Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 124 (2024) 530–535

13th CIRP Conference on Photonic Technologies [LANE 2024], 15-19 September 2024, Fürth, Germany

Influence of full penetration welding on the multispectral signal of the

process radiation during laser beam welding of austenitic stainless steel
Patrick Meyera,*, Christian Schumannb,Dario Spatzeka, André Häuslera
a
Fraunhofer-Institut für Lasertechnik ILT, Steinbachstraße 15, 52074 Aachen, Deutschland
b
4D Photonics GmbH, Burgwedeler Straße 27A, 30916 Isernhagen, Deutschland

* Corresponding author. Tel.: +49 241 8906-135; E-mail address: patrick.meyer@ilt.fraunhofer.de

Abstract

A 100% visual inspection of the welds carried out is not always possible for economic reasons. Different methods of process monitoring are therefore used for non-
destructive testing of process and weld seam quality. In the following, the recorded multispectral signals of the reflective process radiation are evaluated using
descriptive statistics and characteristics of a full penetration weld during laser beam welding of austenitic stainless steel are worked out. The aim is to detect and
localize unintentional full penetration welding at an early stage and thus avoid rejects. For this purpose, adapted sample geometries with varying material thicknesses
are prepared for the test series. As a result, the need to vary the welding process parameters to achieve the different welding states (full and partial penetration weld)
is avoided. By evaluating the measured values of the wavelength-selective signals, a significant influence on the signal intensity and the signal frequency could be
confirmed.
© 2024 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the international review committee of the 13th CIRP Conference on Photonic Technologies [LANE 2024]

Keywords: Laser beam welding; process monitoring; multispectral sensor; stainless steel; full penetration welding

1. Introduction spectrum of the metal vapor plasma. The characteristic vapor

capillary that forms at high intensity during deep penetration
The laser welding process is almost completely automated. welding is called a keyhole [3,4,5]. These radiation processes
External disturbance variables such as fluctuating relative focal are defined in their entirety as process radiation. The individual
positions or contaminated material surfaces have a components of the process radiation contain information; a
considerable influence on the quality of the process and weld targeted wavelength-selective analysis enables conclusions to
seam [1]. Various monitoring systems are used for the early be drawn about the quality of the process and weld seam. The
detection and localization of relevant quality deviations. One spectral radiation intensity can be used for process monitoring
option for process monitoring is the use of multispectral sensor [2,5]. Examples of applications include the non-destructive
[2]. Laser beam welding is based on the interaction between the testing of weld seams to localize weld seam imperfections and
propagating laser beam and the material to be welded. As the determination of local mixing during dissimilar laser beam
result, radiation is emitted and reflected from the area of the welding of copper and aluminum sheets [6]. The keyhole
interaction zone of the joining process into the surrounding during deep penetration welding is highly dynamic [7]. If a
atmosphere. The wavelength range of the thermal radiation penetration weld is achieved, the metal vapor can escape from
emitted during laser beam welding with solid-state lasers the sample not only from above but also from below and the
extends from ultraviolet to the near-infrared range [3]. process becomes more stable [8]. It is assumed that the change
Depending on the intensity of the applied laser radiation, a in process stability has an influence on the process radiation
metal vapor plasma forms in addition to the thermal radiation. emitted. In this paper, weld penetration is examined as a
The characteristic line radiation can be identified in the relevant influencing factor of process quality and

2212-8271 © 2024 The Authors. Published by Elsevier B.V.

This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the international review committee of the 13th CIRP Conference on Photonic Technologies [LANE 2024]
10.1016/j.procir.2024.08.168
 Patrick Meyer et al. / Procedia CIRP 124 (2024) 530–535 531

characteristics of a weld penetration on the multispectral signal 2.2. Material and specimen preparation
are determined.
The austenitic stainless steel 1.4301 (X5CrNi18-10) is used for
Nomenclature the trials. The composition of the alloy is shown in Table 1.

