Introducing a Novel Infrared-Thermography-Based Control Method for Continuous Ultrasonic Welding of CFRPs

1. Introduction

Ultrasonic welding is an established joining process which already represents the state of the art in a wide range of applications and industries. An ultrasonic welding system converts alternating current into a mechanical, oscillating vibration, which is transmitted to the components to be welded via the sonotrode. The oscillating movement between the components generates heat on and within the components. At the appropriate moment, the heated components are consolidated under pressure, resulting in a material-bonded joint. Due to thermoplastic properties of fibre-reinforced high-performance polymers, this process can also be applied to relevant structural components in the aerospace industry. [17]

Regardless of the actual application, this process can be used in either a static or continuous execution [13]. Whilst spot welding has been largely understood and studied by numerous scientists ([4,5,16,19,20,22]), there is still a need to improve common understanding of the continuous process, particularly with regard to process control. First studies into process control for the continuous ultrasonic welding of CFRPs were initiated by a group of researchers at the German Aerospace Center (DLR). Janek et al. have investigated an alternative process control which introduces a constant welding power, while enabling the sonotrode amplitude as the steering parameter of the process. This process controller has been applied to three different material setups (stacking sequences of the CFRPs). It was then compared to the former state of the art controller, which applies a constant amplitude. Janek et al. found a more agile behaviour in the power controller, as the amplitude as a steering parameter was acting in a highly responsive manner. Janek et al. demonstrated that individual steering parameters, such as amplitude or power, exhibit distinct advantages, thereby supporting the implementation of a cascaded control architecture. [11]

Guided by this proposed concept, van Nierop and Jongbloed [12] have taken a significant step towards developing a combined process control method. Their novel control method combines constant-amplitude and constant-power setpoints, while allowing fluctuations in both to be driven by a fuzzy control logic. A symmetrical ratio of amplitude and power was used to progressively refine the system towards an asymmetrical fuzzy logic approach, which was subsequently applied for comparison with state-of-the-art controllers. A thorough investigation and analysis revealed that the combined amplitude-power control approach resulted in improved overall weld seam strength with a lower standard deviation, owing to a reduction in underwelded and, moreover, overheated regions of the weld seam. [12]

Whilst process parameters such as amplitude and power appear to be plausible control variables, ongoing research indicates that relying on these parameters will not be sufficient for future industrial applications. Taking into account system wear and drift, as well as variations inherent in off-the-shelf equipment, process parameters will not be identical across two different machines, setups and manufacturing facilities. Consequently, the authors suggest to identify an independent control variable.

Görick et al. investigated new input factors for an in-line quality assurance system for the continuous ultrasonic welding process. During their studies, various characteristics of the process were evaluated, revealing that the surface temperature of the top adherend correlates significantly with the strength of the weld seam. ([7,8,9])

Despite the conclusion reached by van Nierop and Jongbloed, that surface temperature provides only limited information, this short communication introduces surface temperature as an equipment-independent control variable, based on the preceding obersavtions by Görick. The authors intend that, for future applications, this control variable (temperature) can be calculated in advance using a simplified heat conduction simulation. Ultimately, temperature, as a control variable, can vary as a dynamic guide curve or mapping along a weld seam. Thus, it is independent of the welding equipment, the assembly line and manufacturing facility. [9]

2. Experimental Setup

The experimental setup of the continuous ultrasonic welding machine at the DLR is described in detail by Janek et al. Schematics of the main components are shown in Figure 1 including the top and bottom adherends, as well as the sonotrode and the pure matrix layer between the weld adherends. [11]

2.1. Material

A carbon fiber reinforced composite material provided by Toray® is used to investigate the temperature controller. This commercial material (designated as TC1225) consists of T700G standard modulus carbon fibres and a thermoplastic, semi-crystalline polymer called LowMelt-PolyArylEtherKetone (LM-PAEK). The matrix system is provided by Victrex® and eventually used as the pure matrix film in between the weld adherends as the energy dissipation mean, often referred to as Energy Director (ED). ([17,18])

The procured sheets are made from an unidirectional (UD) tape (stacking sequence $[45,90,135,0,45,135,0]s$), culminating to a sheet thickness of 1.96 mm and resized to the format of 296 mm x 104.6 mm. [11]

2.2. Infrared Thermography (IRT)

Measuring the surface temperature (as a derivative of the weld seam temperature) is accomplished by utilisation of a FLIR®A35 infrared thermography camera. This camera provides thermograms at a frame rate of 60 Hz at the size of 320 by 256 pixels through the GigE Vision interface. In order to make the thermograms serviceable for the controller, a region of interest is defined, which is located immediately behind the sonotrode in the area in front of the compaction unit. Within this area, a line parallel to the trailing edge of the sonotrode is selected, and an average value is calculated across 18 pixels in the joining zone. Eventually, this value is employed within the Programmable Logic Controller (PLC) of the welding machine, based on TwinCAT®Vision.

