Content uploaded by Anthony David Evans
Author content
All content in this area was uploaded by Anthony David Evans on Jan 09, 2020
Content may be subject to copyright.
Content uploaded by Anthony David Evans
Author content
All content in this area was uploaded by Anthony David Evans on Jan 09, 2020
Content may be subject to copyright.
TWENTY-SECOND INTERNATIONAL CONFERENCE ON COMPOSITE MATERIALS
(ICCM22)
DEVELOPMENT OF AUTOMATED DRY FIBRE PLACEMENT FOR
HIGH RATE DEPOSITION
Anthony D. Evans*, Thomas A. Turner and Andreas Endruweit
Composites Research Group, Faculty of Engineering,
University of Nottingham, NG7 2GX
* Corresponding author (anthony.evans@nottingham.ac.uk)
Keywords: Automated Dry Fibre Placement, Aerospace, High Rate Manufacture, Digital Twin,
Online Inspection
ABSTRACT
The details for a lab scale automated dry fibre placement rig is given in this paper with the
information on the real-time data transfer and high rate control (0.1ms) of the temperature and positional
adjustments. This was designed to operate at speeds up to 3m/s deposition to determine the effects of
machine parameters that surpass the limits of existing commercial machines. This was achieved by using
EtherCAT technology to provide high sample rates and communications speeds between control and
storage devices used for this rig. It assesses the current limitations to the deposition rate with the
objective of reducing cycle time. Additionally, experimentation has been performed to demonstrate high
heating rates (>1000°C/s) and temperature control via a Joule heating approach, suitable for heating dry
carbon fibre tows at response rates <1ms.
1 INTRODUCTION
1.1 Background
As aircraft production rates increase towards 60 per month [1], the demand increases for highly
automated manufacturing processes, such as Automated Fibre Placement (AFP) and Automated Tape
Layup (ATL). These aim to produce repeatable complex laminate structures with aerospace precision.
These processes are prolific for the production of wing skins and fuselages [2]. There is also recent
interest within the automotive industry to utilize the low material wastage and general automatability by
producing net-shaped preforms. These often require a secondary processing stage, such as forming or
overmoulding, to enable higher levels of part complexity [3, 4]. However, these processes are still
limited by the high material and processing costs to produce thermoset and thermoplastic prepreg slit-
tapes. This has led to the demand for a dry fibre alternative. One option is the process referred to as
automated dry fibre placement (ADFP), using binder coated tows or tapes which are deposited in strips
before infusion. As a result of the low cohesion of these materials to the substrate, in comparison to
tacky prepreg slit-tapes, increased fibre fuzzing and material variability, preforms are more susceptible
to misalignment and ply gaps/overlaps. Traditionally, these therefore require greater in-process
intervention or reduced deposition speeds to meet aerospace quality standards.
There is no standardised approach to machine architecture of AFP or ADFP, such as gantry or 6-axis
robot with head mounted creels or creel cabinets. However, offline programming takes into account the
machine kinematics (which may vary widely) to improve accuracy. The development of a component
by fibre placement, typically follow that shown in Fig. 1, whereby the part is designed to the required
specification, a tool-path is then generated from this to produce the preform before an infusion or
consolidation stage and then inspecting the part (often by NDT methods). This is very costly, with
considerable time and investment between the part design stage and the part inspection to determine
how closely the final part resembles the initial design. This therefore depends on very accurate models
of machine and material behaviour. In addition to that, consideration must be made for the transfer of a
part geometry into the machine tool path to establish the best approach to achieving the desired fibre
orientations, not characteristic of fabric forming processes, for example, whether the tool path follows
a geodesic or parallel strategies [5].
Figure 1: Flow diagram of the existing ADFP part production.
