Vous êtes sur la page 1sur 7

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/297962661

Seam Welding Monitoring System Based on Real-Time Electrical Signal


Analysis A new approach to in-line seam welding quality evaluation is
presented

Article  in  Welding Journal · October 2010

CITATIONS READS

0 141

3 authors, including:

Bruno Riccò
University of Bologna
387 PUBLICATIONS   7,360 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Wireless Inertial Sensing for Body Area Networks View project

Augmented Graphics View project

All content following this page was uploaded by Bruno Riccò on 15 April 2016.

The user has requested enhancement of the downloaded file.


Lanzoni et al Supplement October 2010:Layout 1 9/8/10 8:56 AM Page 218

Seam Welding Monitoring System Based on


Real-Time Electrical Signal Analysis
A new approach to in-line seam welding quality evaluation is presented

BY MASSIMO LANZONI, MIRKO SALOMONI, AND BRUNO RICCÒ

ered in this paper where the electrodes


ABSTRACT (hence, also the welding point) are con-
stantly moving along a line.
This paper presents a novel welding quality evaluation approach based on the analy- In the specific case of seam welding, a
sis of electrical signals. The method has been implemented in an automatic system fault is represented by a missing welded
developed using field-programmable gate array (FPGA) acquisition boards and custom point segment, possibly due to a variety of
software. The system has been implemented in an industrial machine to detect faults in reasons (misalignments of the parts to be
tinplate welding. Experimental results show that the system is characterized by high welded, border defects, dust or oil on the
sensitivity and high speed. Furthermore, it needs simple conditioning electronics and welding area, and so on). In some cases
tuning procedures. The system is suitable for network operation and remote control. (i.e., welding of cans), this type of defect
WELDING RESEARCH

may result in gas or liquid leakage, unac-


Background to catastrophic results, costly X-ray or γ- ceptable in cans for pressurized gas (that
ray inspections are made, and the shape of must have zero welding defects) or only
Industrial machines are generally parts the image density profile along the weld partially tolerable in cans for solids of
of complex production systems where interface is analyzed (Ref. 3). These tech- high-viscosity liquids.
product quality must be automatically niques, however, can be applied only when Modern seam welding machines can
evaluated by means of several checks dur- welding process times are compatible with work at high speed (several m/s) and pro-
ing intermediate fabrication steps. In re- the duration of suitable image processing. duce welding dots with dimensions in the
cent years, the importance of reliable Furthermore, visual techniques are not order of 1 mm. As a result, thousands of
quality evaluation has greatly increased applicable when the welding area is not welding dots per second must be analyzed
due to the reduction in production gain visible, as is the case of interest for this and processed in real time. To monitor
margins. work. their quality at full speed, all the tech-
In this context, a typical step of a pro- An alternative visual technique is niques mentioned previously (Refs. 1–7)
duction line, which is sometimes responsi- based on the analysis of the welding tem- are not applicable because they do not
ble for important faults, is welding. There- perature (Refs. 2, 5), measured in real meet measurement speed and/or spatial
fore, nondestructive and automatic time using an infrared sensor and welding resolution specifications. In particular, no
methods to monitor welding quality are torch control. In this case, however, at method based on visual inspection (Refs.
normally implemented. Depending on least one side of the welding area must be 1–5) can be used because a) welding takes
welding type and speed, the following visible, and again this is not the case of the place between two surfacing edges of thin
methods are based on different tech- seam welding machines, where both sides plate; b) welding rollers mask the welding
niques: visual (Refs. 1–5), ultrasonic in- of the welding point are masked by the area; and c) complex signal processing al-
spection (Ref. 6), electrode distance (Ref. welding tool. gorithms are not compatible with the ma-
7), and electrical (Refs. 8–13). Another parameter that can be ob- chine speed.
Unfortunately, in the case of the seam served for quality monitoring is the elec- Furthermore, ultrasonic techniques
welding machines considered in this trode displacement during welding (Ref. such as those illustrated in Ref. 6 are also
paper, none of the available techniques is 7). Such a technique, however, is applica- not applicable because a) the part to be in-
applicable. ble only when the welding electrodes and spected is not fixed with respect to a
Visual inspection is based on the use of the parts to be welded are fixed during the probe; and b) the resolution is too low.
template matching techniques; in particu- operation, which is not the case consid- As an alternative, real-time monitoring
lar, images are captured after the welding of seam welding quality can be achieved by
process is completed, and processing algo- capturing the waveform of welding cur-
rithms are used to compare them with ref- rent and electrode voltage to be compared
erence images (for example, Ref. 1, com- KEYWORDS with correct references, because defects
puting the cross correlation between the produce anomalous electrical behavior.
target image and a portion of the acquired Process Monitoring This technique has been successfully
one). In some cases, where faults may lead Resistance Seam Welding implemented in the case of spot welding
Quality (Refs. 8–13) where electrodes remain in
Welding Defects and Flaws the same position during the whole weld-
MASSIMO LANZONI (massimo.lanzoni Signal Processing ing period, lasting several inverter cycles
@unibo.it) and BRUNO RICCÒ are with DEIS (from 6 to 10). Because the inverter fre-
Real Time
University of Bologna, Italy. MIRKO SA-
LOMONI is with Cevolani SPA, Bologna, Italy. Virtual Instrumentation quency is 60 Hz, under such conditions,
the time interval available for measure-

