archives of civil and mechanical engineering 15 (2015) 631–638

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: http://www.elsevier.com/locate/acme

Original Research Article

Inspection method of aluminium extrusion process

P. Garbacz, T. Giesko *, A. Mazurkiewicz

Institute for Sustainable Technologies – National Research Institute, ul. Pulaskiego 6/10, 26-600 Radom, Poland

article info abstract

Article history: The paper presents the method for the monitoring of aluminium extrusion processes. The
Received 25 November 2014 developed hybrid method combines the advantages of computer based, simultaneous
Accepted 4 February 2015 infrared and visible image analysis for surface inspection of the profile directly after leaving
Available online 26 February 2015 the die. Thermograms present the temperature distribution on the surface of the extruded
profile and contain information about the extrusion process. The proposed inspection
Keywords: system can be applied in industry for on-line monitoring of aluminium extrusion processes
Aluminium extrusion and the inspection of defects arising in extruded products.
Surface inspection # 2015 Politechnika Wrocławska. Published by Elsevier Sp. z o.o. All rights reserved.
Multimodal analysis
Thermography

the aspect of quality of products, due to different use of

1. Introduction
A hot aluminium extrusion is a plastic deformation process in
which a preheated aluminium billet, placed in a heated
container, is forced to flow by compression through the die
opening to obtain the desired shape. Compression is caused by
a moving ram pushing the billet. Typically, the hot aluminium
extrusion process is carried out at temperature between 450 8C
and 580 8C, and at a profile exit speed between 5 m/min and
100 m/min. Parameters of extrusion process are dependent on
the shape and size of the profile and the alloy used to produce
the profile [1]. After leaving the die, the extruded profile is
cooled using an air or water quench. The average cooling rate
is determined by the exit speed, the quench temperature, and
the shape of the profile and, in general, its value is lower than
20 8C/s. One of the main issues in the production of aluminium
profiles with use of hot hydrostatic extrusion technology is
constant monitoring process parameters and an on-line
inspection of products just after they leave the extruder. In
aluminium profiles as construction elements that often
require high aesthetics, it is essential to control the state of
their surface. The quality of the produced aluminium directly
depends on the basic parameters of the extrusion process such
as strength and extrusion speed, temperature in the working
area, and other factors like the type of material, the size of the
billet, the billet preparation, profile shape, matrix design, wear
of the die, and the phenomenon of occurring in the process of
friction and lubrication [1]. Some aspects of the extrusion
process and forging process, like tool temperature, press
settings, process speed, lubrication, and tool wear condition
are very similar [2]. The abrasive wear and thermo-mechanical
cracking are dominant mechanisms in the degradation
process of the dies [3]. High-tech, highly automated production
lines are equipped with control systems to ensure the
monitoring of the desired parameters of extrusion process.
Measurements of the surface temperature of the finished
product are usually performed by non-contact infrared
pyrometers [4,5]. In paper [6], the method for local temperature

* Corresponding author. Tel.: +48 48 3644241; fax: +48 48 3644760.

E-mail address: tomasz.giesko@itee.radom.pl (T. Giesko).
http://dx.doi.org/10.1016/j.acme.2015.02.005
1644-9665/# 2015 Politechnika Wrocławska. Published by Elsevier Sp. z o.o. All rights reserved.
632 archives of civil and mechanical engineering 15 (2015) 631–638