ADFP process simulations are often complex with an extensive number of factors compounding into
an overall part variability with no singular root-cause. Some of these factors include properties, which
may be characterised analytically or experimentally, such as the materials thermal response [6], stiffness
parameters [7, 8], tack performance [9, 10] or the compaction of pressure distribution of a particular
machine/roller configuration [11]. However, a greater number of dynamic factors are overlooked
because of their complexity and applicability to specific machine configurations or part geometry, and
are therefore considered unfeasible to predict for every possible scenario. Examples of these include
machine compliance, component wear, material variability caused by bobbin unwinding or
manufacturing method, variability of previously deposited layers, latency or low cycle rates within the
control and inaccuracies in predefined tool paths. This paper proposes that closed-loop control may be
applied to alter motion and temperature parameters based on sensor data. It details a method of numerical
control (NC) interpolation that enables decisions and alterations to be made via a PLC by locating them
within the same industrial PC (IPC). This is opposed to the conventional G-code approaches, which
require motion, feed and other output commands to be predefined, preventing feedback driven control
from being utilised except to interrupt the process in the event of an error.
1.2 Aims and Objectives
This paper proposes a machine architecture independent, virtual process sequence for ADFP, which
improves the control over deposition and has the potential to improve production rate by combining
both real-time (RT) control (0.1ms) and online digital twin for simulation and inspection. This is shown
in Fig. 2, where the coloured process blocks of the flow chart indicate the current areas of interest, whilst
the highlighted region is the focus of this paper.
A high precision, high rate ADFP rig was also developed to demonstrate this at deposition rates up
to 3m/s with a range of sensor data outputs for course-by-course control, and feedback driven online
toolpath generation in the PLC in addition to the digital twin. This data enables further support with
permeability simulation, quality inspection and machine learning.
This paper details the motivation behind these developments and the current limitations are overcome
by introducing closed-loop not existing in typical AFP/ADFP machines. In addition, there will be focus
on the data acquisition, which both the research and industry applications will benefit from the
generation of a digital twin and sensor feedback. This will enable both positional control using surface
measurements as well as temperature control via measured temperature data rather than constant
temperature or power control methods [6].
Figure 2: Proposed flow diagram of the ADFP part production, from design to certification.
2 ADFP RIG DEVELOPMENT
A 4-axis deposition rig (Fig. 3) has been developed for high-speed deposition at a lab-scale. The
deposition rate of this machine aims to achieve 3m/s for 1000mm centrally within the 2500mm x 600mm
tool area. This therefore requires an acceleration of 9m/s2. It is designed to deposit up to four 6.35mm
(1/4”) tows, but can also deposit two 12.7mm (1/2”) or one 25.4mm (1”) dry tow with some minor
adjustments to the belt configuration.
a)
b)
c)
Figure 3: Developed ADFP rig for high-speed deposition
2.1 Dynamics – High Rate Deposition
Although benefits of high deposition rates are evident, existing AFP and ADFP processes are often
restricted up to 1000mm/s [6, 12]. This is because of the dynamic limitations, such as machine rigidity
or the inertial response of large deposition heads, or physical limitations, such as the material adhesion
or response rate of the control system. The ADFP rig is developed to overcome these issues by
minimising the mass of the deposition head. Positioning the creel within a separate module reduces the
head weight by >30kg for a 4 tow system, however fibres must be fed through a conductive channel to
prevent variable distances between the creel and the head, and the friction that is caused by the
electrostatic forces within the channel. The channels also constrain the tows to prevent fibre damage
leading to fibre fuzzing by passing the tows through guides and eyelets. The fibre tension is controlled
independently by servo motors, resulting in a greater accuracy and responsiveness for the control.
A Joule heating system has been incorporated into the head, which increases the fibre length, but
enables both heating between the electrodes and heating between the electrode and the tool surface to
minimise the temperature losses before the nip point. Prior to electrodes fibres are cut using a module
that matches the deposition speed by engaging a shaft driven by the fibre feed and deposition rollers to
the cutter module. This will enable investigation into the effects of cutting speed on the tows quality,
whereby the cutting speed is usually reduced (less than 0.5m/s) in existing systems. After the cutter
modules, a large reinforced silicon belt constrains tows through to the nip point. This is to prevent tow
wander, a phenomenon experienced when the tows are cut and tension is lost causing them to migrate
laterally to the deposition direction.