218-s OCTOBER 2010, VOL. 89


Lanzoni et al Supplement October 2010:Layout 1 9/8/10 8:56 AM Page 219

Fig. 1 — Working principle of the welding machine described in this work. Fig. 2 — Electrical schematics of the welding machine power circuit. VR rep-
The bent tinplate is driven between two welding rollers. Welding takes place resents the voltage between the welding rollers. The system features two trans-
in the contact area between the surfacing can edges (weld interface). formers; one (TA) is used to measure the current flowing through the pri-
mary coil of the other (power) transformer (PT).

ments corresponds to 100–160 ms, and the ification or fine tuning of the system sim- tinplate rises locally to high values, and
shape of the resistance vs. time (in practice ply by acting on the system software, ide- welding takes place.
as a function of the inverter periods) is ally without machine stopping, or the need During this process, several types of in-
used (Refs. 10, 12) to decide about weld- of special equipment as well as trained accuracies localized in the border of the

WELDING RESEARCH
ing quality by comparing it with experi- personnel. tinplate may lead to incorrect or missing
mental or simulated reference data. Un- This paper improves the field intro- welding. In particular, the presence of
fortunately, this approach is not ducing a new approach based on the com- dust or oil can locally increase the resist-
applicable to the case of seam welding of bination of high-performance analog sig- ance, modifying the electrical behavior of
interest in this work, essentially because nal conditioning circuits and off-the-shelf the system; holes in the tinplate can allow
the electrodes are moving along a (weld- advanced digital acquisition boards. The direct contact between electrodes; and
ing) line, and several points need to be result is an advanced monitoring system random variations of the tinplate thick-
welded in a short time. Each weld takes featuring on-board signal processing that ness, misalignment of the border to be
place in a single inverter semiperiod, near can be programmed with a high-level lan- welded, and/or local imperfections can
the current maximum value. Therefore, guage (LabVIEW™). Consequently, algo- significantly change the values of meas-
because in our case the inverter frequency rithm modifications are simple and do not ured resistance. In all these cases, the me-
is about 1 kHz, the resistance (or a related involve specific instrumentation. Further- chanical characteristic of the nuggets pro-
parameter) must be analyzed in a few hun- more, because the whole system is man- duced in the defective area are different
dreds of microseconds; thus, only a single aged by a personal computer (PC), net- than the reference values and the results
(effective) resistance value, instead of a working, remote control, and updating are weak. In all the cases mentioned, vari-
whole waveform, can be taken. can be easily implemented. ations in the electrical behavior of the sys-
In this regard, a major point of our re- To describe the developed system, this tem during welding are produced. Struc-
sults is that they clearly show this infor- paper is organized by giving a brief de- tures appear in the waveforms of the
mation is sufficient to effectively detect scription of the welding machine used as a current and the voltage between elec-
defective welding. target for this work; describing the archi- trodes that can be used for fault detection.
In the case of seam welding, the ap- tecture and main elements of the moni- Figure 2 presents a schematic repre-
proach based on analysis of voltage and toring system; providing some details sentation of the electrical power circuit of
current presents several specific and criti- about software implementation; showing the seam welding machine considered in
cal problems. First of all, as mentioned some experimental results; and ending this work.
previously, analysis speed is much higher with a final discussion. A power inverter produces a dual po-
than in spot welding and requires dedi- larity pulse width modulation (PWM)
cated high-speed signal analysis circuits. Operation of the Welding voltage signal to drive the primary coil of
In this context, the conventional analog Machine a power transformer (PT). The primary