measurement with use of a long-wave infrared sensor was

presented. Temperature control in the areas of the container
and die is carried out by contact methods using thermocouples
[7] and non-contact methods where infrared methods pyrom-
eters are applied [8]. The temperature value is used for
optimization in the feedback control system of the press
extrusion process [1,9]. A limitation of the method of
temperature control by means of a pyrometer is a local
measurement, performed in a fixed location on the surface of
the product. Local measurement method does not allow for
tracking of the temperature distribution on the surface of the
selected sections in a direction transverse to the direction of
movement and the detection of anomalies that may indicate a Fig. 1 – Surface defects in the form of bright streaks (pointed
disturbance in the extrusion process. In current practice, by arrows) [12].
regardless of the automation process control, the extrusion
press operator verifies the basic parameters of product.
Manual control includes a visual assessment of the quality
of the manufactured profiles, in particular, the surface The study indicates that the cause of defects is increased
condition and measurements of the main dimensions of the friction between the surface of the die and material [14]. The
cross-sectional geometry are performed. The main quality result of the uneven wear of the working surface of the matrix
testing of products takes place at a separate station where the may be uncontrolled local increase of temperature and
defective sections of the profiles are identified and rejected. thermal energy in the top layer of the surface of extruded
Analysis of scientific publications and technical studies on material. This phenomenon is associated with dynamic
the problem of quality assurance in the process of hot uneven recrystallization that causes inhomogeneous distri-
hydrostatic extrusion confirmed the lack of advanced solu- bution of grain boundary grooves and the formation of regions
tions with the use of combined infrared and visible light with large grain microstructure shapes elongated in the
techniques for surface inspection of manufactured products. extrusion direction [12]. As a result, anodized or painted
The development progress of methods and systems for on-line profiles in the areas of the change in the microstructure have
monitoring processes of hot hydrostatic extrusion and quality different reflectivity and colour. The key effect of the heat
assurance of manufactured products is, therefore, a current generated in the material that affects the growth of the grains
and important research topic with a significant economic of the microstructure was identified in the works [13,15]. The
importance for aluminium profiles manufacturing companies. theoretical work with use of computer simulation methods
have confirmed that changes in the coefficient of friction
between the die and the material of the billet have a
2. Product defects in aluminium extrusion
substantial effect on the variations in the microstructure of
the material surface layer [16].
A review of publications on the subject of hot hydrostatic Die lines are one of the often-described defects in the
extrusion concludes that the classification of defects of aluminium extrusion process is a die line. It is defined as
aluminium profiles in a manufacturing process have not been longitudinal depressions or protrusions formed on the surface
developed or organized. Therefore, there is flexibility in the of the extruded material due to imperfections on the die
naming of identified defects of the products that hinder proper surface (Fig. 2). However, in practice, there are cases in which
communication between specialists in the same research area. the high-quality surface of the die and the optimal parameters
The main difficulty in comparative studies of extrusion defects of the extrusion process still did not guarantee avoiding this
is the lack of quantifiable data, mainly by giving shapes and defect.
sizes of geometric representations of defects and histograms of Weld defects are defects resulting from the nature of the
the captured images with reference to the manufacturing extrusion process, resulting from the merger of the faces of
process parameters. An analysis of selected scientific publica- two consecutive extruded profiles (Fig. 3). The formation of
tions relating to various types of defects occurring in the process defects contributes to a large oxide layer located on the
of hot hydrostatic extrusion is presented in [10]. The results surfaces of faces and contamination by the lubricant.
indicate a lack of a systemic defect identification method in
current engineering practice. Based on the analysis of available