xi Measured value Table 1. Chemical Composition: 1.4301, according to thyssenkrupp AG

fC/L Focal length (collimator c and lens L) C Cr Ni Si Mn P S
zi Centered measured value (min.)
A Matrix - Partial penetration weld 0,07 17,5- 8,0- 1,00 2,00 0,045 0,015
B Matrix - Full penetration weld 19,5 10,5
IQR Interquartile range Percentages by mass m%, max.
LBW Laser beam welding
NIR Near-Infrared Light For the trials, specimens with a stepped geometry and varying
PL Laser beam power material thicknesses are produced by machining. The stepped
VP Welding speed geometry makes it possible to achieve different welds (full
VIS Visible light penetration weld and partial penetration weld) with constant
ZCR Zero crossing rate welding parameters. The variants of the specimen geometries
ØSpot Spotdiameter are shown schematically in Figure 2. The figure is not to scale
and is for illustrative purposes only.
2. Methods

2.1. Experimental setup

The TruDisk12002 multimode disk laser from TRUMPF SE +

Co KG with an emitting wavelength of λ = 1030 nm and a
maximum output power of 12 kW is used for the experiments.
The focal length is fC/fL = 158 mm/400 mm for a spot diameter
of ØSpot = 0.502 mm and fC/fL = 100 mm/163 mm for ØSpot =
0.168 mm. The focus diameter is determined using a caustic
analysis (MicroSpotMonitor from PRIMES GmbH). Argon
Ar4.8 with a volume flow of 𝑽𝑽� = 15 l/min is used as the
shielding gas for the welding process. The shielding gas is
supplied to the interaction zone via a copper tube with an
internal diameter of Ø8 mm. The 4D.TWO multispectral
sensor from 4D Photonics GmbH is used for the trials and
installed coaxially on the laser optics. With the multispectral
sensor are used, a total of 32 channels to record the
wavelengths. 16 channels for the visible spectrum VIS from
384 nm to 883 nm with an equidistant step width of 33 nm. An Fig. 2. Specimens in half section: (a) step geometry; (b) multi-step geometry
additional 16 channels for the near-infrared spectrum NIR from
1100 nm to 1850 nm with an equidistant step width of 50 nm.
To illustrate this, the multispectral signal of an example weld In addition, specimens with a constant material thickness of
(partial penetration weld) is shown in Figure 1. t1 = 1 mm, t2 = 2 mm, t3 = 4 mm, and t4 = 6 mm are used. At
the beginning of the trials, the surface of the specimens is
cleaned with ethanol C2H5OH.

2.3. Welding parameters

A welding speed of vP = 1,5 m/min, a laser beam power of PL

= 1,5 kW and a spot diameter of ØSpot = 0,168 mm are used for
the trials with the step and multi-step specimens. The welding
depth is approx. 4.5 mm. The welding parameters are adjusted
for the trial with constant material thickness. A welding speed
of vP = 7 m/min and a spot diameter of ØSpot = 0.506 mm are

Fig. 1. Multispectral measured values - raw data

532 Patrick Meyer et al. / Procedia CIRP 124 (2024) 530–535

used. The laser beam power PL is increased from 3,2 kW to 3 and 4) [9]. The variable I defines the total number of centered
4,0 kW with an equidistant step width of 0,2 kW. measured values zi per interval. The function sgn (abbreviation
for sign change) is used to differentiate between the cases.
2.4. Determination of a relative key figure

The wavelength-selective analysis is carried out for discrete- �

𝑍𝑍CR � ∑����|𝑠𝑠𝑠𝑠𝑠𝑠�𝑧𝑧� �𝑛𝑛�� � 𝑠𝑠𝑠𝑠𝑠𝑠�𝑧𝑧� �𝑖𝑖 � 1��| (3)
time stochastic signals. Stochastic signals cannot be fully �
described using mathematical methods. For this reason,
methods of descriptive statistics and frequency analysis are
used. The comparison of wavelengths is carried out by 1, 𝑧𝑧� �𝑖𝑖 � � 0,
sgn�z� �𝑖𝑖 �� � � (4)
calculating matrices. �1, 𝑧𝑧� �𝑖𝑖 � � 0.