approach to signal analysis has several dis- peak current is in the order of 50 to 150 A
advantages because a) the system design is The welding machine considered in and is monitored by means of the current
expensive, necessarily customized for a this work is able to produce a sequence of transformer (TA). The secondary of the
specific machine, difficult, and expensive welding points, called nuggets, along the PT is connected to the rolling electrodes,
to modify; and b) accurate and critical tun- edge of a tinplate suitable for can fabrica- and the voltage across these (essentially
ing is normally required. Alternatively, tion. The principle of the welding opera- the voltage drop on the resistance in the
signals can be sampled and digitally tion is illustrated in Fig. 1. welding point) is monitored by means of
processed in real time, but high-perfor- The tinplate is bent to form a cylinder, sliding contacts. Welding current and volt-
mance embedded systems based on digital then the surfacing edges are welded by age difference across the rollers are the
signal processing (DSP) or field-program- means of a couple of rolling electrodes two signals to be analyzed for welding
mable gate array (FPGA) devices are connected to a power inverter. When the quality evaluation. Their typical wave-
needed to meet machine speed specifica- superimposed edges pass under them, forms are shown in Fig. 3.
tions. Compared with its analog counter- high current flows between the rolling As can be seen, in spite of the impulsive
part, the digital approach allows the mod- electrodes; thus, the temperature of the behavior of the voltage across the rollers,

WELDING JOURNAL 219-s


Lanzoni et al Supplement October 2010:Layout 1 9/8/10 8:56 AM Page 220

Fig. 4 — Schematic representation of the proposed welding monitoring system.

B
WELDING RESEARCH

Fig. 3 — Typical primary current and roller voltage during welding. Different Fig. 5 — Schematic representation of the algorithm implemented in the FPGA,
shapes of the PWM generator are used. A — The voltage across the rollers has showing the parallel execution of analog and digital signal processing. Analog
few pulses and produces a quasi-triangular behavior of the current. B — The signal sampling and conversion takes about 4 µs. The whole cycle duration is
number of pulses (within a period) is larger, and the current has a quasi-sinu- 5 µs, corresponding to 200-k samples/s.
soidal shape.

the welding current has a smoother shape welding period (0.7 ms). Such a high pro- able to perform real-time signal process-
due to the large inductance associated cessing speed can be reached only by means ing defined by algorithms written in a spe-
with the coils of the PT and secondary of dedicated analog circuits or high-speed cial version of LabVIEW™.
load. The peak value of the current flow- mixed analog/digital electronics. Figure 4 gives a schematic representa-
ing in the PT primary is in the order of 50 As already anticipated, the latter op- tion of the system.
A, corresponding to a peak welding cur- tions were chosen, which offer many ad- A Windows®-based PC is used for sys-
rent of 2000–4000 A depending on the op- vantages such as reduced custom design of tem management, operator interface, net-
erating conditions (tinplate thickness, both hardware and software, networking work connection, data collection, and
welding speed, and so on). capabilities, and low cost. storage. The PC is equipped with a NI
Because a nugget is produced both 7830 PCI acquisition board featuring a
when the current reaches a maximum and The Proposed System FPGA device that can be programmed to
minimum value, in the (typical) case of process in real time both analog and digi-
Fig. 3B, the duration of the welding period The monitoring system of this work, tal signals at 40 MHz clock rate, corre-
is about 0.7 ms. consisting of a high-speed section for real- sponding to a clock period of 25 ns. The
In our monitoring system, the wave- time signal processing and digital unit for hardware blocks inside the FPGA can be
forms of Fig. 3 are sampled at 200-k sam- the management of defective cans, is organized for parallel processing of sev-
ples/s; thus, a double set of about 140 sam- based on a new family of acquisition eral signals.
ples must be processed in less than a boards featuring on-board FPGA devices In our system, analog signals from the