publications [1,10–13] and the authors' research carried out in a
light alloys manufacturing company, the following major
defects can be identified that occur most often in practice.
Streak defects are defects that manifest themselves only
after anodizing or painting in the form of elongated streaks
that differ from the surrounding surface brightness and colour
shade (Fig. 1). The lack of the possibility of direct visual
detection of these defects on the processing line is a major
problem in quality control. Fig. 2 – Die lines defect [10].
 archives of civil and mechanical engineering 15 (2015) 631–638
Fig. 3 – Weld defect [10].
Fig. 4 – Transverse defect lines on the surface of the
aluminium profile.
Fig. 5 – Scratch defect.
Transverse lines are defects in the form of a transverse line of
a different colour and microstructure, resulting from a
temporary stopping the plastic flow of material through a
die (Fig. 4).
Scratches – surface defects in the form of scratches caused
by hard inclusions (impurities) deposited on the working
surface of the die, which result in the damage of the surface
layer or local lowering of the hardness of the coating (Fig. 5).
Fig. 6 – Blister defect [17].
633
Blisters – surface defects in the form of bubbles (Fig. 6). The
mechanism of the defect is not fully recognized. It is generally
accepted that the cause of the defects are small volumes of air
or lubricant vapours trapped in the material in the extrusion
process.
Pick-up – surface defects in the form of longitudinal recesses
intermittent lengths from 3 mm to 12 mm (Fig. 7). Since the
cause of their formation is the given inclusions in the billet
material, the improperly conducted process of homogeniza-
tion of the billet material and geometrical distortion of die are
imputed.
Cracking, tearing – a series of surface cracks perpendicular to
the extruding direction caused by the high extrusion speed or
extrusion temperature (Fig. 8).
From the analysis of the available quality control statistics
of profiles manufactured from aluminium alloys, the majority
of the defects are streaks, transverse lines, and blisters.
Currently in the industry, the streak defect detection proce-
dure is carried out by painting or anodizing previously cut out
parts of the profile and manual visual control by an
experienced technologist. Outcome of the assessment is the
basis for grading a batch of products.
3. The monitoring method of aluminium
extrusion process
The developed original hybrid method for monitoring the
aluminium hot extrusion process involves the use of a
multimodal imaging and combined analysis of images of
the aluminium profile surface captured by the infrared camera
and visible light camera [18]. The multimodal imaging is a
technique that merges the complementary and redundant
information from sensors of different wavelengths, simulta-
neously viewing the same scene, in order to get one composite
image. The fused image is formed to improve image content
and to make it easier for the user to detect, recognize, and
identify targets. The capabilities of applications and advan-
tages of the use of methods for monitoring processes using
observations of the object in the infrared and visible light are
presented in [19,20].
The structure of the system is shown in Fig. 9. The system
dedicated for continuous on-line monitoring is based on a
feedback loop, which results in a message about the probable
irregularities in the process that generated based on the
analysis of detected defects of the products. The main module
of the monitoring system is the camera head equipped with
infrared (IR) and visible light (VIS) cameras.
For observation of the surface of aluminium at high
temperatures up to approx. 600 8C, a shortwave infrared
sensor camera of the spectrum of 0.9–1.7 mm was selected. The
634 archives of civil and mechanical engineering 15 (2015) 631–638
Fig. 7 – Surface defect in the form of pick-up [10].
Fig. 8 – Crack defect [10].
proposed solution provides the most favourable conditions of
temperature measurement due to the characteristics of the
emissivity [21]. The applied IR camera sensor has a resolution
of 320  265 pixels. The accuracy of temperature measure-
ment by the camera is 1 8C. The infrared camera is designed
for recording fast processes with a maximum frequency of
340 frames/s. The applied camera for observations in the
visible light has a high resolution of 2456  2058 pixels, which
allows precise inspection of the surface of products. The
Fig. 10 – Experimental stand.
Fig. 9 – Monitoring system scheme.
optical axes of cameras were adjusted to ensure that captured
images contain the same area of the surface being inspected.
The measurement resolution of images depends on the size of