Quadratic matrices A (partial penetration weld) and B (full

penetration weld) are formed to determine a wavelength The zero-crossing rate ZCR indicates how often the sign of
intensity quotient that is sensitive to full penetration weld. The successive (centered) measured values changes from positive
matrices consist of n = 32 columns and m = 32 rows to negative or from negative to positive in a defined interval.
corresponding to the number of wavelengths detected (compare
equation 1). The row and column indices m and n stand for the 3. Results and discussion
recorded channels from 1 ≙ 384 nm to 32 ≙ 1850 nm.
Based on the calculated state matrices A and B, a maximum
percentage deviation of 54.20% is determined for the
𝑎𝑎1,1 ⋯ 𝑎𝑎1,32 wavelength intensity quotient of 1500 nm (NIR) and 384 nm
A32,32 � � ⋮ ⋱ ⋮ � (1) (VIS) when comparing the states full penetration weld and
𝑎𝑎32,1 ⋯ 𝑎𝑎32,32 partial penetration weld. The percentage deviation along the
main diagonal of the difference matrix corresponds to 0 %. The
For the matrix elements a of the matrix A and the matrix result can be explained mathematically by subtracting two
elements b of the matrix B, the quotients are calculated from quotients with the value 1, which result from the division of
the respective wavelength combinations and specified as a equal measured values due to equal wavelengths. All 32
median. With this method, all 1024 possibilities of quotient wavelengths considered are similar in terms of the curve
formation are considered. The difference matrix from A and B progression depending on the process feed rate and the weld
is specified as a percentage deviation. seam length and differ in terms of the intensities. For all
wavelengths, a drop in signal intensity can be observed in
2.5. Zero crossing rate ZCR compare of the full penetration welding. The resulting
difference matrix is shown in Figure 3 as a contour map.
The aim is to quantify the dynamic behavior of the signal Deviation [%]
intensity curve as a function of the welding condition (full
penetration and partial penetration). The multispectral data set 0,00 6,78 13,55 20,33 27,10 33,88 40,65 47,43 54,20

of a weld seam consists of 32 channels with a measured value

acquisition of 5.000 data per second and per channel. The data 1750

sets are intervalized for the analysis. Each interval consists of

a number of measured values xi per channel, where i defines the 1500
exact measured value in the interval as a running variable. The
recorded measured values xi of the raw signal are centered on 1250
Wavelength [nm]

the calculated arithmetic mean value 𝒙𝒙 � of the respective

interval. For centering, the difference between the measured
850
values (minuend) and the arithmetic mean of the interval
(subtrahend) is formed within the defined intervals. The
procedure for centering the measured values xi is shown in 683

equation 2. Where zi is the centered measured value.

516

𝑧𝑧� � 𝑥𝑥� � 𝑥𝑥̅ (2) 500 750 1000 1250 1500 1750
Wavelength [nm]

Fig. 3. Comparison of full penetration weld (material

To determine the zero-crossing rate ZCR of the interval, the
thickness t = 2 mm) and partial penetration weld (material
quotient of the number of sign changes and the number of thickness t = 6 mm)
measured values is formed for each interval (compare equation
 Patrick Meyer et al. / Procedia CIRP 124 (2024) 530–535 533

The percentage deviation of the signal intensity differs

significantly from the percentage deviation determined in the
Based on this result, the ratio of 1500 nm and 384 nm is matrix calculation. The difference matrix is calculated based on
considered as a relative parameter for the next trials. It is measured values generated in trials with specimens of constant
assumed that the influence of full penetration welding is material thickness. It is assumed that the ratio of welding depth
greatest for this parameter (for the specified conditions). to material thickness (for a full penetration weld) has an
The progression of the signal intensity and the zero-crossing influence on the multispectral signal. In the next step, the
rate of the quotient of 1500 nm (dividend) and 384 nm (divisor) influence of full penetration welding on the multispectral signal
is shown as a function of the weld length in Figure 4. The length is investigated as a function of the ratio of weld penetration
of the weld seam is 70 mm. For the analysis, the data set of the depth and the material thickness. All specimens with a material
entire weld seam length is divided into 70 intervals. Each thickness of t1 = 1 mm are full penetration welded. All
interval contains 200 measured values. The step transition and specimens with a material thickness of t3 = 4 mm are partially
thus the start of the full penetration weld is marked as a dashed penetration welded. For specimens with a material thickness of
line at 37 mm. The intensity and frequency (number of zero t2 = 2 mm, full penetration welding is achieved from laser beam
power of 3,6 kW. The length of the weld seam is 60 mm. The
crossings per interval) of the intensity quotient decreases with
data set of each weld seam is divided into 5 intervals. Each
the start of the full penetration welding. The value of the
interval contains 515 measured values. The centered intensity
interquartile range as a measure of the statistical scatter also
quotient of 1500 nm and 384 nm is evaluated according to
decreases. equation 3 and 4. The results of the zero-crossing rate are
shown in Figure 5. Filled symbols represent full penetration
weld seams. Symbols without filling represent partial
penetration weld seams. The connecting lines only serve to
illustrate related test series with constant material thickness and
are not to be understood as a linear interpolation of the
intermediate values.