220-s OCTOBER 2010 , VOL. 89


Lanzoni et al Supplement October 2010:Layout 1 9/8/10 8:57 AM Page 221

Fig. 6 — Schematic representation of the transport and ejection systems of Fig. 7 — Results of the analysis of a single can. Each point in the figure
the welding machine considered in this work. represents the quality factor (QF) of a nugget along the weld interface. As
can be seen, a set of nuggets exceeded the allowed thresholds (also see
Fig. 8).

sensors (few, but needing time-consuming of our system. certainness in material characteristics.
sampling and AD conversion) and many Two parallel data flows are imple- To implement the ejection function by
digital signals are processed in parallel. mented exploiting the parallel execution counting the transport encoder pulses, the
This allows performance of a whole sam- capability of the FPGA device. Starting FPGA algorithm evaluates the position of

WELDING RESEARCH
pling (AD conversion and partial process- when a new trigger event is detected (i.e., the welded cans within the transport sys-
ing cycle) in less than 200 clock periods the current crosses the zero line while ris- tem after the welding rollers and gener-
(i.e., 5 μs in total), resulting in a 200-k cy- ing or falling), a new data collection is ates an expulsion command when a faulty
cles/s continuous operation of the started. A loop consists in parallel sam- can reaches the right position in front of
algorithm. pling and conversion of the analog signals the eject system — Fig. 6.
Therefore, within a welding period of and sampling of digital inputs (can trans- As can be seen in Fig. 6, a number of
0.7 ms, 140 16-bit digital samples of the port encoder pulses, new can synchroniza- cans, depending on transport speed and
analog signals are acquired and processed. tion signal, and so on). At each loop, the tinplate format, is normally present within
This performance is sufficient to obtain analog samples are stored and digital sig- the transport system at the output of the
suitable accuracy in welding quality evalu- nals are processed. At the end of the weld- welding point. Therefore, a set of registers
ation, hence high resolution in defect ing period (i.e., when a new trigger event containing the position of every can is re-
detection. is detected), the collected analog samples quired. These registers store the number
A suitable signal conditioning stage has are evaluated to obtain the nugget quality of pulses of the transport encoder from
been designed to a) amplify analog signals factor (QF) expressed by the average con- the end of the welding and must be incre-
from the sensors to fit the dynamics of the ductance (i.e., the reciprocal of the resist- mented (in parallel) at every encoder
AD converters present on the FPGA ance) during a single welding period. The pulse. The eject system is activated if the
board; and b) shift the level of the 24-V sig- QF, determined by the current and voltage can in the ejection position (correspond-
nals from digital sensors and actuators waveform during welding, is automatically ing to the fixed distance from the welding
placed on the welding machinery to tran- compared with operator-defined limits, rollers) has been recognized as defective.
sistor-transistor logic (TTL) compatible and if these are exceeded, a fault signal is Because real-time operation is needed, a
levels. The gain of the top-quality ampli- produced and the faulty can is ejected. To portion of the FPGA is dedicated to up-
fiers used for the analog channel can be avoid errors due to random noise super- dating and sampling the register’s status.
varied to optimize the system for the par- imposed to the current signal, a Schmitt Therefore, the FPGA is able to auto-
ticular machine to be monitored. Further- trigger (realized via software) is used to matically manage both welding analysis
more, the accuracy of signal sampling and detect the trigger event. and defective can ejection.
conversion guarantees the effectiveness of Product quality standards can be easily Furthermore, the quality of all the
the analysis even if the input signals do not defined by the operator choosing the num- nuggets of the last welded can are stored
completely span the converter dynamics. ber of tolerable faults that depends on the in an internal memory, and the data can be
This is important from the point of view of application, because, for example, cans for transferred to the host program, without
system flexibility, because different signals aerosol normally require zero defects affecting the system operation, exploiting
(due to differences in tinplate thickness, while those for solid contents are more tol- the direct memory access (DMA) option.
welding current, and so on) can be treated erant. To manage this parameter, the op- These data provide important information
without the need to adjust the condition- erator can easily define/modify the limits to the operator (tinplate quality, adjust-
ing amplifiers gain to any specific case. for defect detection. This manual and em- ments to the machinery setup, and so on).
As for software, the operator/network pirical calibration is necessary because in The main LabVIEW™ program (host),
interface and system management program real production environments, the process running on a PC, is dedicated to service
are written in LabVIEW™. The same holds tolerances as well as the intrinsic random- tasks such as data collection and visualiza-
for the FPGA program (slave), that, how- ness in fault nature and entity may easily tion, operator interface, network connec-
ever, must be compiled and transferred to lead to a different variance in a statistical tion, and remote control of the monitor
the FPGA before system operation. data of good nuggets. Therefore, in some system. The main program panel allows
Figure 5 shows a block diagram of the cases, the sensitivity of the system must be the operator to view the data (of course, in
FPGA algorithm that represents the core suitably adjusted to compensate for un- compact form) concerning a set of cans.

