64 IEEE SENSORS JOURNAL, VOL. 12, NO. 1, JANUARY 2012 Sensor for the Detection of Protective Coating Traces on Boron Steel With Aluminium–Silicon Covering by Means of Laser-Induced Breakdown Spectroscopy and Support Vector Machines Francisco Anabitarte, Jesus Mirapeix, Olga Maria Conde Portilla, José Miguel Lopez-Higuera, Senior Member, IEEE, and Adolfo Cobo, Member, IEEE Abstract—Welding processes are one of the most widespread in- dustrial activities, and their quality control, in both online and offline methods, is an important area of research. In the partic- ular process of laser welding of boron steel with aluminium–sil- icon covering in the automotive industry, one problem is the pres- ence of residual traces from the protective antioxidant coating, an aluminium–silicon alloy, which can result in a significant reduc- tion of the welding seam strength. This work proposes a sensor system based on a laser induced breakdown spectroscopy (LIBS) setup to detect and discriminate aluminium residues in the welding area without destroying the sample before the welding procedure. A spectral algorithm based on support vector machines (SVMs) is used as a classifier to automatically identify areas with aluminum presence. Index Terms—Feature selection, laser induced breakdown spectroscopy, laser-welding, plasma spectral analysis, quality monitoring, support vector machines. I. INTRODUCTION W ELDING processes are of great importance in many industrial sectors, and a great deal of research effort has been devoted to the design, monitoring, and control of the process, in order to assure the final quality of the weld seam. Many process parameters have influence on the resulting quality: type and composition of materials, joint preparation and cleanness, relative position between the pieces to be joined, the welding technique (laser, arc, friction, etc.), filler material, protection gas, speed, etc. [1]–[4]. One particular problem faced by many industries, and in par- ticular, the automotive industry, is the laser welding of boron steel with an aluminium–silicon covering, which is a common Manuscript received November 11, 2010; revised December 27, 2010; accepted February 13, 2011. Date of publication March 03, 2011; date of current version November 29, 2011. This work has been co-supported by the Spanish Government through the project TEC2007-67987-C02-01 and the Grant AP2007-02230. The associate editor coordinating the review of this paper and approving it for publication was Prof. Jose Santos. The authors are with the Photonic Engineering Group—Universidad de Cantabria, Edificio I+D+i Telecomunicación-Dpto. TEISA, 39005 San- tander, Spain (e-mail: anabitartef@unican.es; francisco.anabitarte@unican.es; mirapeix@unican.es; jesus.mirapeix@unican.es; olga.conde@unican.es; miguel.lopezhiguera@unican.es; adolfo.cobo@unican.es). Digital Object Identifier 10.1109/JSEN.2011.2121902 process for the manufacturing of several automotive compo- nents. Laser welding has many advantages in this case, as it is a clean and rapid process, produces narrow welds with a small heat affected zone (HAZ), and is a proven technology [5]. However, boron steels with an aluminium–silicon covering are supplied with an antioxidant protective coating usually made of an aluminium and silicon alloy, only some tens of microm- eters thick. The presence of residues of aluminum from this coating during welding has been proven to negatively affect the welding quality, resulting in a final joint with a significantly reduced strength [3]. For this reason, this protective layer has to be removed from the joint area before the welding process. Laser ablation is an effective and rapid process commonly em- ployed to remove the protective coating, but the possibility of traces of the protective material remaining is still not negligible. Thus, a sensor system to detect these residues before the welding process would be desirable. Some attempts to solve this problem using artificial vision have been unsuccessful due to the change in reflectivity of the metallic surface. For this purpose, laser induced breakdown spec- troscopy (LIBS) can be a useful tool, because this technique is very powerful for detecting traces of metals and other materials [6]–[13], with a minimum impact on the material surface. LIBS is a method sensitive to all the elements of the periodic table [8], [9], and it is considered as a nondestructive process because the laser ablates a few nanograms of the sample [8], [9]. Accordingly, LIBS could be a suitable solution to the problem of detecting remaining aluminum traces on the steel surface, before the welding process. However, for this approach to be practical, not only is the sensitive detection of impurities an important feature, but also the automatic and real-time identification of the contaminated surface. This could enable the prevention of potentially de- fective weld with a noninvasive system that could operate after the ablation process or right before the welding process. For this purpose, a fast and efficient algorithm is needed to process of the vast amount of spectral information. Support vector machines (SVMs) [14]–[16] can meet this requisite. There have been many investigations into SVMs applied to spectral data [17]–[20], some specifically regarding image classification in remote sensing [21], medicine [22], and the 1530-437X/$26.00 © 2011 IEEE