Fig. 5. Influence of the weld penetration on the signal frequency

Considering the statistical scattering, a comparatively higher

zero-crossing rate (at constant laser beam power) was
determined for the full penetration weld seam specimens
examined. The value of the zero-crossing rate increases with
Fig. 4. Analysis: Specimen (a) step geometry increasing laser beam power for specimens with a constant
material thickness of t2 = 2 mm and t3 = 4 mm. If the laser beam
power is increased from 3,4 kW to 3,6 kW for the specimens
The signal intensity of the quotient does not decrease abruptly with a material thickness of t2 = 2 mm, a full penetration weld
at the beginning of the full penetration weld, but steadily over seam is achieved, and the zero-crossing rate is reduced by
a weld seam length of approx. 4 mm. The arithmetic average approx. 24,2%. If the laser beam power is increased further, the
deviation of the signal intensity when comparing both states zero-crossing rate increases again. The zero-crossing rate of
(full penetration weld and partial penetration weld) is 21,65%. specimens with a material thickness of t1 = 1 mm increases up
The marginal areas and the transition areas are neglected. to laser beam power of 3,6 kW and decreases when the laser
beam power is increased to 4,0 kW. At a laser beam power of
534 Patrick Meyer et al. / Procedia CIRP 124 (2024) 530–535

3,2 kW, the arithmetic mean values of all specimens (full

penetration weld and partial penetration weld) are close to each
other. The maximum percentage deviation is determined
between the specimens t1 = 1 mm and t2 = 2 mm with a value
of 12,2%. If the laser beam power is increased, the percentage
deviation of the specimens from each other increases. At a laser
beam power of 4,0 kW, the maximum percentage deviation of
the zero-crossing rate between the specimens t1 = 1 mm and
t3 = 4 mm is determined with a value of 31,2%
To further investigate the influence of the material thickness, a
trial is carried out with the multi-stage specimen. The material
thickness of the specimen is reduced from t = 6 mm to 1,8 mm
over the length of the specimen. The difference in material
thickness between two stages is Δt = 0,1 mm. For the analysis,
an evaluation interval with a step length of 2,3 mm (≙ 460
measured values� is defined for each step. With the welding
parameters used (welding speed of vP = 1,5 m/min, a laser beam
power of PL = 1,5 kW and a spot diameter of ØSpot = 0,160 mm),
the first full penetration weld spots are localized from a
material thickness of t = 4,7 mm. From a material thickness of
t = 4,5 mm, a continuous weld seam can be seen on the bottom
side of the specimen. The bottom side of the welded specimen
is shown in figure 6.

Fig. 6. Weld seam: Specimen (b) multi-step geometry (bottom side)

Fig. 7. Specimen (b): multi-step geometry

A reinforcement of approx. 0,5 mm is measured on the weld
seam (top side of the specimen). In the area of the full
penetration weld seam, the reinforcement corresponds to
approx. 0 mm. The value of the wavelength intensity quotient 4. Conclusion
of 1500 nm and 384 nm decreases with the start of the full
penetration weld. The decrease does not occur abruptly to a The impact of full penetration welding on the multispectral
constant intensity, but steadily with varying values of the signal during laser beam welding was investigated. The emitted
gradient values. An approximately constant intensity is only process radiation was recorded with a multispectral sensor and
achieved 46 mm after the start of the full penetration welding the measured values were evaluated wavelength-selectively.
process at a material thickness of 2,5 mm. The value of the Wavelength ratios were also calculated to observe a relative
interquartile range IQR as a measured of the statistical scatter key figure.
decreases with the start of the full penetration welding process.
The statistical scatter within an interval of the step decreases. The findings can be summarized as:
The statistical scattering of the interquartile ranges along the  The intensity of the multispectral signal decreases at
weld seam is increased compared to the partial penetration constant welding parameters when a full penetration
weld area. The result of the intensity analysis (signal intensity weld occurs,
quotient of 1500 nm and 384 nm and the IQR analysis is shown  The signal intensity does not drop abruptly to a
in Figure 7 as a function of the material thickness. The step (the constant level, but decreases steadily,
material thickness) from start of a continuous weld seam on the  In addition to the signal intensity, the statistical
bottom of the specimen is marked with a dashed line at
scattering of the individual measured values and the
t = 4,5 mm.
frequency (zero-crossing rate) of the signal also
decreases,
 Patrick Meyer et al. / Procedia CIRP 124 (2024) 530–535 535

