1. Introduction
More holistic integrity assessments involving integrated non-destructive testing are becoming more widely adopted with greater emphasis on a connected manufacturing philosophy via Industry 4.0. Such initiatives are now enabling manufacturers to analyse those welding characteristics and determine whether the integrity of the weld and its heat-affected zone (HAZ) satisfies the defined quality standard. Fusion welding may be the defined requirement or only solution available for fabricating complex assemblies commonly found in the production of safety-critical structures deployed within the nuclear and aerospace sectors, and as such, its integrity and thus its quality are of paramount importance for such structures to achieve design performance. Present inspection methods can be time-consuming, destructive, and often not representative of the whole part. With the increasing fidelity of non-destructive evaluation (NDE) methods through the development of improved sensor technology, it is possible to monitor the weld quality in situ and in real time, thereby offering cost-effective solutions that contribute to improved sustainability and mitigate rectification complexity, which further reduce the assigned non-value-added effort that is unfairly placed on the inspection process.
Through the integration of various acoustic emission (AE) sensors within the mechanised MAG welding set-up, utilising an eight-axes robotic system and a Fronius TPS 500i CMT™ welding power source, experiments were completed using a “bead-on-plate” configuration to detect emitted elastic waves discharged during material solidification and cooling. Any change in the solidification and cooling noise patterns was investigated to identify anomalies or gross imperfections within and around the thermally disrupted material. This is to correlate any modifications in noise frequency at any point in time, with specific microstructural analysis at that time interval, utilising recorded soundwave signals of both airborne and contact ultrasonic emissions. This also included monitoring any changes in amplitude and energy, throughout the event duration, which may potentially lead to exploiting opportunities to create autonomous welding and inspection systems, based on enhanced decision-making protocols.
1.1. Fusion Welding Characteristics Involving Arc Welding
Whilst fusion welding is recognised as being a major and complex manufacturing engineering discipline, it nonetheless needs to be recognised as a method for indicatively inducing flaws into structures, even if the weld and HAZ regions are classified as 100% dense. This is because a fusion weld induces significant phase changes, as a result of converting the substrate and any filler metal into a liquid state that subsequently solidifies non-isothermally, forming a cast structure, encouraging varying levels of compositional variance (segregation). This conversion locally modifies the component material’s original condition of supply. This is where the engineering community would clarify the original base metal structure as now being “flawed”.
There are several key conditions that can significantly contribute to morphing such conditions into flaws or even defects, some examples of these are as follows:
-
Material chemistry combinations and microstructures;
-
Filler metal chemistry (if used);
-
Geometry and thickness;
-
Heat source power density;
-
Heat flow/thermal cycle;
-
Gas–metal reactions.
The constituency and complexity of a material’s chemistry can have a substantial effect on its mechanical and geometrical response to welding, e.g., hardenability and toughness, and distortion, respectively, and in some instances, its sensitivity to forming low-melting point compounds and brittle intermetallic phases. High-strength steels are prone to hydrogen cracking if their hardness values are >350 HV, austenitic stainless steels are susceptible to solidification cracking if they contain no or very low Delta ferrite contents, or if their chromium equivalent to nickel equivalent ratios are below 1.6
[1]. Nickel-base alloys are prone to micro-fissuring in the weld and HAZ when significant compositional gradients exist
[2].
The profile or shape of the weld, the type of joint used, the volume of liquid metal produced, and the type and magnitude of restraint all contribute to the level of strain, shrinkage, and residual stress encountered within the final weld.
Power density, also known as heat source intensity, is a factor that has a major effect on both the heating and cooling rates encountered within the weld and the size of the isotherm, extending axially and perpendicularly to the direction of heat source travel. In this study involving the MAG process, the typical values produced were in the range between 1 × 10
4 and 5 × 10
4 W/cm
2 [3].
The weld’s thermal cycle is also a critical parameter that is dependent upon three key variables:
The quotient value between that of welding power and travel speed determines the amount of energy delivered per unit length and is always required in formulating a weld procedure specification. This is due to it being a key parameter in qualifying a weld’s mechanical properties, specifically impact toughness and hardness.
These interrelating heat input variables are known to affect the final transformation products. This is especially so when considering allotropic materials wherein metastable phases such as martensite and bainite are predominantly produced when welding high-strength steels and low-alloy steels. The evolution of such phases associated with welding these types of materials results from rapid cooling through the austenite region (800 °C and 500 °C), also referred to as the Δt
8−5. Dixon and Håkansson
[4] researched the effect from increasing the welding heat input between 1 and 5 kJ/mm and reported a notable difference in hardness.
