1. Introduction
Countries around the world are making efforts to reinforce regulations on emissions and expand the use of eco-friendly vehicles to reduce greenhouse gas (GHG) emissions in the transport sector1). In particular, there is growing interest in electric vehicles (EVs) among the eco-friendly vehicles. The core of EVs is batteries, and ultrasonic welding has been frequently applied to the bonding of battery cells2). Since welding defects in the manufacturing of EV batteries may involve battery performance degradation and fire (explosion) risks3), welding quality management is very important. For welding quality management, monitoring technology that can provide feedback by observing and analyzing phenomena and results that occur during the welding process is required. Quality monitoring in the ultrasonic metal welding process involves the following methods: a method of predicting quality by analyzing the data obtained using sensors that can measure energy, the vibration of the horn and anvil, and the temperature of the weld, which are variables related to the ultrasonic metal welding process, and a method of observing the weld and indentation depth after ultrasonic metal welding and then predicting quality through various mechanical evaluation results4,5). Data collection was very challenging for the ultrasonic welding process because welding is performed by vibrations of more than 20,000 cycles, but research has been conducted on various techniques as the development of science and technology made monitoring possible. This study aims to introduce technologies for the monitoring of the ultrasonic metal welding process.
2. Ultrasonic Welding Characteristics
Ultrasonic welding is a process that generates solid binding between two sheets by inducing oscillatory shear and applying pressure using frequencies higher than 20 kHz6). The ultrasonic welding process can be divided into four stages as shown in Fig. 1. In Fig. 1(a), two sheets are fixed under the application of a pressure among the process parameters. In Fig 1(b), heat energy is created through the oscillatory shear motion of the horn as well as the friction between the horn and the sheet, between the two sheets, and between the sheet and anvil. Bonding is performed through pressurization as shown in Fig. 1(c), and the ultrasonic welding process is completed as the horn is raised as shown in Fig. 1(d)7).
Fig. 1
The ultrasonic welding process divided into four steps, (a) clamping, (b) welding, (c) holding, and (d) unloading steps7)
The main process parameters of ultrasonic welding include the geometry of the horn/anvil, ultrasonic frequency, oscillation amplitude, pressure, welding time, and energy. The frequencies of ultrasonic welding generated by the transducer designed to operate at specific frequencies are 20, 30 or 40 kHz8).
Ultrasonic welding is generally classified into two categories depending on the applied material. As shown in the left figure in Fig. 2, it is used for plastic welding when the vibration of the horn is perpendicular to the upper sheet surface of the material. It is used for ultrasonic metal welding when the vibration of the horn is parallel to the upper sheet surface of the material as shown in the right figure in Fig. 2,9). When the vibration is applied in the parallel direction, the upper specimen is fixed and vibrated by the horn according to the amplitude while the lower specimen is fixed, thereby causing friction and heat. This is the bonding process caused by the pressure of the horn, which is favorable for metal welding10). In plastic welding, heat is obtained through the vertical vibration of ultrasonic waves between plastic molecules, and the heat causes the deformation of the base metal for welding11).
3. Ultrasonic Welding Process Monitoring Technique
3.1 Ultrasonic weld temperature measurement
For ultrasonic metal welding, it is known to be very difficult to measure the heat of the weld during welding due to the small weld size and very short welding time12). In addition, since ultrasonic metal welding is most effectively used in the field of bonding thin materials, such as thin sheets or foils13), it is very difficult to measure the temperature of the material using a thermocouple in ultrasonic metal welding.
Ren et al. measured the temperature of the weld during ultrasonic metal welding for Mg alloy AZ31B (thickness: 1.5 mm) and Ti alloy Ti6Al4V (thickness: 1.5 mm) by drilling a 0.5mm-diameter hole at 0.5mm below the surface of the Ti6Al4V alloy, which is the lower sheet, and inserting a thermocouple. They indicated that it was impossible to accurately measure the temperature even though the thermocouple was close to the weld as much as possible14).
Bakavos et al. located a thermocouple as close as possible to the weld of ultrasonic metal welding to measure the temperature. They could not measure the accurate value even though the temperature rose as the energy of ultrasonic metal welding increased, and reported that the temperature at the center of the weld is expected to be 100°C higher than the measured temperature15).
To accurately measure the heat generated in the weld, Zhao et al. processed the center of the anvil and inserted a modified thin film thermocouple as shown in Fig. 3. The modified thin film thermocouple was fabricated using a 0.4mm-thick Si wafer. The schematic of the modified thin film thermocouple is shown in Fig. 4. They developed a thermocouple that can be installed at the center of the weld and conducted research under the judgment that it is difficult to measure the temperature change of the material during the process using a conventional K-type thermocouple because ultrasonic metal welding is used in the field of bonding thin sheets or foils as aforementioned. They suggested that the newly developed heat measurement sensor is suitable for field process monitoring and welding system control16).
Shin et al. measured the heat around the horn, anvil, and material using an infrared (IR) thermal imager instead of a thermocouple for weld heat measurement. The IR thermal image was located as close as possible to the structure of ultrasonic metal welding to ensure the reliability of the measured temperature distribution. As shown in Fig. 5, the temperature distribution was different depending on the positions (top/bottom) of aluminum and magnesium during ultrasonic metal welding, and they reported that it was unclear whether there was a relationship between the temperature of the specimen and the tensile shear load17).
