Cảm biến trong sản xuất P14

machining system by moving the workpiece back to the machining position and performing additional machining. In other words, the dimensional error of ma- chining can be corrected before demounting the workpiece from the machine. Concerning the VS method, applications for the measurement of practical parts and improvements in reproducibility and accuracy are currently being investigated. 4.7 Welding H. D. Haferkamp, Universität Hannover, Hannover, Germany, F. v. Alvensleben, Laser Zentrum Hannover, M. Niemeyer, Universität Hannover, W. Specker, Laser Zentrum Hannover, M. Zelt, Universität Hannover 4.7.1 Introduction Quality in welding depends not only on the various process parameters such as welding speed, current, and voltage, eg, for arc welding. In most cases, the weld- ing process is affected by: · tolerances and mismatch of the workpiece geometry; · tolerances in machines and clamping devices; · variations in groove shape; · tack beads; and · welding deformations. In manual welding, trained welders are able to compensate for all these influ- ences, because their senses, especially the eyes and ears, give them the informa- tion they need to produce a high-quality weld. For mechanical and automatic welding, all this information must be detected by sensors. Such a sensor for weld- ing can be defined as follows [1]: 4 Sensors for Process Monitoring286 Fig. 4.6-10 Set-up for on-the-machine measurement Sensors in Manufacturing. Edited by H.K. Tönshoff, I. Inasaki Copyright © 2001 Wiley-VCH Verlag GmbH ISBNs: 3-527-29558-5 (Hardcover); 3-527-60002-7 (Electronic) ‘A detector, if it is capable of monitoring and controlling welding operation based on its own capacity to detect external and internal situations affecting welding results and transmit a detected value as a detection signal, is called as a sensor. Moreover its whole control device is defined as a sensor system (control system)’. In this definition, the external situation refers to all workpiece-related geometric values such as changes in dimensions of the welding groove, position of the weld- ing line, and presence of component obstacles or tack welds. The internal situa- tion covers factors such as the shape of the welding arc and molten pool, the penetration depth, and all kinds of effects related to the welding process itself [2]. In general, every physical principle which is able to deliver information about an object’s shape and position may be the basis of a sensor. For welding sensors, the special ambient conditions and the industrial constraint for economic effici- ency, however, cause many additional restrictions, such as: · process-induced disturbances, such as light, heat, fume, spatter, and electromag- netic fields, must not influence the sensor; · the sensor must be satisfactorily durable for welding ambience; · it must be compact in size and light weight, so that there are no restrictions in handling and accessibility; · the sensor system must only generate low costs; · it should have easy maintenance. Owing to this and because of the very complex process, so far no universal sensor for welding is available which meets all these requirements and is able to detect all the various kinds of information by which the welding process is influenced. For the user, it is necessary to select the most satisfactory sensor type for every special welding task [2–5]. In general, a classification of welding sensors can be made by their functional principle. In Figure 4.7-1 such a classification of sensors for welding is shown in accordance with [3]. Further classification is made by the physical principle on which the sensor is based. 4.7.2 Geometry-oriented Sensors Geometry-oriented sensors gather their information from the geometry of the welding groove itself or from an edge or a surface which has a defined relation to the seam. Geometry-oriented sensors can be divided into contact and non-contact types. 4.7.2.1 Contact Geometry-oriented Sensors Contact sensors permit the detection of the welding start/end point and seam tracking with comparatively low technical expenditure. The first seam tracking 4.7 Welding 287 systems in welding were based on the mechanical tracing of a gap or an edge by a probe and the direct transformation of movements to the torch by way of ful- crums. The form of the probe, eg, as a ball, roll, or ring, must be adjusted to the groove geometry. Further technical development leads to electromechanical sensor systems, which convert the probe movement into electrical signals. Generally, there are two different kinds of these sensors. One group operates with limited switches, which deliver an on/off output signal, to track the seam stepwise. The other group uses potentiometers or differential transformers to generate a dis- tance-proportional output signal, which allows continuous seam tracking. In Fig- ure 4.7-2, the principle of these two sensor groups is shown [1, 6, 7]. The probe may have