Gas–metal reactions used within this context concern those chemical reactions that take place at the interface between the gas phase and the liquid metal. Such reactions involve the dissociation of carbon dioxide and the dissolution of nitrogen, oxygen, and hydrogen. Gas–metal reactions are complex and known to affect the enthalpy of the welding arc and the level of interstitial elemental pick-up, which can adversely affect the response to embrittlement and porosity in the weld and HAZ region
[5][6]. The combination of gas mixtures and varying material chemistries within the weld pool were chosen to accentuate the exothermic reaction to create insoluble compounds that promote the onset of porosity or cracking.
In summary, several variables need to be considered in determining a successful weldment. The effects arising from the interrelationship between heat input, cooling rates, shrinkage, and weld metal chemistry are major influences on cracking sensitivity during fusion welding. Accentuating these interactions to induce flaws was the main driver within this research to create an acoustic signature that can be remotely detected.
1.2. Welding Imperfections and Non-Destructive Evaluation (NDE)
Previous research
[7] into the use of acoustic emission to determine the arc stability of the gas–metal arc welding process has identified the potential use of this technology to perceptibly detect instabilities in the arc that could offer the potential of locating defects in a welded foil. Furthermore, the use of acoustic emission techniques to evaluate hydrogen cracking in slow strain-rate tensile (SSRT) tests in various hydrogen-charged specimens proved that cracking could be estimated and monitored rather successfully from the total acoustic emission (AE)
[8].
The work presented here focuses on the weldability response of carbon–manganese (C–Mn) steel utilising a fully austenitic stainless steel filler metal and a combination of varying alloy inserts.
The main imperfections commonly found when welding these types of steels are as follows:
Such imperfections normally found in a weld or its HAZ can be deemed innocuous under certain operating conditions defined by the design and quality assurance standard, hence the term “imperfection” rather than “defect”. However, if certain conditions arise, then these conditions may induce a growth in regions of higher stress intensities, which can ultimately result in failure. This work encompasses studies aimed at inducing “cracking” within the weld, which at the same time provides energy emissions for exploitation by an NDE technique. The chosen NDE technique this research is investigating involves the application of AE sensor technology to detect crack formation in a weld. Cracks are classified under two distinctive headings: hot cracking, also known as solidification cracking and cold cracking or hydrogen-induced cracking.
Solidification cracking occurs during the initial stages of solidification and are commonly associated with the formation of low-temperature interdendritic phases and the shape of the weld incapable of withstanding the terminal solidification stresses. The termination stresses and corresponding cracking sensitivity are further accentuated in joints that are highly restrained and of large cross-sectional thickness. Thin weld beads are more sensitive to cracking when the alloy contains volume fractions high in sulphur, Phosphorus, and carbon, and when the depth-to-width ratio of weld promotes a horizontal solidification profile. Prevention methods that must be considered when adopting arc welding technologies are as follows:
-
Selection of appropriate filler wire size and welding parameters;
-
Consumables with low carbon and higher manganese and silicon;
-
Fusion profile that promotes depth-to-width ratio no greater than 2:1;
-
Weld pool surface profiles;
-
Avoidance of significant gaps;
-
Removal of surface contaminants.
Hydrogen-induced cracking (HIC) is more common in C-Mn and low-alloy steels where the hardness values are >350 HV, and occurs when four factors have been simultaneously satisfied, these being microstructural sensitivity, highly stressed joints, increased hydrogen enrichment and an Mf temperature near to normal ambient.
-
Microstructural sensitivity is due primarily to the HAZ having a hard and brittle microstructure or if the martensitic finish temperature (Mf) is below normal ambient temperature.
-
Significant tensile stresses arising from high restraint and steep cooling curves, or a high yield strength filler metal used to join a lower yield strength substrate.
-
Hydrogen sources arising from decomposition of hydrocarbons and moisture sources, e.g., flux-based welding processes.
-
A temperature near to ambient that retards the rate at which nascent hydrogen diffuses to a position where the above three factors overlap, causing the coalescence of hydrogen molecules (gas) to induce a significant hydrostatic pressure.
The HAZ microstructure is more liable to result in failure from hydrogen cracking due to low-heat inputs, giving rise to steep cooling curves, which in turn simultaneously increases hardness and tensile stress. Solutions to overcome this problem involve using higher heat inputs or pre-heating the substrate.
In a weldment, HIC is more prone to occur at the toe of weld (the junction between the fusion zone and near-HAZ), due to a combination of geometrical change and higher hardness experienced in the HAZ region. Many anomalies identified can be detected using existing NDE techniques but, as previously mentioned, AE sensors are particularly sensitive to detecting such phenomena and offer significant benefits to monitor time-delayed cracking. Another method assessed for inducing flaws within a weld was undertaken through introducing the fusing of non-compliant materials to stimulate cracking responses during welding. In the work completed by Ser’eznov et al.