WELDING JOURNAL 221-s


Lanzoni et al Supplement October 2010:Layout 1 9/8/10 8:57 AM Page 222

Fig. 9 — Example of defect due to border


imperfection.

Fig. 8 — Example of welding analysis in the case of quasi-triangular welding current. Each point repre- abnormal value of the maximum or mini-
sents the result of the analysis of a set of nuggets within a can. The following three curves are displayed: mum QFs (red and green curves,
The red and green curves represent the maximum and minimum values of QFs recorded during the analy- respectively).
sis of the can while the yellow line represents the average of all the QFs. Defective welding leads to ab- Through the main panel, the operator
normal values that can be detected by comparison with thresholds indicated by the horizontal red and can easily set boundary levels for can se-
green dotted lines. Nondefective cans are characterized by maximum and minimum QFs within the range
delimited by these thresholds. A third (blue) line indicates the threshold allowing detection of the absence lection calibrating the position of the max-
of tinplate that produces high values of the QF. In this case, the welding machine inverter is instantly imum and minimum horizontal lines
switched off to prevent electrode damage. (thus, selecting an allowed range for QF
values). In addition, the lack of tinplate to
WELDING RESEARCH

be welded can be easily recognized be-


cause QF assumes values well above the
The minimum, maximum, and average exception of a few nuggets located near maximum limit (red horizontal line). This
values of QFs measured for each nugget of the center of the can where a defect has condition can be detected by comparing
a single can are shown. been encountered, presenting QF exceed- the minimum QF detected (green curve)
As stated above, the host program is ing the maximum QF threshold (red hori- with a third threshold level (blue horizon-
able to manage data storage in the hard zontal line) set by the operator. In the spe- tal line). If this latter is found below min
drive of the PC. Furthermore, the opera- cific case in the figure, border nuggets QF, the system can immediately inhibit the
tor can store significant sets of analysis pa- (normally presenting slightly abnormal welding machine inverter.
rameters corresponding to various situa- QF values) have acceptable QFs; in oth- In the same way, insulation between
tions of can size, tinplate characteristics, ers, the threshold is exceeded. In any case, rollers (not shown in the figure) can also
etc. These sets of parameters can be trans- the operator can decide to exclude from be detected. Because this situation is also
mitted to the FPGA via the internal PCI the analysis border nuggets or apply for a potentially dangerous for the PWM in-
bus without affecting the measurement different decision threshold. verter, the FPGA immediately asserts a
speed. In general, the welding regions can be disable digital line that will be reset at the
subdivided in different areas to be ana- beginning of the next can.
Experimental Results lyzed with different sets of parameters. A The same analysis performed on a set
typical subdivision of practical interest is of cans welded using quasi-sinusoidal cur-
The proposed system has been suc- the beginning, center, and end of the cans, rent waveforms gives similar results in
cessfully implemented in commercial with the borders featuring different values spite of the fact that the voltage across the
welding machines. Different prototypes with respect to the center. In this way, the rollers features a high number of pulses,
have been applied to machines presenting effects of the cans’ borders on the final hence the noise induced on the measuring
different characteristics such as the output quality evaluation can be eliminated using system is higher.
waveform of the PWM inverter. specific sets of threshold values. Noise produces peaks in the current
As already mentioned, the system The operator also has the possibility of waveform and extra peaks in the voltage
speed is sufficiently high to allow data masking the analysis of the first and last waveform. However, the analysis is not af-
transfer from the FPGA to the main pro- nuggets. This feature is important to avoid fected because the differential amplifiers
gram via a DMA channel. Thanks to this mistakes due to the fact that the voltage used in the analog front end have a low slew
feature, the operator can verify the distri- waveform across the welding rollers is not rate limiting the effect of pulse noise. As a