the inspection area determined on the surface of a profile. For
example, for the inspection area of 144 mm  189 mm, and the
observation distance of 1.1 m, the measurement resolution for
the IR camera applied is app. 0.56 mm, and for the VIS camera,
app. 0.073 mm. The process of image analysis and defects
detection is carried out using dedicated algorithms, and it is
less than 50 ms. While calculating exposure times of both
cameras, the image blur effect was considered [18]. An
important factor in the inspection process carried out by the
VIS camera is to ensure proper lightning conditions in the
observation area. Halogen illuminator and pulse LED illumi-
nator sources were selected for implementation. The inspec-
tion process takes place in on-line mode in the course of the
movement profile. Developed software implements a com-
bined image analysis and the identification of possible defects.
4. Experimental studies
The experimental studies were carried out on a laboratory
stand equipped with a heating chamber to allow heating of
selected aluminium profiles (Fig. 10). The observed aluminium
 archives of civil and mechanical engineering 15 (2015) 631–638
Fig. 11 – Diagram of combined images analysis.
profile is moved on the linear guide simulating the movement
of the extruded product when leaving the press.
Diagram of the combined analysis of IR and visible light
images is shown in Fig. 11. The IR image is overlaid onto the
visible image by use of homographic transformation. To
calculate a homographic matrix, a previous calibration process
is required which involves identification of minimum four
corresponding pairs, of points from the IR image and visible
image. Once the homographic matrix is calculated, the IR
image is projected onto the visible image. The fusion of images
helps to identify the location of individual defects occurring on
the surface of the inspected object.
The implemented algorithms in the developed software
enable the detection of the most common defects in the
surface of aluminium profiles with use of both visible light and
the infrared cameras. These defects can be classified into three
groups due to their specificity: longitudinal, transverse, and
local defects. In order to identify submitted defects, two
categories of algorithms for the detection of edges and the
blobs (binary large objects) are used.
Edge detection algorithms are used for monochrome image
segmentation, the aim of which is the location of pixels having
a high gradient of the brightness. Another important issue is to
determine the appropriate level of threshold gradient that will
ensure the desired edge detection. This is particularly
important for the analysis of thermal images of objects for
which the surface temperature changes dynamically.
Blob detection algorithms are used for detecting the binary
image areas, consisting of interconnected pixels that are of
Fig. 13 – Acquired thermogram: (a) thermogram of aluminium profile surface and (b) temperature profile in the section 1.
635
Fig. 12 – Algorithm for detection of longitudinal and
transverse defects.
equal logical value. The decision about whether an image
element belongs to a given area is determined based on the
criterion of the connection [22]. In order to ensure the
detection of the desired object in the analysed image, a
binarization is performed with a suitably selected threshold
before using this method.
The algorithm for the classification of defects occurring
longitudinally and transversely on the analysed profile surface
is shown in Fig. 12.
After the acquisition of images from the infrared camera
and visible light cameras, they are processed for edge
detection in vertical and horizontal direction. The thermo-
gram of the profile surface is shown in Fig. 13a. The procedure
for the detection of lines with the desired width is applied
based on the approach to regard lines as objects having
parallel edges [23]. First, the differentiation is performed for
the detection of edges of the desired line. The areas of the
analysed image with the greatest changes in the grey level
have the largest absolute differential value (Fig. 13b). Two
maximum values are searched after the threshold operation is
applied. The detected pair of edges must also fulfil a condition
of gap distance to be identified as the sought defect type.
636 archives of civil and mechanical engineering 15 (2015) 631–638
Fig. 14 – Image of aluminium profile surface with local
defects (pointed by arrows).
It should be noted that, in the case of the thermal camera,
image analysis is performed not by the value expressed in
Celsius degrees but with the use of a 14-bit value that
corresponds to the measuring resolution of the sensor. For
visible light camera brightness, the level of each pixel is