 Based on the results, a correlation between the signal [6] Hollatz S. Funktionsorientiertes Laserstrahl-Mikroschweißen von
intensity of the full penetration welds and the ratio of Aluminium-Kupfer-Verbindungen mit örtlicher Leistungsmodulation.
Edition Wissenschaft Apprimus Aachen.
weld penetration depth to material thickness is DOI: 10.18154/RWTH-2023-10190; 2023
assumed [7] Volpp J, Vollertsen F. Keyhole stability during laser welding – part I:
modeling and evaluation. In: Production Process Volume 10.
 The consideration of the ratio of NIR (1500 nm) and DOI: 10.1007/s11740-016-0694-3; 2016
VIS (384 nm) was evaluated as effective for the [8] Wagner J, Hagenlocher C, Hummel M, Olowisnky A, Weber R, Graf T.
High-speed synchrotron X-ray investigation of full penetration welding of
determination of a full penetration weld during the aluminum sheets. Lasers in Manufacturing Conference 2021
trials. [9] Giannakopoulos T, Pikrakis A. Introduction to Audio Analysis: A Matlab
Approach. ISBN: 978-0-08-099388-1; 2014
5. Outlook

Further research will be based on the results presented here.

One question that needs to be investigated further is the
influence of different welding parameters and alloy
compositions. In a next step, the correlation coefficient
between the wavelengths should also be investigated. By
superimposing the reflected laser radiation with the emitted
radiation in the VIS and NIR, new insights can be gained.
Applying signal smoothing, for example a moving average to
interpret the results and performing a Fast Fourier
Transformation FFT for frequency analysis could provide
further insights. Finally, a closer look at the transition from full
to partial penetration weld and its influence on the signal offers
new research opportunities.

Acknowledgements

This work is funded by the Federal Ministry of Education and

Research (BMBF) of the Federal Republic of Germany and
supervised by the Project Management Agency Karlsruhe
(PTKA) as part of the joint project "Digitaler Prozess-
Onlineoptimierer für intelligente Lasermaschinen - DIPOOL"
under the funding code 02P20A000. We would like to thank all
our project partners.

Furthermore, the authors thank Marc Hummel and Jan-Patrick

Wille for their support during the experiments.

References

[1] Helm J. Prozessstabilität und Prozesseffizienz beim Laserstrahlfügen von

hoch reflektiven Kupferwerkstoffen. Edition Wissenschaft Apprimus
Aachen.
DOI: 10.18154/RWTH-2022-09483
[2] Yu J, Lee H, Kim D-Y, Kang M, Hwang I. Quality Assessment Method
Based on a Spectrometer in Laser Beam Welding Process. In: Metals,
10(6), 839.
DOI: https://doi.org/10.3390/met10060839; 2020
[3] Huber S. In-situ-legierungsbestimmung beim Laserstrahlschweißen zur
Prozessüberwachung. Herbert Utz Verlag München.
ISBN: 978-3-8316-4370-7; 2014
[4] Gatzweiler W, Maischner D, Beyer E. Messung von
Plasmadichtefluktuationen und Schallemissionen beim
Laserstrahlschweißen zur Prozessüberwachung. In: Laser und
Optoelektronik; 1988
[5] Huber S, Glasschroeder J, Zaeh MF. Analysis of the Metal Vapour during
Laser Beam Welding. In: Physics Procedia 12:712-719.
DOI: 10.1016/j.phpro.2011.03.089; 2011