bution of QF values along the weld inter- synchronous with the beginning of the tin- result, in the normal case (no faults), the
face of a single can — Fig. 7. This is im- plate to be welded. Thus, the first and last QF presents only slightly higher dispersion
portant to optimize the analysis, because nugget may feature incomplete welding; around the mean value than the values ob-
nuggets near the borders present slightly hence, their values must be discarded. tained in the case of Fig. 3A. Thus, small
different quality values compared to those Figure 8 shows typical results of the defects are not masked by the noise.
in the center of the can. This leads to a analysis of a set of cans. Each point of the Figure 9 shows a typical defect safely de-
widening of the min.-max. window and de- curves in this plot represents a synthetic tected by the system of this work. In the
creased analysis sensitivity. evaluation of the QF analysis of a whole case of the measurements presented in this
Figure 7 presents the typical results of set of nuggets within a can. The yellow line section, the speed of the tinplate was 1.3
the analysis of nuggets along the weld in- represents the average of the QF values. m/s, resulting in a distance between nuggets
terface of a single can welded using the As can be seen, the average is almost con- of 0.9 mm. Therefore, the defect shown in
waveforms presented in Fig. 3B. As can be stant from sample to sample. Fig. 9 results in two defective nuggets at
seen, the QF is almost constant with the On the contrary, defects produce an most.

222-s OCTOBER 2010, VOL. 89


Lanzoni et al Supplement October 2010:Layout 1 9/8/10 8:58 AM Page 223

System Features tion systems and high-quality conditioning X. 2007. Research on image resolution in ultra-
electronics. Such prototypes have been sonic imaging inspection of welding defect.
Calibration implemented on commercial machines Proc. IEEE 8th Intl. Conf. on Electronic Meas-
with excellent results. urements and Instruments (ICEMI2007), Xi’an,
China, pp. 965–968.
The system developed in this work The system is able to perform accurate 7. Lin, Z., Zhang, Y., Chen, G., and Li, Y.
does not require critical calibration proce- analysis in real time and allows remote as- 2004. Study on real-time measurement of
dures. The main operation is to optimize sistance, management, and data collection. nugget diameter for resistance spot welding
the gain of the amplifiers used in the con- The system is based on widely available using a neuro-fuzzy algorithm. Proc. IEEE Intl.
ditioning circuit to account for the charac- components and can be easily extended to Conf. on Instrumentation and Measurement
teristics of the particular machine where include additional sensors. Technology (IMTC 2004), Como, Italy, pp.
the monitor is installed. In fact, the signal 2230–2233.
at the output of the TA transformer can References 8. Savage, W. F., Nippes, E. F., and Wassel,
F. A. 1978. Dynamic contact resistance of series
vary within a large interval depending on
1. Kim, S.-M., Lee, Y.-C., and Lee, S.-C. spot welds. Welding Journal 57(2): 43-s to 50-s.
the machine, and the same holds for the 2006. Vision based automatic inspection system 9. Chien, C. S., and Kennatey-Asibu Jr., E.
voltage across the welding electrodes. for nuts welded on the support hinge. Proc. 2002. Investigation of monitoring systems for
Therefore, to exploit the acquisition board IEEE SICE-ICASE Intl. Joint Conf. 2006, resistance spot welding. Welding Journal 81(9):
AD converter (ADC) dynamics, the peak Busan, Korea, pp. 1508–1512. 195-s to 199-s.
values of the amplified signals must be 2. Govardhan, S. M., Wikle, H. C., Nagara- 10. Dickinson, D. W., Franklin, J. E., and
maintained at about 80% of the converter jan, S., and Chin, B. A. 1995. Real-time welding Stanya, A. 1980. Characterization of spot weld-
range. process control using infrared sensing. Proc. ing behavior by dynamic electrical parameter
A further calibration concerns the syn- IEEE American Control Conf., Seattle, pp. monitoring. Welding Journal 59(6): 170-s to
1712–1716. 176-s.
chronization of the transport and acquisi-
3. Li, Y., and Liao, T. W. 1996. Weld defect 11. Livshits, A. G. 1997. Universal quality
tion systems to be achieved, varying the detection based on Gaussian curve. IEEE Proc. assurance method for resistance spot welding
delay between a mechanically produced of the Twenty-Eighth Southeastern Symposium based on dynamic resistance. Welding Journal
electrical trigger and the beginning of the on System Theory, pp. 227–231. 76(9): 383-s to 390-s.
acquisition. 12. Cho, Y., and Rhee, S. 2002. Primary cir-

