Device for process monitoring during resistance spot welding and resistance projection welding

Abstract

The present invention relates to a device for process monitoring during resistance spot welding and / or resistance projection welding, comprising at least one electrode (1) for resistance spot welding and / or resistance projection welding of at least one workpiece (4) and at least one current / voltage source (6) connected to the at least one electrode (1). Furthermore, at least one detection unit (2) is provided, which is designed to detect at least one airborne sound parameter during welding. Furthermore, at least one evaluation unit (5) is provided, which is designed to decompose the detected airborne sound parameter into a continuous spectrum (8) by means of a Fourier transformation, wherein the evaluation unit (5) is designed to compare the continuous spectrum (8) with a template signal (7), whereby a defect pattern can be clearly determined.

Description

[0001] The present invention relates to a device and a method for process monitoring in resistance spot welding and resistance projection welding.

[0002] Projection welding is often performed using capacitor discharge (CD). This CD welding technique is not controllable. There is no standardized method for non-destructive testing and process monitoring (PÜ) for projection welds. Therefore, non-destructive and contactless process monitoring is desired in industrial environments, capable of detecting fluctuating quality and even rejects while maintaining constant machine parameters.

[0003] For CD projection welding, process monitoring primarily involves measuring sound emissions (structure-borne sound) at the electrode using piezoelectric sensors, and then determining a correlation between the strength of the joint and the sound signal. Similar approaches exist for other pressure welding processes. For example, in resistance spot welding, the structure-borne sound in the ultrasonic range is also evaluated during the process. Furthermore, a correlation between the airborne sound signal and the strength is determined. In this case, a microphone is used next to the process to detect the emitted airborne sound.

[0004] A disadvantage of using piezoelectric sensors in the electrodes is that they require modification, which can lead to problems in a highly automated process. Accessibility to the electrodes in question plays a particularly crucial role, as significant integration effort is required to implement process monitoring. Furthermore, the sensors used are expensive, and precise determination of the defect pattern is not readily possible.

[0005] For example, KR 101 859 604 B1 discloses a projection welding device capable of detecting defects in real time. Ultrasonic sensors arranged on an inclined surface of a rotor are used. The sensors are capable of examining defects on the surface and inside the object being inspected by detecting and evaluating the sound reflected from the surface. As already mentioned, these devices have the disadvantage that the corresponding sensors must first be integrated, which is very complex.

[0006] The present invention is therefore based on the object of monitoring the welding process in real time, whereby the costs for a device and a method for process monitoring in resistance spot welding and / or resistance projection welding are to be reduced.

[0007] This problem is solved according to the invention by a device and a method according to the independent claims. Advantageous embodiments and further developments are described in the dependent claims.

[0008] A device for process monitoring during resistance spot welding and / or resistance projection welding comprises at least one electrode for resistance spot welding and / or resistance projection welding of at least one workpiece and at least one current / voltage source connected to the at least one electrode. Furthermore, at least one detection unit is provided, which is designed to detect at least one airborne sound parameter during welding. Furthermore, at least one evaluation unit is designed to decompose the detected airborne sound parameter into a continuous spectrum using a Fourier transformation, wherein the evaluation unit is designed to compare the continuous spectrum with a template signal, whereby a defect pattern can be clearly determined.

[0009] A "template signal" or "reference signal" is understood to be a signal that was recorded during a faultless weld, i.e. the template signals comprise the reference values of a faultless or defect-free weld, which are compared with the signals recorded later in order to identify any deviations and derive clear defect patterns from them. The template signals are stored in a memory unit of the evaluation unit. The advantage of this invention is that no measuring equipment is required to record structure-borne noise. The easy integration into existing systems results in particular from the simple and uncomplicated design of the invention. Despite the simple installation, it is possible to detect and evaluate the detected airborne noise in real time in a non-contact and therefore non-destructive manner.In this case, the Fourier transformation can be, in particular, a fast Fourier transformation (FFT) or a discrete Fourier transformation. The at least one electrode and the at least one current / voltage source can be controlled by a control unit, which typically also includes the detection unit and the evaluation unit or is at least (electrically) connected to them, so that control signals and / or data can be communicated between the individual units.