The four belts and tows are then delivered to the nip point, a segmented roller enables ±2mm
independent travel in the Z-direction to improve pressure distribution over non-uniform geometry. This
enables a low compliance material to be used, which will constrain the lateral position of the belts. The
compaction force of the segmented roller is controlled by the servo motor which may either be torque
or positionally controlled (to apply a force up to 1kN). Trailing this roller, a pneumatic actuator applies
a force (up to 700N) to a second deposition roller, which continues to apply a force to the deposited
material for a greater duration once making contact with the tool surface. This roller is also retractable,
to prevent shearing of the deposited material in the event of steering the deposition. The deposition belt
connects between the two rollers, ensuring that pressure is maintained between them whilst the tows
begin to cool and adhere to the substrate.
2.2 Sensors
The primary objectives of this rig are to further understand and characterise the manufacturing
process and the variability between a part design and the actual outcome in the form of a digital twin.
To achieve this, the rig features three IR thermometers, axis encoders, a laser line scanner, a positional
micrometre and a force torque sensor.
The contactless temperature measurement of the fibre before and after passing through the electrodes
enables heating to be controlled and adjusted on-the-fly during the deposition process. These sensors
are mounted to a pneumatic actuator to adjust their position, which enables different tows to be
investigated. The third thermometer measures the temperature of the substrate ahead of the nip point to
add a further degree of heating control, dependent on the cooling of the previously deposited material.
The axis encoders determine the position of each of the 4 axes. Although, the servo’s each contain
built-in encoders to establish position of each motor, these additional encoders measure the actual
position once compliance of the machine assembly or error due to backlash are factored into the motion
of the deposition of the head. These will enable greater accuracy to the motion of the machine whilst the
difference between servo encoders and additional encoders will provide data into the compliance and
backlash within the machine design.
The laser line scanner trails the deposition roller assembly to measure surface profiles, at a resolution
of 2µm, each used in real time to control the motion, ‘course-by-course’, using PLC interpolation of
Bezier curves rather than predefined G-code or discretized toolpaths that are used in existing offline
programmed layup. In addition to using this data in real time, the topology, position and orientation of
each tow are transferred to a ring buffer and exported to a servo at 1s intervals via Ethernet. This enables
a generation of a digital twin without causing delay to the cyclic control, which is then transferred to
simulations, and assesses the quality of the preform in parallel to part infusion.
The positional micrometre, similar to the laser line scanner, measures the width of the tows across a
profile. However, positioned either side of the tow, enables higher sampling rates and measurements of
each tow prior to the deposition of the machine. This is currently used for characterisation purposes,
measuring the variability in the width of the tow, as well as the lateral position of the material. If
significant, this will provide further insight as to whether independent control of the lateral tow position
is required to compensate or whether changing other machine or material parameters may reduce this
impact.
A force/torque sensor, typically used in the robotic industry, obtains the forces and moments that act
on the deposition head. These provide force and torque information in all 3 orientations. This will be
used to measure both the compaction force as well as the resistance to motion in any direction. It will
also later be developed into a kinematic control for a 3D deposition rig, whereby this data will provide
a basis for determining how the deposition head is impacted by changes in tool geometry. Combining
this with a feed forward control, where the part geometry is known, it will enable how normal-to-surface
motion may be achieved on-the-fly, without the requirement of predefining the toolpath.