one or two degrees of freedom, so it is able to compensate for most two-dimensional deviations of the weld seam. Usually, the contact probe sensor is in a fixed position ahead of the welding torch, which causes some lim- itations in the shape of the welding seam. Because of the distance between sensor 4 Sensors for Process Monitoring288 Fig. 4.7-1 Classification of sensors Fig. 4.7-2 Contact probe sensors: (a) limited switch type; (b) potentiometer type and welding torch, which leads to deviations on curved seams, this method is commonly used for welding straight-lined seams. In order to apply contact sensors to welding curved seams, several techniques have been developed. Bollinger [8] described a method that is based on the turn- ing of the complete fixed torch-sensor unit around defined axes. From the mea- sured turn angles, the path feed rate of the different axes is calculated. Neverthe- less, this system leads to some deviation on seams with a small radius of curva- ture. To avoid this, it is necessary to monitor the welding seam using the sensor, prior to welding, and store the deviation values of the seam. With the stored coor- dinates, the system allows the welding of small curve radii with constant torch ori- entation. However, the system is not suitable for welding closed contours. The maximum deviation angle to the mean welding direction is given by the author as ±608. Another method for contact sensors for welding curved seams is based on the mechanical decoupling of the sensor and torch motion [9]. In these systems, called memory delay playback, the sensor is mounted on a separate x-y drive block, which allows tracing of the shape of the groove independent of the torch movement. The groove deviation values are stored, and based on the welding speed and the distance between sensor and torch, the correct position of the weld- ing torch is adjusted when it is moving towards the former sensor position. This sensor system leads to satisfactory results, even in welding small bending radii. Nevertheless, this system also is not capable of welding closed contours. It is just able to compensate deviation angles of ±308 to the mean welding direction. Another way of using a contact sensor for welding any curved seam is detection of the seam deviation prior to welding, and compensation of the programmed seam line by the measured deviation values, as described by Schmidt [10]. Prior to the first weld, a contact probe sensor is mounted in lieu of the torch gas cup, and the welding robot senses the deviation of the workpiece weld line to the pro- grammed one. After this sensing cycle, the normal gas cup is mounted, and the robot starts to weld the rectified seam line. This method calls for a separate mea- sure cycle prior to every weld, and is not able to compensate for deviations that occur during welding, eg, due to thermal distortion. A seam tracking system has been described [11, 12] which was developed for laser beam welding of three-dimensional fillet welds using an industrial robot. The mechanical sensor consists of a metal tracer pin, which is dragged along the fillet joint at a fixed distance ahead of the laser spot. Positional changes induce potentiometric variations at the head of the pin. The sensor feeds these variations to an electronic controller as a stream of analog data. The controller then guides the laser spot accordingly, by means of two servo motors which give the vertical and transverse motion of the laser head, using commercial systems, within a range of ± 7.5 mm and with an accuracy of ± 0.1 mm. Welding speeds up to 6 m/ min are possible with current systems. Generally, the use of contact probe sensors is limited by the wear of the probe itself. Because of permanent contact to the workpiece, there is marked friction wear, which decreases the probe lifetime. Further limits are caused by the accessi- 4.7 Welding 289 bility of the joint. The additional sensor mounted in front of the torch means a significant enlargement of the tool, and so welding in confined areas is restricted. Nevertheless, contact probe sensors have been widely used in many industrial applications for a long time, especially owing to their simple design and easy han- dling and maintenance. Another special contact sensor is the electrode or wire contact sensor. This is a sensing method which was developed for arc welding robots. It is able to detect deviations between a taught point of the robot path and the present position of the welding torch. In Figure 4.7-3, the principle of this sensing method is shown. The basic idea of this sensor is to use the torch as a switch in an electric cir- cuit. In this circuit, the workpiece surface and the welding wire have different po- larity. When the wire comes into contact with the workpiece, a change in electric current or voltage can be detected. The difference of the taught point and the ac- tual position can be calculated, and the real position of the workpiece