2.3. Temperature Controller

The surface temperature controller developed is based on a general approach for closed-loop control systems, see Figure 2. This requires the measurement of a sensor variable, which in this case is provided by the IRT camera. This sensor value, designated as the temperature $y_M\left(t\right)$, is subtracted from a setpoint temperature $w\left(t\right)$. The resulting difference $e\left(t\right)$ is fed to the controller. The controller acts accordingly and sends a correction signal in the form of a control variable, the sonotrode amplitude $u\left(t\right)$, which activates the actuator (welding generator) and acts upon the controlled system, which in turn is subject to certain external disturbances $d\left(t\right)$. Afterwards, the system response is recorded by the IRT camera and fed back to the controller in a closed-loop approach. ([1,15])

Table 1 provides the process parameters set to the adjustable variables, such as sonotrode force, consolidation pressure and weld velocity. The weld parameters have not been optimised, but are instead based on empirical knowledge. The controller was developed and fine-tuned during a series of preliminary welding trials. The remaining material was used to carry out as many replicate tests as possible. Hence, future studies should include proper parameter analyses in order to understand the influence of various variables.

2.4. Coupon Testing

For the purpose of investigation and evaluation two coupon tests are employed. At first a non-destructive inspection by applying water-coupled ultrasonic flaw detection facilitated with an OLYMPUS®OmniScan SX device, as described in detail by [11] Secondly, a destructive testing method to gather the Lap Shear Strength (LSS) as an established measure of weld seam strength and quality. ([6,17,19,20,23])

While ultrasonic flaw detection is applied to each continuous weld seam at full length, tensile testing is executed with individual test specimens. Therefore, each continuous weld seam is cut to ten individual Single Lap Shear (SLS) test coupons, as shown in Figure 3. Subsequently, each test coupon is manufactured to the size of 25.4 mm in width and 190.5 mm in length. Hence, an overlap length of 12.7 mm results accordingly. Eventually, LSS values are obtained with tensile testing on a Galdabini QUASAR 5 tensile-compression testing machine.

3. Discussion

3.1. Surface Temperature

The outset of the analysis is described by investigating the surface temperature as the main control parameter. Figure 4 illustrates the surface temperature extracted from the region of interest as described in Section 2.3. Two temperature curves are presented in a synchronised plot with the SLS coupon samples in ascending order. The graphs exemplary illustrate the consistent behaviour along the welding seam suggesting that the control algorithm functioned as intended. The temperature rises very rapidly as a response to the applied vibration, so that the target temperature is reached as early as the first test coupon. The temperature then oscillates at a low frequency and with an amplitude of approximately 15 °C around the desired setpoint of 165 °C.

This behaviour continues almost right up to the end of the weld, until it finally begins to decline again in the area right after the tenth sample in the outer region of the weld arrangement (see Figure 3). Weld trials No.20 and No.25 have been selected as these weld runs represent the extremes of the temperature control behaviour. No.20 consistently fluctuates around the desired temperature, whilst during weld trial No.25 a serious disturbance caused the controller to lose balance, resulting in a high peak temperature deviation. According to the authors’ observations, one reason for certain malfunctions is the PLC’s real-time capability. As indicated by black arrows in Figure 4, real-time capability, automatic shutter closure of the camera and related issues temporarily result in a lack of temperature measurements. Due to that downtime, the controller is delayed in trying to catch up with the actual temperature and state of the material system to be welded. For the time being, whilst the controller architecture was being developed, the authors sought to improve real-time performance by reducing the camera’s frame rate and by disabling the save function for the IR camera’s native thermograms.

3.2. Ultrasonic Flaw Detection

Water-coupled ultrasonic flaw detection is applied to assess the quality of the weld seams, as introduced by Janek et al. [11]. Hence, each weld trial performed using the final process parameters is scanned accordingly. In this short communication, ultrasonic C-scans are employed purely as a qualitative tool and measure for identifying similarities and differences between the individual weld trials.