2.3 Control and Communication
In order to process this data, an Industrial PC (IPC) uses EtherCAT technology establish
communication between a human-machine interface (HMI), server, the PLC, inputs and outputs (I/O),
and Numerical Control (NC). The IPC is a Windows based system contains the PLC, I/O and NC within
the same location, enabling real-time data exchange between each of them. The HMI and server are
operated within a PC, which is connected to the IPC via Ethernet. The software that links the devices,
TwinCAT (Beckhoff Automation) contains a built-in feature Automation Device Specification (ADS)
which ensures that each device may send, receive and write data in real-time once communication is
established. This is the backbone of the developed rig, simultaneously performing cyclic loops at a base
rate of 0.1ms and performing cyclic tasks offset from one another to prevent latency. Combined with
the ADS, enables large quantities of data to be exchanged between devices without impacting the motion
or other RT control.
Although the EtherCAT technology allows for 0.1ms sampling rates, hardware such as the sensors
and servo drives limit the rate for which the system may operate (Table 1). These sampling times have
an effect on the adjustments during the high speed deposition processes. For example, travelling at 3m/s,
a 2ms sampling rate suggests that the minimum travel distance before a change will be made is 6mm.
Therefore, selecting a control that can surpass the hardware connected to it, prevents a ‘bottle-neck’ as
the data is processed and exchanged between devices.
Component
Sampling rate (ms)
Limit
PLC (Base Rate)
0.1
Processing power
Heating
0.5
Solid state relay (opening/closing)
Force/torque sensor
1
Data output
Encoders
1
Data output
Servo drives
2
Terminals
Positional Micrometre
2
Data output
Laser Line Scanner
10
Exposure time to image carbon
IR Thermometer
150
Data output
Table 1: Sampling rates and limits for various sensors and
Factor
Unit
Levels
Rate
m/s
0.5, 1, 3, 5
Acceleration
m/s2
0.5, 1, 5, 10
Part Area
m2
0.5, 18, 72, 288
Aspect Ratio (Length/Width)
-
1, 2, 4, 8
Tow width
mm
(inches)
3.175 (1/8”), 6.35 (1/4”), 12.7 (1/2”), 25.4 (1”)
Tows per course
-
1, 4, 8, 16
Number of plies
-
4, 8, 12, 16
Table 2: Factors and levels used for full-factorial analysis
3 METHODOLOGY
3.1 Full-factorial analysis - Rate Model
Firstly, to evaluate the rate limitations, a model determines the relative contribution of the ADFP
machine and part parameters (individually and interactions) to production rate by performing full
factorial analysis (Table 2). The model negates the dynamic limitations of the machine architecture and
assumes a zig-zig tool path of each rectangular part with a quasi-isotropic layup. As there is no standard
unit for quantifying rate of ADFP machines, separate conclusions were drawn in terms of kg/hr or parts
per annum (ppa). The rate model operates by determining duration for each course along with the length
of each tow. Assuming a bobbin contains 6kg of carbon fibre, this enabled the number of bobbin changes
to be quantified to determine the ‘downtime’ (3 minutes per bobbin change) of each structure. At this
stage, it is assumed that 100% of the tows are successfully deposited, and therefore the only downtime
experienced during manufacture is resultant from bobbin changes.
3.2 Heating experimentation
As part of the development of the high rate ADFP rig, multiple heating methods were considered:
infrared (IR) heating, laser heating, flashlamp technology and Joule heating. IR heating is a common
solution for its low cost, however the response rate can be large (up to 1s) making them only suitable
for constant temperature deposition [6]. Laser heating significantly improves this heating rate and
enables the temperature to be controlled by altering the laser power or shape [6]. However, laser heating
is expensive whilst also posing safety issues for a lab scale rig. Flashlamp technology is a relatively new
technology development, used pulsed light to enable high levels of control of temperature at the nip
point by changing duration, frequency and energy of the light [13]. Although, this improves the safety
of the heating over the laser heating, it is still an expensive option for a lab scale setup. The selected
option was Joule heating, using the electrical conductivity in the fibre direction to generate heat once an
electric current is induced into the material. This has previously been achieved for thermoplastic
materials and has shown to improve bulk factor when producing thick preforms, by reducing heat
transfer into the previously deposited materials compared to heating a nip-point, and is capable of high
heating rates (>4000°C/s) [14, 15].