is defined. For using the welding wire as a probe, the stick-out length of the wire must be de- fined. Therefore, wire extension may be determined by automatically cutting it to length prior to sensing, or it can be calculated by sensing a machine reference point, which has no initial deviations prior to workpiece sensing. Another method is the use of the welding torch gas cup as the contact dip [13–16]. The electrode contact sensor is industrially used in robotic welding to detect var- iations of the starting and end points of welds and of the length of welds and in sensing the form of the welding gap prior to welding. They are simply designed, easy to use, and not subject to wear. This sensor type is able to achieve an accu- racy of ± 0.2–0.3 mm in position detection. Beyond that, they cause no restrictions in accessibility of the joint, because there are no additional extensions to the torch [14–17]. In general, the use of these kinds of sensor can be limited by all kinds of insu- lating coatings on the workpiece, such as primer or oxide layers. Furthermore, the electrode contact sensor is not able to allow for deviations which occur during welding, eg, due to thermal distortion. Hence, it is usually used for short welds or in combination with an additional seam tracking sensor, eg, a through-the-arc sen- sor [13–17]. 4 Sensors for Process Monitoring290 Fig. 4.7-3 Principle of an electrode contact sensor 4.7.2.2 Non-contact Geometry-oriented Sensors A further development in sensor systems is the non-contact geometry-oriented sensors. These sensors are based on various physical principles of measurement (see Figure 4.7-1). Generally, they deliver information about the workpiece shape and its position in space. Depending on the sensitivity and accuracy of the sens- ing system, non-contact geometry-oriented sensors are able to detect the start and end points of welds and track weld seams. The most commonly used types in this category of sensors are based on optical, electromagnetic, and acoustic measure- ment. The fourth category of pneumatic sensors from the list in Figure 4.7-1 use the impact pressure of a gas nozzle to detect the distance between the workpiece surface and the sensor. This sensing method is not commonly used in welding processes at present. In laser welding long seams, the problems resulting from the geometric accu- racy of the workpiece become a decisive factor. Industrial robots are often used to guide the laser head or the welding torch along the workpiece. In laser welding the robot-guided beam must follow the (three-dimensional) seam geometry accu- rately, because focus diameters are typically in the range 0.15–0.5 mm. Addition- ally, any movement out of the focal plane (eg, the distance workpiece lens changes) can cause a defective weld. The robot is usually programmed manually, using a time-consuming point-by-point basis, so that curves are often estimated. In addition to that in arc welding, the process caused thermal distortion of the workpiece, often leading to geometric deviations of the joint line. Optical sensors, which use the topography of the workpiece surface in order to detect the weld seam, belong to the non-contact geometry-oriented type. The basic principle of the optical measurement used in this sensor group is a light-section procedure. Using a laser diode, a line-shaped laser beam is projected on to the 4.7 Welding 291 Fig. 4.7-4 Principle of a laser-stripe sensor workpiece (see Figure 4.7-4). A variation in the distance between sensor and work- piece leads to a change of the reflected beam position. This reflected beam is mea- sured by a charge-coupled device (CCD) camera whose data are processed by a PC, in order to calculate the workpiece surface contour. These data can be used for seam tracking, groove shape detection, and detecting weld start/end points [18–23]. The data of the sensing system are transmitted to the handling system, in order to correct the beam or torch position on the workpiece. Usually, it controls, eg, the robot directly via CNC commands. The measurement accuracy of commercial systems is 0.025 mm, and these systems are suitable up to maximum welding speeds of 15 m/min. The positioning accuracy also depends on the handling sys- tem. Both optical components, laser diode and CCD camera, are adapted to the la- ser head and to the welding torch, which makes it sensitive to alignment, dust, fumes, and spatter. The optical method has the drawback that reflections and scratches on the workpiece surface may cause the system to go astray. Electromagnetic sensors are non-contact geometry-oriented sensors, which gain their information by the effect of metallic materials on electromagnetic fields. These sensors, used to detect position or displacement, are classified into capaci- tance and eddy current types. Capacitance sensors measure the capacity between the workpiece and a small electrically conductive plate. They offer the possibility of