[0010] Furthermore, the detection unit can be designed as a microphone, in particular as a condenser microphone. By using a microphone, airborne sound in the audible frequency range, i.e., 1 Hz to 1 MHz, can be detected. In particular, inexpensive condenser microphones can be used here to reduce costs. The use of the invention in an automated production environment can thus be further simplified. Furthermore, the detection unit is typically not arranged in direct, i.e., immediate contact with the workpiece to be welded.

[0011] In addition, the defect pattern can include the coating composition, the base material of the component to be welded, the geometry of the welded component, deviations in the welding parameters, the positional deviation of the component during welding, component damage, operating errors, or the degree of component contamination. The comprehensive options for classifying and determining the signals allow a wide variety of defect patterns to be distinguished, enabling precise determination. Differences are formed between the continuous spectrum and the template signal. From these, characteristic signals can be identified, and thus the defect patterns can be derived.

[0012] In addition, the template signal can comprise at least one arithmetic, geometric, or harmonic mean of at least one detected airborne sound parameter from at least three defect-free welds, preferably from at least 10 defect-free welds, as well as an upper limit and a lower limit for each of the at least one detected airborne sound parameter. By using mean values as well as upper and lower limit values, value ranges can be defined within which the weld is classified as "defect-free." This makes it possible to determine or specify a tolerance range individually for each airborne sound parameter, so that monitoring can be adapted to the respective welding task.

[0013] Furthermore, the upper and lower limit values can be defined as at least one standard deviation, preferably at least twice the standard deviation, and particularly preferably at least three times the standard deviation of the respective mean value of the detected airborne sound parameter. This means that a tolerance range is defined by the upper and lower limit values. By selecting the appropriate standard deviation, the monitoring can be individually adapted to the welding task at hand.

[0014] Additionally, the airborne sound parameter can include sound pressure or sound velocity. Using these parameters, measurements can be performed easily. Both parameters can be recorded and processed simultaneously.

[0015] In a method for process monitoring during resistance spot welding and / or resistance projection welding, at least one current / voltage source supplies at least one electrode with electrical current, thereby welding at least one workpiece. During the welding process, at least one detection unit detects at least one airborne sound parameter, and at least one evaluation unit subsequently decomposes the detected airborne sound parameter into a continuous spectrum using a Fourier transformation. The evaluation unit then compares the continuous spectrum with a template signal, thereby unambiguously determining a defect pattern.

[0016] The main advantages are that it proposes a cost-effective method for non-destructive, contactless, system-integrated, and real-time process monitoring. By comparing the signal with one or more template signals from sample welds without process defects, deviations or anomalies can be detected, allowing conclusions to be drawn about the existing process defects. Furthermore, it is also possible to establish a correlation between the sound signal and the strength of the joint. Pattern recognition optimized by machine learning could also be easily integrated.

[0017] In addition, the detection unit can record at least one airborne sound parameter during the activation phase and the bonding phase of the welding process. In the "activation phase," current flows through the electrode, causing a projection to heat up (typically, two separate electrodes are used, separated from the workpiece, but the workpiece itself can also serve as one of the electrodes). The workpiece to be welded has not yet bonded. In the "bonding phase," the bonded connection is established. By limiting the detection process to the activation and bonding phases, the detection process is focused, thus avoiding interference signals that may occur in other phases of the welding process.

[0018] In addition, the duration of the welding process can range from 2 ms to 10 ms. By shortening the welding process, the overall process duration can be reduced, thereby increasing process efficiency and reducing costs.

[0019] Furthermore, the error patterns can be identified using machine learning, enabling further automation of process monitoring. Algorithms can be trained to subsequently detect deviations and assign the signals to error patterns.

[0020] Furthermore, a computer program product with a computer program comprising instructions or software means for carrying out the described method and / or for controlling the described device when the computer program is executed in an automation system such as a computer can be provided.

[0021] The method is designed to be carried out with the described device, ie the device is suitable for carrying out the method.

[0022] Embodiments of the device and the method are shown in the drawings and are described below with reference to Fig. 1 - 3. Recurring features are provided with identical reference symbols.

[0023] They show: Fig. 1 a schematic drawing of a device for resistance projection welding with a detection unit and an evaluation unit, Fig. 2 exemplary curves of a template signal and a measurement signal of a faultless weld according to the FFT and Fig. 3 exemplary curves of a template signal and three measurement signals of faulty welding after the FFT.