Consequently, Figure 5 presents the C-scans of the weld tests examined, in ascending order from No. 20 to No. 28. If one compares the various scans from a global perspective, a common trend can be observed along the direction of the weld (in the positive X-direction), with common quality characteristics. The attenuation of the ultrasonic test signal is represented by a colour map, in which high attenuation, significant scattering and, consequently, poor joint quality are illustrated in blue. Low signal attenuation or a strong back-wall echo results in a red colour, which indicates a high-quality joint. Each weld is characterised by a marked increase of signal attenuation in the section between approximately 20 mm and 30 mm, followed by a relatively homogeneous, continuous weld up to the position x = 280 mm. This second section, which exhibits a high degree of continuity, is characterised locally by narrowings, some of which are more or less pronounced. The remaining section of each weld is characterised by a gradual weakening and variations associated with narrowings in the cross-sectional direction.

Assuming that the first and last sections are affected by the edges of the adherends to be welded, it seems plausible that these are of lower quality and exhibit deviations. The main section, on the other hand, exhibits extremely consistent weld seam quality, which suggests that a controlled surface temperature leads to consistent pressure-temperature conditions, thereby resulting in consistent weld seam quality. On inspecting the top surface after welding, the authors observed hot spots resulting in a change in gloss and reflections characteristic of the thermoplastic matrix. These hotspots, sometimes appearing as narrow lines running parallel to the direction of the weld seam, correspond to the narrowings visible in the C-scans, as shown in Figure 5. Given that such observations have been investigated ([2,3,14]) the results presented here lead us to believe that these hot spots are related to the mechanism of through-thickness heating by conduction and an unfavourable alignment of the adherends, rather than to the temperature-driven process control described. Further investigations need to include micrographic analysis in order to understand certain process relationships and correctly classify the origin of such effects.

In summary, it can be concluded that the temperature controller ensures a high degree of reproducibility in weld seam quality, as demonstrated by the qualitative analyses of the C-scans.

3.3. Tensile Shear Testing

The results of the single-lap shear tests are shown in Figure 6. These box plots illustrate the findings of the present study in comparison with the results for the amplitude and power controllers as published by Janek et al. [11]. Whilst the material and ply stacking are the same for the three different controller applications, slight differences, such as the welding speed, are adjusted for the temperature controller. Nevertheless, it is evident that process robustness and reproducibility are significantly improved when the temperature controller is used. This is supported by a reduction in the sample standard deviation $SS{D}_{abs}$ alongside a substantially higher mean lap shear strength value.

On closer examination of the LSS results of the temperature control, a certain trend becomes apparent. At sample #01 an overall reduced LSS value is attained. Subsequently, a plateau of comparatively high LSS values is established, characterised by individual outliers (indicated as fliers) and a median LSS that fluctuates slightly. These observations are consistent with similar behaviour observed in water-coupled flaw detection scans (see Figure 5) and the measured cross-sectional surface temperature (see Figure 4).

Evidently, sample #01 is joined at a lower average temperature due to the start of the welding process. Consequently, this results in a viscoelastic state that is not representative when compared with the other samples along the weld seam. Ultimately, this leads to reduced weld seam quality, as indicated by the C-scans, and lower weld seam strength, as quantified by the lap shear strength. Hence, a clear causal relationship is evident.

Outliers, whisker span widths and deviations between the remaining samples cannot be definitively concluded in this short communication. However, the authors believe that this fluctuation is mainly attributable to the unstable real-time capability of the PLC, as well as the nature of the fracture in each individual SLS specimen. The authors observed a variety of fracture characteristics, particularly as a result of the 45° orientation of the interface ply along the direction of the weld seam, involving different failure modes.

Consequently, further weld trials need to be carried out in order to comprehend the influence of ply stacking on LSS and its standard deviation within a set of samples.

4. Conclusions

The presented short communication introduce a novel process control method for the continuous ultrasonic welding of CFRPs. Initial investigations show that the controller effectively utilises the correlation between surface temperature and weld seam strength, translating this into consistent weld seam quality as indicated by homogeneous C-scans.

The authors believe that by using surface temperature as a reference variable, they have found a rational approach. Considering that no genuine benchmark against conventional control strategies (e.g., constant-amplitude or constant-power) was performed, the quantitative advantage of the newly proposed approach cannot be fully established. Nevertheless, the results reported in this short communication already indicate a discernible performance improvement.

Based on the low standard deviation, robust behaviour can be inferred, which yields reproducible results in replicated experiments. Building on this finding, further investigations must be carried out to determine the effect of selected parameters (sonotrode force, consolidation pressure, velocity, etc.) on the weld seam strength and its quality.