To achieve deposition rates of 3m/s, heating rates of over 4500°C/s are required to heat the material
within 95mm heating length. Therefore, to test the feasibility of this heating methodology, a table top
test rig was set up (Fig. 4) to establish communication with the IPC, and measure the heating rates with
respects to voltage. This used three copper electrodes positioned 47.5mm apart to enable both 47.5mm
and 95mm heating lengths to be measured. The electrodes were compacted together via springs, and
were insulated from the bearings and other apparatus by ceramic bushes. Two IR thermometers
measured the temperature of the carbon tapes, half way between each of the electrodes.
Figure 4: Resistance heating testing rig for measuring the temperature response of carbon fibre at a
rate of 150ms using IR thermometers
In addition to hardware, a UI was created on a connected PC, which transmitted data to and from the
PLC in RT, with timestamps to ensure the measured data could be synchronized with motion data
obtained from servo motor encoders. When transferring the data to the PC cyclically, every 2s, the PLC
stored temperature data at to the ring buffer at a sampling rate governed by the limit of the three IR
thermometers (150ms) before being transferred and deleted from the buffer every cycle. This method
ensured that the PLC remained unaffected as the data was transferred between IPC and PC.
Simultaneously, a separate ring buffer stored motion data from the servo drives at a rate of 2ms to
demonstrate that the large quantities of data could be transferred by this method. The UI created is shown
in Fig. 5, and shows how the IPC communication is established, live data is added to the text boxes,
input data is sent to the IPC and graphs show temperature and motion data, updated every 2s once the
tests have started. The data is then saved as a CSV file on the PC to be post processed.
Figure 5: UI for establishing RT two-way communication and data logging between data acquired in
the PLC environment and the server (PC)
Figure 6: Graphical visualization of the I/O signal to for controlling temperature via TwinCAT
Scopeview
Within the PLC, safety limitations were added to ensure that the current was cut upon exceeding
250°C. The current was applied cyclically to a single ¼” tape (Cytec DRY UD-24K IMS65-194-6.35),
heated for 30s and cooled for 30s, three times at a range of voltages, incrementing 0.5V each time until
a steady state temperature greater than the limit temperature was achieved. Proceeding this a second
order exponent curve was fitted to the temperature/time curves for each period of heating and cooling.
This data was then extrapolated to a higher voltage (30V), whereby a PID control was added to the PLC
to demonstrate the temperature control ability. For this, a duty cycle time and frequency was used to
control the heating (Fig. 6). This uses a digital I/O signal to control a solid-state relay (SSR) at duty
cycle times as short as 1ms (minimum ton/toff =0.5ms).
4 RESULTS AND DISCUSSIONS
4.1 Full-factorial analysis - Rate Model
The percentage contributions of the rate model are shown in Fig. 7. This shows that the overall rate
of the machine is largely governed by the part size and the course width. The primary contributions are
part area of the preform (48.2%) when characterised by ppa and number of tows per course (31.3%) and
tow width (21.3%), by kg/hr. However, in this simple model, it does not factor the complexity of the
geometry, assuming it to rectangular, and therefore the reductions in rate as a result of non-uniform
pressure distribution and increased head inertia are neglected. The results highlight the difference of the
units used to measure the production rate. Parts per Annum (ppa) is a mainly influenced by part geometry
whilst the unit itself has no indication of the overall quantity of material deposited. This unit is
considered to be geometry specific and not suitable as a machine specification. It is therefore suggested
that kg/hr is better suited to quantity the rate of a fibre placement machine, as this value may be
Input
temperature &
motion
parameters
Temperature data
Motion data
Connect to PLC
toff
ton
tcycle
Digital
Signal
Time (ms)
considered an upper limit, and therefore the impact of downtime can quantified directly from this rate.
For example, if a machine is capable of 30kg/hr but the machine is stopped to correct a failed tow for
10 minutes and 4 bobbins were changed (3 minutes) during a single hour, a simple calculation would
determine the actual deposition rate to be 19 kg/hr.