[0024] Fig. 1 depicts a device for resistance projection welding, comprising two electrodes 1 arranged opposite one another. A projection part 3 is arranged between the electrodes 1 as part of one of the electrodes 1, which in turn is in contact with a sheet metal workpiece 4. The electrodes 1 are electrically connected to a current / voltage source 6. Furthermore, a detection unit 2 with an evaluation unit 5 is arranged outside the device for resistance projection welding. These units have no mechanical contact with the welding device and are spaced apart from it. The detection unit 2 together with the evaluation unit 5 can be connected to a control unit, or the control unit can comprise the detection unit 2 and the evaluation unit 5.The control unit typically also controls the electrodes 1 and can be designed as a computer on which a computer program product for carrying out the method described below can be executed. In the embodiment shown in . Fig. In the embodiment shown in Figure 1, the detection unit 2 is designed as a condenser microphone.

[0025] Process monitoring is enabled by the illustrated embodiment through the evaluation of airborne sound, i.e., in a frequency range from 1 Hz to 1 MHz. For projection welding, the evaluation of the activation and bonding phases may be of particular interest. A fast Fourier transformation (FFT) is used to decompose the acoustic signal or the detected airborne sound parameters into continuous spectra.

[0026] Sample welds are then used to create a template signal. This is compared with the welds to be monitored. In this example, the template signal comprises the mean values of the sound pressure and the sound velocity from 10 defect-free (sample) welds, as well as an upper and a lower limit for each of the sound pressure and the sound velocity. The upper and lower limits of the two airborne sound parameters are calculated from twice the standard deviation of the sample welds. This defines a value range in which the weld should ideally run without the presence of a defect being output, i.e. if the measured values for the airborne sound parameters are within this range, the weld is classified as "defect-free".

[0027] As soon as the detected airborne sound parameters exhibit values that lie outside the error-free range, a defect has occurred. Depending on the type of deviation of the detected airborne sound parameters from the template signal, various defect patterns can be distinguished. These include, among other things, the coating composition, the base material of the component to be welded, the geometry of the welded component, the deviation of the welding parameters, the positional deviation of the component during welding, component damage, incorrect conditions, or the degree of component contamination.

[0028] This means that with the help of the present invention, it is not only possible to determine whether a fault exists, but also what type of fault it is. This allows the source of the fault to be directly identified, which in turn allows it to be rectified quickly and easily. This helps avoid long downtimes of the system.

[0029] Monitoring can also be performed using machine learning to detect additional patterns (error patterns) in the signal that are undetectable by humans, i.e., in the continuous spectra of the detected airborne sound parameters, thereby further improving the precision of monitoring. This also increases the degree of automation.

[0030] The classification can be performed using binary classification (i.e., divided into flawless and flawed welds) or multi-classification (differentiating into a suitable defect pattern). Logistic regression, super vector machines, k-nearest neighbors, decision trees, random forests, naive Bayes, neural networks, gradient boosting, and similar methods can be used to perform the classification. Categorized data obtained from flawless test welds is used to train each method.

[0031] Since the welding process lasts only a few milliseconds (2 ms to 40 ms), fast and precise detection of potential defects is necessary to ensure a high-quality process. This is ensured in the present invention by analyzing the airborne sound with a condenser microphone.

[0032] The use of a non-contact detection unit 2 eliminates the need for measuring instruments for detecting structure-borne noise, which would require complex integration into the welding device. Cost-effective condenser microphones are sufficient to detect airborne noise in the audible range. This simple design also allows for use in automated production. Furthermore, the evaluation is real-time capable and, thanks to the non-contact and non-destructive detection of airborne noise signals, can be easily integrated into production.

[0033] Thus, in addition to process monitoring (correlation of the sound signal and strength of the joint), numerous types of defects can also be detected under the same machine conditions with the present invention.

[0034] Fig. Figure 2 shows a template signal 7, which was determined in advance as the mean value of ten flawless welds, with a tolerance range 10 formed by an upper limit and a lower limit, whereby in this exemplary embodiment, twice the standard deviation from the mean value described above is used. In addition, a continuous spectrum 8 of a detected airborne sound parameter according to the fast Fourier transformation (referred to here as the first measurement signal), which was recorded during the welding of a galvanized weld nut as workpiece 4, is shown. It can be clearly seen here that the course of the template signal 7 and the course of the first measurement signal have a similar course. In addition, the first measurement signal lies predominantly within the tolerance range 10, so that the weld during which the first measurement signal was recorded can be classified as flawless.