distance sensing. Matthes et al. [24] described a capacitance sensor for seam tracking in V-grooves. The sensor signal of capacitance sensors is heavily vitiated by deviations in flatness or parallelism of the workpiece surface. Hence this kind of sensor ordinarily is not used in welding, but sometimes is in thermal cutting [2, 3]. The eddy current type is based on the interaction of metallic materials and alter- nating magnetic fields. The sensor induces eddy currents in the near-surface range of the workpiece. These eddy currents influence the inductance of the sen- sor coil, depending on the distance between sensor and surface. From this influ- ence, a distance-dependent electrical signal is obtained [2, 3, 25–28]. The principle is shown in Figure 4.7-5. Depending on the frequency of the eddy current, the sensor reacts in different ways to the various magnetic characteristics of the workpiece materials. Sensors 4 Sensors for Process Monitoring292 Fig. 4.7-5 Electromagnetic sensor, eddy current type with low-frequency eddy currents are only suitable for ferromagnetic materials, whereas high-frequency sensors are applicable to both ferromagnetic and non- magnetic materials [27, 29]. Electromagnetic sensors with one coil system are limited to detecting the dis- tance to the workpiece in one direction. Hence they are only able to adjust the torch’s height or lateral deviation. Sensors with a combination of several coil sys- tems, however, allow sensing a welding groove in every direction. In addition to height and lateral deviation, these systems can detect changes in the direction of the welding groove, the beginning and the end of a groove and some changes in the setting angle of the welding torch [29–32]. Because of the geometric distance between the torch and the electromagnetic sensor, these systems are affected by some deviations on curved welds (compared with contact probe sensors). To avoid these deviations, several methods have been developed. In one system [29], the sensor rotates around the torch, and the sensor signal is connected to the direction by a turn angle transmitter. Considering the welding speed and direction, the deviation between the sensor and torch can be compensated by the control system. This allows one to track curved seams in every direction with satisfactory precision. For tracking fillet welds, another possibility is to sense the weld flanks by a col- lateral arrangement of two sensors to the torch [2, 33]. Every sensor is arranged perpendicular to one flank. In that way, by sensing and adjusting the distance to them, the torch follows the seam. An eddy current sensor has been described [34] which is concentric to the torch and integrated in the gas cup. This leads to a very compact design, so accessibility problems are minimized. In general, eddy current sensors are able to compensate deviations with an ac- curacy of ± 0.15–0.5 mm. They are suitable for detecting almost every kind of groove shape. In butt joint welding they are able to track gaps up to a width of 0.05 mm. Nevertheless, the use of these sensors is limited in several ways. In gen- eral, some additional extension of the torch is necessary, so the accessibility of seams is limited. When welding butt joints, filler and cover passes are difficult to track using eddy current sensors. Edge misalignment on butt joints causes devia- tions to the center of the weld, so very exact preparation of the workpiece and reli- able fixture is essential for accurate seam tracking. Moreover, eddy current sen- sors are affected by any foreign magnetic field in the sensing area. Even geo- metric changes in the region of the seam, such as clamping fixtures, tack welds, workpiece thickness, and material non-homogeneities, can influence the sensor signal [3, 26, 29, 34–36]. In spite of these disadvantages, eddy current sensors are widely used in many industrial applications. Because of their robust design, universal application cap- ability, and comparatively low cost, their application is economical for a great vari- ety of sensor tasks. Another type of non-contact geometry-oriented sensor utilize ultrasonic signals to gather information. The principle of this kind of sensor (see Figure 4.7-6) is based on the fact that ultrasound waves are reflected from material surfaces, and that the propagation of these waves in air is related to the distance between the 4.7 Welding 293 ultrasonic transmitter and the receiver. For tracking weld seams, it is possible to use either the reflected energy amplitude or the range information, or both [37, 38]. In the first case, the sensor scans the area in front of the welding torch and finds the seam by detecting a modification of the reflected energy in the course of a scan cycle. This energy modification is caused by any change in the wave’s angle of incidence to the reflecting surface, eg, on groove edges. In the sec- ond case, the distance between the sensor and the workpiece surface is monitored