Figure 7: Percentage effects of machine and part parameters.
Although the deposition rate of the machine showed some influence up to 3m/s, the increased
distance required to achieve higher deposition rates therefore depends on greater acceleration or larger
part area to benefit, indicated by the divergence of the results shown in Fig. 8a and 8b, respectively. The
mean rates for changing tow width and number of tows show near linear trends. However, the number
of bobbin changes each require generates the discrepancy between these variables. This is evident in
Fig. 8c, which shows that although multiple combinations of tow width and number achieve equal course
widths, higher rates are achieved by using wider tows as a result of the reduced number of bobbins and
therefore bobbin changes they experience even though the quantity of material per bobbin remains equal,
6kg. For example, four 25.4mm (1”) tows can achieve a greater mean deposition rate, 35.0kg/hr, than
sixteen 3.175mm (1/4”) tows, 22.6kg/hr.
a)
b)
c)
Figure 8: Mean rates across full-factorial analysis study. Results combinations of (a) rate and
acceleration, (b) rate and part area and (c) tow width and number of tows per course. The blue line
indicates the average across all results, 27kg/hr.
4.2 Joule Heating Control
The initial trials were performed to determine the effects of tow length and compaction force applied
to the electrodes to induce Joule heating into carbon fibre. Example results are shown in Fig. 9a, showing
the temperature curves for varying voltage, heating times of 30s was sufficient to achieve steady state
temperature for which 2nd order exponential curves could be fitted to determine the coefficients of
Equation 1 and 2, each with R2 values greater than 0.99.
(1)
3.81%
8.13%
5.47%
0.06%
48.18%
0.22%
0.18%
0.34%
31.34%
21.26%
0.08%
10.40%
0.97%
5.80%
0% 10% 20% 30% 40% 50% 60%
Number of Plies
Tows per Course
Tow Width
Aspect Ratio
Part Area
Acceleration
Rate
Contribution to Production Rate
kg/hr
ppa
0
10
20
30
40
50
0 1 2 3 4 5
Mean rate (kg/hr)
Rate (m/s)
0.5ms-2
1ms-2
10ms-2
5ms-2
Accerelation:
0
10
20
30
40
50
0 1 2 3 4 5
Mean rate (kg/hr)
Rate (m/s)
Part area:
0.5m218m2
288m2
72m2
0
20
40
60
80
100
0 5 10 15 20 25 30
Mean rate (kg/hr)
Tow width (mm)
1 4
816
Number of tows:
(2)
a)
b)
c)
d)
Figure 9: Joule heating results for carbon tow (L - 47.5mm, Compaction – 39N). (a) Raw data for
varying voltage, (b) fitted temperature vs. time curves, (c) fitted heating rate vs. time curves and (d)
heating rate vs temperature.
Fig. 9b and 9c show the fitted curves and enable the heating rates to be determined with respects to
temperature (Fig. 9d), which would be used for the temperature control. The 2nd order exponential curve
fit may also be applied to the cooling rates, which was used for the temperature control. Fig. 10
demonstrates the effects of altering the length of the heating zone, L, and the compaction force of the
electrodes. Reducing the distance between the electrodes reduces volume of heated material and the
surface area of the convection and radiation heat loses. Additionally, reducing the heated length linearly
reduces the resistance of the material and increases the current drawn for a constant voltage; from 1.8A
(L – 95mm) to 2.4A (L – 47.5mm) at 5V. This therefore increases the degree of Joule heating. Initial
results (Fig. 10) achieve greater than double the heating rate across the full temperature range at 5V
when the compaction force remains constant and the length halved. Increasing the compaction force of
the electrodes further increases the heating rate by reducing the contact resistance with the electrodes
and therefore increases the degree of Joule heating by increasing the current further to 3.7A (at 5V).