[0035] Fig. Figure 3 shows, in addition to the template signal 7 and the associated tolerance range 10, three continuous spectra curves of three different welds. The dotted curve of the recorded spectrum 8a of the second weld is referred to as the second measurement signal, the dotted and dashed curve of the recorded spectrum 8b of the third weld is referred to as the third measurement signal, and the dashed curve of the recorded spectrum 8c of the fourth weld is referred to as the fourth measurement signal.

[0036] The first thing to note is that all of the measurement signals lie outside the tolerance range 10 at several points. This means that at certain frequencies, all of the measurement signals have amplitudes that are not within the tolerance range 10 of the template signal 7. Thus, each of the displayed measurement signals is categorized as faulty. Furthermore, significant differences in the curve can be observed, for example, between spectrum 8b and spectrum 8a. These differences can be traced back to the type of defect present and allow different defect patterns to be directly classified. Each of the displayed spectra 8a, 8b and 8c represents a unique defect pattern by way of example, so that not only can a distinction be made between fault-free and defect-free welds, but the source of the defect can also be directly determined. QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] KR 101 859 604 B1

[0005]

Claims

[1] Device for process monitoring during resistance spot welding and / or resistance projection welding, comprising: at least one electrode (1) for resistance spot welding and / or resistance projection welding of at least one workpiece (4), at least one current / voltage source (6) connected to the at least one electrode (1), at least one detection unit (2) which is designed to detect at least one airborne sound parameter during welding, and at least one evaluation unit (5) which is designed to decompose the detected airborne sound parameter into a continuous spectrum (8) by means of a Fourier transformation, wherein the evaluation unit (5) is designed to compare the continuous spectrum (8) with a template signal (7), whereby a defect pattern can be clearly determined. [2] Device for process monitoring during resistance spot welding and / or resistance projection welding according to claim 1, characterized by that the detection unit (2) is designed as a microphone, in particular as a condenser microphone. [3] Device for process monitoring during resistance spot welding and / or resistance projection welding according to one of the preceding claims, characterized by that the defect pattern includes at least the layer composition, the base material of the workpiece (4), the geometry of the welded component, the deviation of the welding parameters, the positional deviation of the workpiece (4) during welding, workpiece damage, incorrect operation or the degree of workpiece contamination. [4] Device for process monitoring during resistance spot welding and / or resistance projection welding according to one of the preceding claims, characterized bythat the template signal (7) comprises at least one mean value of at least one detected airborne sound parameter from at least 3 faultless welds, preferably from at least 10 faultless welds, as well as an upper limit value and a lower limit value of the at least one detected airborne sound parameter. [5] Device for process monitoring during resistance spot welding and / or resistance projection welding according to claim 4, characterized by that the upper limit value and the lower limit value are formed by at least the single standard deviation, preferably at least the double standard deviation, particularly preferably at least the triple standard deviation of the respective mean value of the detected airborne sound parameter. [6] Device for process monitoring during resistance spot welding and / or resistance projection welding according to one of the preceding claims, characterized bythat the airborne sound parameter includes the sound pressure or the sound velocity. [7] Method for process monitoring in resistance spot welding and / or resistance projection welding, wherein at least one current / voltage source (6) supplies at least one electrode (1) with electrical current, so that at least one workpiece (4) is welded and at least one detection unit (2) detects at least one airborne sound parameter during welding and subsequently, at least one evaluation unit (5) decomposes the detected airborne sound parameter into a continuous spectrum (8) by means of a Fourier transformation and then the evaluation unit (5) compares the continuous spectrum (8) with a template signal (7) and This allows a fault pattern to be clearly determined. [8] Method for process monitoring in resistance spot welding and / or resistance projection welding according to claim 7, characterized by that the detection unit (2) records the at least one airborne sound parameter during the activation phase and material bonding phase of the welding process. [9] Method for process monitoring in resistance spot welding and / or resistance projection welding according to one of claims 7 and 8, characterized by that the duration of the welding process is in a range of 2 ms to 10 ms. [10] Method for process monitoring in resistance spot welding and / or resistance projection welding according to one of claims 7 to 9, characterized by that the error patterns are recognized by machine learning.

Images