determined while using the 12-bit value.
In order to identify defects in the form of streaks occurring
after the anodizing process, combined analysis of thermal and
visible light camera images is required. In the applied
procedure, images from both thermal and VIS cameras are
analysed using a previously described algorithm for longitu-
dinal defect. However, the proposed algorithm also identifies a
situation where there are no longitudinal defects detected in
the same location on the profile surface, and there is a
significant increase in the temperature gradient in the vertical
direction, which suggests the occurrence of defects in the form
of a die streak. An important element of the streak detection
function is to guarantee an exact match to each image areas
with a thermal imaging camera and visible light. For this
purpose, the operation of homography [24] is used which
Fig. 15 – Visualization of experimental result: (a) 3D visualization of temperature distribution on the surface and (b) thermal
image after the process of binarization.
provides the desired image registration. Homographic trans-
formation requires a pre-calibration process with use of a
pattern whose elements are especially pronounced in the
images from both thermal and infrared cameras [25].
The second group of defects for which detection methods
have been developed are blisters and pick-ups. In order to
identify them, a blob detection algorithm has been applied. An
example of the image captured by the visible light camera
where there are areas with identified local defects is shown in
Fig. 14.
In the first stage of the developed procedure, a binarization
function with two thresholds was used to define the upper and
lower limit of the range of acceptable values. On the image
captured by infrared camera, these levels are used to detect
areas of non-uniformity of the temperature profile on the
camera. In case of the visible light camera, the detection of
areas with different levels of brightness in relation to the areas
corresponding to the image of proper surface profile is
implemented. The 3D visualization of the temperature
distribution on the surface of the profile is shown in
Fig. 15a. Each temperature deviation from the assumed correct
level is marked by the white colour on the image after the
binarization process (Fig. 15b). Areas corresponding to the
correct temperature are marked in black.
To increase the stability of the method by reducing the
influence of disruption, the combination of morphological
operations was used for image processing. First, an erosion
operation is performed, and then a dilation process is applied.
Then, on the processed images, areas with suitable geometries
are searched that are then identified as local defects in the
form of blisters or pick-ups (Fig. 16).
An important element of the proposed method of inspec-
tion is the module to prevent false positive detection of
defects. This is particularly important in the case of longitu-
dinal defects on the surface and profiles of atypical cross-
sections (Fig. 17), for which both the image captured by
infrared and the visible light camera edges resulting from the
structure of the profile will be observed.
In order to avoid erroneous detection of defects, it is
necessary to develop a software system to support a final
 archives of civil and mechanical engineering 15 (2015) 631–638
Fig. 16 – The identified areas of potential local surface
defects of aluminium profiles.
Fig. 17 – Sample shapes of aluminium profiles [26].
decision inspection. The schematic structure of the system is
shown in Fig. 18.
All the defects identified for profiles should be collected and
categorized in the manufacturer database. Every new defect
detected and described may be put into the database of
defects. Algorithms of image analysis suitable for a given
product are used to detect and recognize potential defects.
Fig. 18 – Schematic structure of the software system.
637
5. Summary and conclusions
The results of experimental studies have confirmed the
possibility of using a hybrid method of inspection with use
of infrared and visible-light cameras for the monitoring of the
hot hydrostatic extrusion process. The applied thermal
imaging camera, which is characterized by high sensitivity
and speed measurement registration, allows the observation
of thermodynamic effects on the surface of aluminium
extrusion. The visible light camera with a high-resolution
sensor allows the detection and identification of surface
defects with a high accuracy. The conducted experiments
confirmed that the combined hybrid image analysis allows for
increasing the reliability of the detection of surface defects of
aluminium profiles in the case of interference measurement in
one of the optical paths. Multimodal analysis also allows one