020 40 60 80 100 120 140 160 180 200
Time (s)
0
50
100
150
200
Temperature (°C)
1V
1.5V
2V
2.5V
3V
3.5V
4V
4.5V
5V
0 2 4 6 8 10 12 14 16 18 20
Time (s)
0
50
100
150
200
Temperature (°C)
1V
1.5V
2V
2.5V
3V
3.5V
4V
4.5V
5V
0 2 4 6 8 10 12 14 16 18 20
Time (s)
0
10
20
30
40
50
60
70
80
dT/dt (°C/s)
1V
1.5V
2V
2.5V
3V
3.5V
4V
4.5V
5V
050 100 150 200
Temperature (°C)
0
10
20
30
40
50
60
70
80
dT/dt (°C/s)
1V
1.5V
2V
2.5V
3V
3.5V
4V
4.5V
5V
a)
b)
Figure 10: Comparison between the fit curves for changing length of the heat
Temperature control was applied to the Joule heating test rig. This used greater voltages than what
could be tested under steady state conditions due to the high temperatures and heating rates (up to 30V).
This used a PID controller to determine the ton time for a duty cycle, as shown in Fig. 6. An example of
this is shown in Fig. 11, for a 8ms duty cycle and 30V supply, and demonstrates a changing set
temperature despite a heating rate greater than what could be measured by the 150ms sampling rate
(>1000°C/s).
Figure 11: Demonstration of the temperature control for heating carbon fibre tows by Joule heating
5 CONCLUSIONS
This paper has detailed the development of the ADFP rig capable of using RT data acquisition and
sensor feedback. Initially, this feedback has been used for temperature control during the heating of the
fibre deposition, which has been demonstrated, but current work is investigating using additional data
for positional adjustments and “course-by-course” changes to the toolpath determined by variability of
the previously deposited tows. The design of this rig is capable of depositing at speeds of up to 3m/s,
which requires high heating rates (>4500°C/s) and fast response times (<1ms) which it has been
determined that Joule heating is the most suitable method of heating.
The analysis of the rate model has shown that most informative method of charactering the rate of
an ADFP process is by the volume of material deposited with respects to time, rather than production
rate. This quantifies the machine performance in a non-part specific unit. It has also demonstrated
deposition speed and acceleration are only part of the limitations to an ADFP machine, with interactions
between multiple factors leading to influence in the deposition rate. However, it would also be of benefit
to add a ‘robustness’ factor to this model, as this would enable machine downtime to be added to the
analysis, which would be expected to have a significant impact on the results. Electroimpact propose a
0
50
100
150
200
250
0 5 10 15 20
Temperature, T (°C)
Time (s)
0
10
20
30
40
50
60
70
80
90
050 100 150 200 250
Heating Rate, dT/dt (°C/s)
Temperature, T (°C)
0
50
100
150
200
250
050 100 150 200
Temperature (°C)
Time (s)
suitable metric for quantifying robustness by “mean strips before failure” (MSBF) to provide ratio of
the number of successful tows to failed tows during a manufacturing process [16].
ACKNOWLEDGEMENTS
This work was supported by the Engineering and Physical Sciences Research Council [grant
numbers: EP/I033513/1 and EP/P006701/1], through the EPSRC Centre for Innovative Manufacture in
Composites (CIMComp) and the EPSRC Future Composites Manufacturing Research Hub
REFERENCES
[1] “Airbus Commercial Aircraft delivers record performance”. Accessed on: 17/12/2018; Available
from: https://www.airbus.com/newsroom/press-releases/en/2018/01/airbus-commercial-aircraft-
delivers-record-performance.html, 2018.
[2] D.H.J.A. Lukaszewicz, C. Ward and K.D. Potter, The engineering aspects of automated prepreg
layup: History, present and future, Composites part B, 43, 2012, pp. 997-1009, (doi:
10.1016/j.compositesb.2011.12.003).