by timing the interval between wave transmission and echo return. Based on these distance measurements in the scanned area, the workpiece profile along the scanning path can be determined. For tracking curved seams using ultrasonic sensors, commonly the same meth- ods are used as for electromagnetic sensors. In most applications, the ultrasonic sensor moves on a circuit path around the welding torch, and the seam direction is determined considering the sensor’s angular position, and the measured dis- tance values [37–42]. In order to increase the sensitivity and lateral resolution of non-contact ultrason- ic sensors, different methods are used. Zhang et al. [39] described a sensor sys- tem that uses high-frequency ultrasonics of 1.15 MHz to improve tracking accu- racy. In general, the width and wavelength of an ultrasonic beam increase as the frequency decreases. Lower frequencies correspond to poorer resolutions, but longer travel distances. Thus, for traditional applications of non-contact ultrasonic sensors, such as long distance measurement of large objects, low-frequency sen- sors are suitable. However, in weld seam tracking, small gaps, eg, in butt welding, are difficult to detect. On the other hand, unlike low-frequency ultrasonics, high- frequency ultrasonic signals deteriorate significantly in air. So the range of these signals is limited. To overcome these limits, the transmission efficiency of the high-frequency ultrasonic transmitter was improved and an adjusted transducer was developed to focus the beam [39]. The system shows a tracking accuracy of 0.5 mm. Another method for improving the tracking accuracy of non-contact ultrasonic sensor systems was presented [40]. The system uses the level of the reflected en- ergy from the workpiece surface to detect the welding seam. The ultrasonic trans- mitter operates with an ultrasonic frequency of 150 kHz, because at this fre- 4 Sensors for Process Monitoring294 Fig. 4.7-6 Principle of non-contact ultrasonic sensor quency the sensor is least sensitive to the arc noise in welding. For improving the resolution of the acoustic signal detected, and to eliminate the influence of noise from other than the scanned direction, a waveguide is used. This waveguide im- proves the sensor system’s tracking accuracy using several effects. On the one hand, it concentrates the receiver’s sensitivity to a small area, being positioned near it. On the other hand, a matched dimensioned waveguide acts as a resonator, and thereby increases the receiver’s sensitivity. In addition, it can be used as a fil- ter to attenuate interfering signals which arise from, eg, process noise. This sys- tem also shows a seam tracking accuracy of 0.5 mm. The major limitations in using non-contact ultrasonic sensors in welding are caused by the significant enlargement of the torch. The application of the sensor itself and its guiding mechanism leads to accessibility problems in welding small and complex structures. Further, for seam tracking on butt welds, there must be at least a groove of 0.5 mm depth and width that the sensor is able to detect. In general, ultrasonic sensors are distinguished by their simple configuration and low cost. The sensors are robust in harsh welding environments with arc- light, fumes, dust, and sputter, without any degeneration in sensing sensitivity. In addition to common industrial welding conditions, ultrasonic sensors are also suitable for underwater wet welding [41, 42]. 4.7.3 Welding Process-oriented Sensors Process-oriented sensors gain their information from the primary and secondary process phenomena. In this classification, primary process phenomena data re- lated to laser welding are the beam quality and the laser power. In arc welding, primary process phenomena are arc related and they can be acquired directly in the welding circuit, such as welding current and voltage. Arc welding sensors, for example, use this information. Secondary process phenomena data, on the other hand, are gathered by observation of the joining area while welding. From the radiation of the arc and welding pool and from the geometry of the joining area, information for torch positioning is generated. 4.7.3.1 Primary Process Phenomena-oriented Sensors Referring to the classification of welding sensors in Figure 4.7-1, the through-the- arc sensor is a primary process phenomena-oriented sensor. This sensor type uses the electrical characteristics of the welding arc in order to detect the distance be- tween the torch and workpiece surface. Generally, in arc welding the ohmic resis- tance in the welding circuit is closely related to the arc length. Depending on the welding process and the characteristic curve of the power source used, the weld- ing voltage or current is more influenced by this [43]. The general principle of a through-the-arc sensor in gas-metal-arc (GMA) weld- ing with consumable wire electrodes is shown in Figure 4.7-7. While the torch is 4.7 Welding 295