to identify areas of potential defects revealed after the process
of anodizing profiles.
Further work will be carried out to optimize the algorithms
for image processing and analysis and to verify them directly
on the line in industrial conditions. The next stage of the
project is to integrate the developed system, the control
system, and quality control in a selected manufacturing plant.
A prerequisite for this is to implement a function that uses a
database of geometric parameters of produced profiles in order
to avoid false positive detection of defects. The hybrid
monitoring method proposed in the article can be easily
adopted to other hot metal forming processes.
references
[1] K. Saha, Aluminum Extrusion Technology, ASM International,
2000.
[2] Z. Gronostajski, M. Hawryluk, The main aspects of precision
forging, Archives of Civil and Mechanical Engineering 8 (2)
(2008) 39–55.
[3] Z. Gronostajski, M. Kaszuba, M. Hawryluk, M. Zwierzchowski,
A review of the degradation mechanisms of the hot forging
tools, Archives of Civil and Mechanical Engineering 14 (4)
(2014) 528–539.
[4] KELLER MSR. http://www.keller-msr.com/infrared-temperature-
measurement/, 2014.
[5] Land Instruments International. http://www.landinst.com/
industries/aluminium-extrusion/, 2014.
[6] T.S. Wiśniewski, W. Pachla, D. Kukla, A. Mazur, K.J.
Kurzydłowski, Application of infrared thermography in
investigation of hydrostatic extrusion, in: Proceedings of the
7th International Conference on Quantitative InfraRed
Thermography, 2004, pp. F.8.1–F.8.6.
[7] M. Terčelj, R. Turk, G. Kugler, P. Fajfar, P. Cvahte, Measured
temperatures on die bearing surface in aluminium hot
extrusion, RMZ – Materials and Geoenvironment 53 (2) (2006)
163–173.
[8] P. Wright, S. Metcalf, Recent innovations and improvements
in infrared thermometry within the aluminum extrusion
industry, in: Proceedings of 8th International Symposium
‘‘Temperature, Its Measurement and Control in Science and
Industry’’, vol. 7, 2002, pp. 801–806.
[9] T. Sheppard, Extrusion of Aluminium Alloys, Kluwer
Academic Publishers, 1999.
638 archives of civil and mechanical engineering 15 (2015) 631–638
[10] A.F.M. Arif, A.K. Sheikh, S.Z. Qamar, M.K. Raza, K.M. Al-
Fuhaid, Product Defects in Aluminum Extrusion and Their
Impact on Operational Cost, KFUPM, Dhahran, 2002137–154.
[11] H. Zhu, M. Couper, A. Dahle, Etching Effects and the
Formation of Streaking Defects on Al Extrusions, 2011
http://www.aluminiumtoday.com.
[12] H. Zhu, X. Zhang, M. Couper, A. Dahle, Effect of initial
microstructure on surface appearance of anodized aluminum
extrusions, Metallurgical and Materials Transactions 40A
(2009) 3264–3275.
[13] Z. Peng, T. Sheppard, Study of surface cracking during
extrusion of aluminium alloy AA 201, Materials Science and
Technology 20 (2004) 1179–1191.
[14] I. Flitta, T. Sheppard, Nature of friction in extrusion process
and its effect on material flow, Materials Science and
Technology 19 (2003) 837–846.
[15] M. Lewandowska, Wytwarzanie materiałów nanokrystalicznych
metodą wyciskania hydrostatycznego, Obróbka Plastyczna
Metali XVII (4) (2006) 9–13.
[16] P. Menezes, K. Kumar, K. Kishore, S. Kailas, Influence of
friction during forming processes – a study using a numerical
simulation technique, International Journal of Advance
Manufacturing Technologies 40 (2009) 1067–1076.
[17] R. Peris, Z. Chen, T. Pasang, Effects of Extrusion Process
Conditions on ‘‘Die Pick-up’’, VDM Verlag Dr. Mueller e.K.,
2008.
[18] T. Giesko, P. Garbacz, Application of hybrid vision method for
the hot aluminium surface inspection, Solid State Phenomena
19 (2013) 267–272.
[19] A. Christie, What's Inside: Thermal Process Imaging Permits
a Real Time View of Your Extrusion Process, 2012 http://
www.optexprocesssolutions.com/news/white-papers.shtml.
[20] E. Whitenton, High-Speed Dual-Spectrum Imaging for the
Measurement of Metal Cutting Temperatures, National
Institute of Standards and Technology, 2010.
[21] K. Chrzanowski, Non-contact Thermometry. Measurement
Errors, SPIE Polish Chapter, 2001.
[22] T.B. Moeslund, Introduction to Video and Image Processing,
Building Real Systems and Applications, Springer-Verlag,
London, 2012.
[23] T.M. Koller, G. Gerig, G. Szekely, D. Dettwiler, Multiscale
Detection of Curvilinear Structures in 2-D and 3-D Image
Data, IEEE Computer Society, Washington, 1995864–869.
[24] E. Vincent, R. Laganiere, Detecting planar homographies in
an image pair, in: ISPA, 2001, 182–187.
[25] L. St-Laurent, D. Prévost, X. Maldague, Fast and accurate
calibration-based thermal/colour sensors registration, in:
Proceedings of the 10th International Conference on
Quantitative InfraRed Thermography, 2010, pp. 127–134.
[26] Cardinal Aluminum Company. https://cardinalaluminum.
com/, 2014.