WELDING RESEARCH
4. Kim, C.-H., Choi, T.-Y., Lee, J.-J., Suht,
Both these procedures are made easy J., Parkt, K.-T., and Kangt, H.-S. 2006. Devel- cuit dynamic resistance monitoring and its ap-
by the execution of special calibration opment of welding profile measuring system plication to quality estimation during resistance
programs. with vision sensor. Proc. IEEE Sensors, EXCO, spot welding. Welding Journal 81(6): 104-s to
Daegu, Korea, pp. 392–395. 111-s.
5. Nagarajan, S., Benerjee, P., Chen, W. H., 13. Cho, Y., and Rhee, S. April 2004. Qual-
Human Interface
and Chin, B. A. Feb. 1992. Control of the weld- ity estimation of resistance spot welding by
ing process using infrared sensors. IEEE Trans. using pattern recognition with neural networks.
The human-machine interface of the Robotics and Automation 8(1): 86–93. IEEE Transactions on Instrumentation and
system presented in this paper has been 6. Guangzhi, D., Tiequn, C., and Jiaxiang, Measurements 53(2): 330–334.
carefully designed to allow nonskilled op-
erators to perform basic operations by
means of easy-to-use commands imple-
mented on a touch screen.
Of course, operators with higher skills
can log on the system and access extended
Do You Have Some News to Tell Us?
features such as statistics and calibration
procedures.
If you have a news item that might interest the readers of
Network Integration the Welding Journal, send it to the following address:
Welding Journal Dept.
The monitor system of this work is suit-
able for network integration in the manu-
Attn: Mary Ruth Johnsen
facturing system. The main program is 550 NW LeJeune Rd.
written in LabVIEW™, which provides an Miami, FL 33126.
integrated Web server allowing remote Items can also be sent via FAX to (305) 443-7404 or by e-
control. The system can be completely
controlled from a remote workstation run- mail to mjohnsen@aws.org.
ning any Web browser. This feature is use-
ful for effective surveillance of the plant
from a single location or to improve re-
mote assistance in the case of faults.

Conclusions
Change of Address?
This work presented a welding moni-
Moving?
toring system able to detect faults in real
time and with high sensitivity. Make sure delivery of your Welding Journal is not interrupted.
The system is based on measurements Contact the Membership Department with your new address
and processing of electrical signals from
sensors placed on the welding transformer information — (800) 443-9353, ext. 217; smateo@aws.org.
and electrodes.
Working prototypes of the system have
been developed using commercial acquisi-

WELDING JOURNAL 223-s

View publication stats

Vous aimerez peut-être aussi