[3] G. Gardiner, Automated Preforming Part 6: Coriolis Composites, Article of CompositesWorld 23-
3-2018, available from: https://www.compositesworld.com/blog/post/automated-preforming-
coriolis-composites (on 12-6-2019)
[4] B. Uzman, Automated Manufacturing for Mass Production and Low-Cost Materials, 2015 SPE
Automotive Composites Conference Exhibition, Michigan, USA, September 9-11, 2015
[5] P. Zhang, Z. Zhou, G. Chen and S. Chen, Optimizing the lay-up of composite tapes based on
improved geodesic strategy for automated tape placement, Proceeedings of the Institution of
Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, 232 (22), 2018, pp.
4084-4097, (doi: 10.1177/0954406217750187).
[6] M. Di Francesco, L. Veldenz, G. Dell’Anno and K. Potter. “Heater power control for multi-material,
variable speed Automated Fibre Placement”. Composites Part A: Applied Science and
Manufacturing, Vol. 101, pp 408-421, 2017.
[7] J.P.H. Belnoue, T. Mesogitis, O.J. Nixon-Pearson, J. Kratz, D.S. Ivanov, I.K. Partridge, K.D. Potter
and S.R. Hallett, Understanding and predicting defect formation in automated fibre placement pre-
preg laminates, Composites Part A: Applied Science and Manufacturing, 102, 2017, pp 196-206,
(doi: 10.1016/j.compositesa.2017.08.008).
[8] M.Y.Matveev, P.J. Schubel, A.C. Long and I.A. Jones, Understanding the buckling behaviour of
steered tows in Automated Dry Fibre Placement (ADFP), Composites Part A: Applied Science and
Manufacturing, 90, 2016, pp 451-456, (doi: 10.1016/j.compositesa.2016.08.014).
[9] R.J. Crossley, P.J. Schubel and N.A. Warrior, The experimental determination of prepreg tack and
dynamic stiffness, Composites Part A: Applied Science and Manufacturing, 43, 2012, pp 423-434,
(doi: 10.1016/j.compositesa.2011.10.014).
[10]R.J. Crossley, P.J. Schubel and D.S.A De Focatiis, Time–temperature equivalence in the tack and
dynamic stiffness of polymer prepreg and its application to automated composites manufacturing,
Composites Part A: Applied Science and Manufacturing, 52, 2013, pp 126-133, (doi:
10.1016/j.compositesa.2013.05.002).
[11]I.D.J.G Ojeda, O. Patrouix and Y. Aoustin, Pressure based approach for Automated Fiber Placement
(AFP) with sensor based feedback loop and flexible component in the effector, IFAC-
PapersOnLine, 50, 2017, pp. 794-799, (doi: 10.1016/j.ifacol.2017.08.511).
[12]B. Denkena, C. Schmidt and P. Weber. “Automated Fiber Placement Head for Manufacturing of
Innovative Aerospace Stiffening Structures”. Procedia Manufacturing, Vol. 6, pp 96-104, 2016.
[13]Heraeus, Intelligent heat for Automated Fibre Placement, available from:
https://www.heraeus.com/media/media/hng/doc_hng/products_and_solutions_1/arc_and_flash_la
mps_1/humm3.pdf (on 20-6-2019)
[14]F Helber, A Amann, S Carosella and P Middendorf, Intrinsic fibre heating: a novel approach for
automated dry fibre placement, 13th International Conference on Textile Composites (TEXCOMP-
13), IOP Conf. Series: Materials Science and Engineering, 406, 2018, pp. 1-10, (doi: 10.1088/1757-
899X/406/1/012064).
[15]Y. Grohmann and J. Stueve, CoRe HeaT - Continuous Resistance Heating Technology for fast and
efficient high volume production processes, 28th SICOMP Conference on Manufacturing and
Design of Composites, Pitea, Sweden, June 1st-2nd, 2017
[16]O. Lu, Servo Creel Development, Electroimpact, 2019, available from:
https://www.electroimpact.com/WhitePapers/2019-01-1348.pdf (on 21-6-2019)