Schwarz et al. examined the temperature distribution by measuring 18 points (6 for the upper sheet, 10 for the lower sheet, and 2 for the horn) using an IR thermal imager. The 18 points are shown in Fig. 6. They measured the temperature for ten seconds for welding under 60 conditions through the design of experiments. The obtained temperature distribution curve showed that the temperature sharply rose within a short period of time during the ultrasonic metal welding process and it rapidly decreased when the horn was raised (Fig. 7). They also revealed that the temperature generated during ultrasonic metal welding is related to tensile shear strength, and that tensile shear strength tended to increase as the temperature during welding increased. They also reported that single data showed a high mean absolute error (MAE) when the tensile shear strength was predicted by performing machine learning based on the data for each element obtained through the experiment, but the MAE value decreased when machine learning was performed by combining various elements18).
Fig. 6
Top view of welded sheets with indicated cross section for microstructural examinations and clamping for tensile testing. The temperature measurement points on the sheets are indicated with black crosses4)
Fig. 7
(a) Temperature measurement results at measurement point T1 on the terminal. (b) Sonotrode temperature at measurement point S14)
As aforementioned, it is known to be very difficult to measure the heat of the weld during ultrasonic metal welding due to the small weld size and very short welding time16). Many researchers have increased methods to monitor the weld temperature by utilizing multiple IR cameras4), locating thermocouples close to the weld 15,17), and inserting them into the anvil18,19,20) to accurately measure the heat of the weld for ultrasonic metal welding.
3.2 Weld vibration measurement
Schwarz et al. measured the vibration of the anvil using a laser vibrometer. The 18 points are shown Fig. 8(a) shows the anvil vibration velocity under all conditions and Fig. 8(b) the tensile shear strength under each condition.
Fig. 8
(a) Anvil vibration velocity at 20 kHz coloured according to the weld quality. (b) Scatterplot of TSS over the mean vibration velocity extracted from the second process phase of the curves from (a)4)
The tendency of dispersion between the average vibration velocity and tensile shear strength can be seen, and they revealed that the Pearson correlation coefficient was 0.83. In addition, machine learning was performed using both temperature data and vibration data to improve the model performed based on the temperature data mentioned in section 3.1. As shown in Fig. 9, a lower MAE value could be secured compared to the machine learning model that used individual data4).
Haddadi et al. conducted a study to show the effect of the interface reaction on ultrasonic vibration behavior during the ultrasonic metal welding process using a laser vibrometer. As shown in Fig. 10, the vibration of the horn was measured by irradiating the laser vibrometer to the horn (sonotrode). Ultrasonic metal welding for dissimilar metals was performed using aluminum, galvanized steel, and non-plated steel. It was found that ultrasonic vibration behavior varied depending on the material combination, and the resulting change in the microstructure of the joint was observed21).
Guo et al. conducted research to find ways to reduce the false detection rate that wrongly detects good welded products as errors through the monitoring system used in the field to ensure the quality of aluminum copper ultrasonic welding of lithium-ion batteries. Data were collected using a linear variable differential transformer (LVDT) that measures the displacement of the horn, a watt meter that measures the power and frequency of the ultrasonic welding system, and a frequency sensor. They constructed the system as shown in Fig. 11. They proposed a new algorithm using the measured data, and reported that the false detection rate of the ultrasonic welding monitoring system could be reduced through the algorithm22).
In addition, an approach was developed based on Uncorrelated Multi Linear Discriminant Analysis (UMLDA), one of the machine learning techniques, to analyze the correlation between ultrasonic metal welding and the data measured with various sensors, including LVDTs, in the ultrasonic metal welding system used in the field. It was reported that data were secured through the preprocessing of the data extracted from the sensors, and that UMLDA, especially regularized UMLDA (R-UMLDA), could classify 100% of defects that occurred during ultrasonic metal welding with the data23).
3.3 Other monitoring techniques
Shao et al. explained the possibility of defects caused by the degree of wear of the horn and anvil, and determined the timing of replacement by measuring the degree of wear of the horn and anvil as a measure to ensure welding quality. They extracted the degree of wear of the tool by processing the profile of the tool as images through the weld of the ultrasonic metal welding specimen as shown in Fig. 12 rather than directly measuring the geometry of the tool. They proposed an algorithm for classifying the degree of wear of the horn and anvil by applying leave-one-out cross-validation (LOOCV) to classify the tool condition using the collected data, and reported that excellent welding quality in product production can be ensured through the monitoring system5).
Nazir et al. performed machine learning using the fact that the energy consumption of the welder and the noise generated during the process vary depending on the degree of wear of the horn and anvil in the ultrasonic metal welding process. As shown in Fig. 13, Four tool conditions for the states of the horn and anvil (new/new, new/worn, worn/new, and worn/worn) were presented. Data were collected for 50 welding conditions. After preprocessing them, a machine learning classification model was developed. When machine learning was performed using the individual data for each sensor, the validation value was 99.0% for the displacement sensor, 83.5% for the energy sensor, and 86.5% for the sound sensor. When machine learning was performed by combining the three sensors, however, the accuracy (99.5%) was close to 100%24).
In this study, various ultrasonic welding quality monitoring technologies were introduced. Based on the review of various monitoring technologies listed above, the advantages, disadvantages, and field applicability of each monitoring technology are summarized in Table 1.
Table 1
Comparision of the advantages, disadvantages, and field applicability for various ultrasonic welding monitoring technologies
4. Conclusion
In this review paper, various ultrasonic metal welding quality monitoring technologies were introduced. It was found that the majority of studies were conducted on monitoring technologies through the measurement of the heat generated during the process, but there were few studies on quality prediction that analyzed and applied the measured temperature data. On the other hand, studies have been actively conducted to predict quality by measuring and monitoring the vibration generated during the process of the horn and anvil. To ensure the quality reliability of ultrasonic metal welding, it is necessary to perform quality monitoring by accumulating data through various sensors and analyzing them rather than using single sensor-based monitoring techniques.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Print
