


Figure 5. Numerical results of the evolution of the glass domain and its temperature throughout the forming

process of the SIERRA 60 bottle.


Accepted Article
Figure 6. Numerical results of the (A) initial simulation step, (B) parison and (C) final simulation bottle

compared with a (D) cross section profile of the SIERRA 60 real bottle.

Figure 7. ALFA 200 manufactured bottles using the two blank molds: blank mold cavity A (left) and blank

mold cavity B (right).

Figure 8. Numerical results of the initial step, parison and final bottle compared with a cross section profile

of the bottle ALFA 200. Results of blank mold cavity A top (A) initial step (B) parison, (C) final bottle in

simulation (D) final bottle experimental and cavity B bottom (E) initial step (F) parison, (G) final bottle in

simulation (H) final bottle experimental.

Figure 9. Numerical results using axisymmetrical model (left) of the (A) initial step, (B) parison and (C)

final compared to three-dimensional model (right) (D) initial step, (E) parison and (F) final of the SIERRA 60

bottle.

Figure S1. Influence of the glass gob temperature on the final glass thickness distribution results. (A) is the

experimental measured temperature (1,030ºC), (B) adds +50ºC and (C) adds +100ºC to the initial glass gob

temperature.

Figure S2. Influence of the mesh sizes and remeshing algorithms on the glass temperature results in the

glass-mold interface. Temperature results using a uniform mesh size (top) and a non-uniform mesh size

(bottom).

Table Caption List

Table 1. Specifications of the two selected perfume bottles.

Table 2. Relevant forming times to define the numerical simulation stages according to the IS machine

operation times.

This article is protected by copyright. All rights reserved.


Accepted Article Table 1. Specifications of the two selected perfume bottles.

SIERRA 60 ALFA 200

Brimful capacity (ml) 68 211

Mass (g) 95 293

External bottle volume (ml) 107.5 328.5

Cross sectional area (cm2) 31.6 72.5

Production machine rate (bpm) 58 32.3

IS forming process cycle (s) 11.1 20.7

Table 2. Relevant forming times to define the numerical simulation stages according to the IS machine

operation times.

SIERRA 60 ALFA 200

Gob loading (s) 0 0

Counter blow start (s) 1.7 3

Counter blow end (s) 2.9 5

Blank mold open (s) 3.1 5.7

Inversion (s) 3.9 7.7

Final blow start (s) 7 11.5

Final blow end (s) 9 17.2

Blow mold open (s) 9.6 18

Total simulation time (s) 7.3 12.2

This article is protected by copyright. All rights reserved.


Accepted Article
Accepted Article
Accepted Article
Accepted Article
Accepted Article
Accepted Article
Accepted Article
Accepted Article
Accepted Article