[9], titanium and aluminium inserts were fused within the substrate to cause brittle intermetallic phases within the weld pool to promote a cracking tendency. Such flaws caused an increase in the amplitude and total count of AE signals, characterising the process of nucleation and development of cracks in the solidified weld.
Standard fusion welding of alloys (non-pure systems) indicatively produces a columnar grain structure that results from heterogeneous nucleation promoted through epitaxial growth. This mechanism promotes the direction of crystal growth from the substrate–liquid interface through to the formation of dendrites within the liquid. Within this thermodynamically controlled mechanism, it is possible that steep compositional gradients are created that incubate the formation of intermetallic and low-melting point compounds, which under a favourable stress level rupture during the cooling phase, and it is here where the emission of acoustic signals would exist.
1.3. Acoustic Emission (AE) Sensors
AE alters the shape of piezo-electric material, which in turn alters the respective electrical signal. This is due to the elastic waves deforming the atomic structure, which is further refined using specialist amplification electronics/measuring equipment. The analysis of AE signals can help to understand the details in respect to the origin of the defect/discontinuity in the material.
AE is different from most other sensing technologies for two reasons:
-
Instead of transmitting energy to the material being examined, it simply records emitted energy over a defined threshold. Weld solidification provides enough energy for generating elastic waves that are above such defined thresholds.
-
AE testing is concerned with changes in material discontinuities or density, as it provides the detection of active features like crack growth or spot indication, respectively, which can be used to determine the quality of the weld.
AE emissions can be produced from crack initiation and growth in a metal during dislocation and slip movements or phase transformations in allotropic metals. AE detection takes place mostly when the material under load exhibits plastic deformation or is loaded at the onset of its yield point. The presence of cracks in a metal implies that the level of stress at the crack tip is many times higher than the surrounding area; therefore, AE can provide distinguishing features when the material along the crack tip undergoes plastic deformation. The output scale of energy produced by AE is in the form of waveform amplitude, velocity, and signal attenuation (due to damping effects). The stress intensity factor, load at the crack tip, and material modulus will then affect the crack growth rate, which is proportional to the acoustic emission amplitude.
2. Investigating Sensor Technologies Applied to Welding Quality
Zhang et al.
[10] investigated image-based acquisition systems utilising charge couple devices (CCDs) to monitor weld pool profiles within the keyhole mode of the plasma arc welding (PAW) process. During the PAW process, the keyhole status could be characterised by indirect signals (generated when the keyhole is formed) and direct signals (keyhole shape and size is maintained). However, all the detected information could only be used to demonstrate the establishment of an open through-thickness keyhole but could not give its dimensional information, the weld pool size, or any volumetric imperfections. Rosado et al.
[11] investigated a new non-destructive testing (NDT) system that focused on the application of a new ionic eddy current method to assess the evolution of micro-size superficial defects in metallic joints. Another addition is plasma optical spectroscopy by Ferrara et al.
[12], which is also an advancement in weld monitoring, where the emitted plume provides indications of the weld quality; however, there is little publicised literature available suggesting barriers to industrial practicalities. Yi et al.
[13] carried out trials using the AE system applied to assessing the metal droplet transfer characteristics within the MAG process and Kamal et al.
[14] compared the arc sound of a pulsed MAG sequence at various process parameters and found that the welding arc sound was strongly related to both metal droplet transfer and weld quality, which enhances the ability to detect welding stability and defects. Paoletti et al.
[15] investigated the impact of material transform parameters influencing the joint quality. Spot bonding using the friction stir technique (FSSB) applied to polycarbonate sheets was monitored through torque signals via piezo-electric dynamometers distinguishing abusive from normal conditions. In addition to this, temperature (infrared radiation) could be extracted from an infrared pyrometer. Eriksson et al.
[16] created a method to characterise imperfections during laser welding employing a mix of sensors that utilised spectrometry, photodiode analysis, and visible imaging; such techniques can lead to numerous physical features being quantified. Sibillano et al.
[17] investigated a technique for on-line identification of joint penetration utilising spectroscopic analysis of the optical emission collected from the laser–metal interaction zone. From examining the optical emission from the connection zone between the laser and metal, the resulting spectra reveal unique signatures of different phenomena. Guokai et al.
[18] also carried out studies using a CCD camera and an infrared camera to visually monitor the weld pool. The dynamic dimensional transformations that occur in the weld pool/keyhole for short periods were obtained with confidence. Suárez et al.
[19] explored different avenues using fibre Bragg grating sensors during tungsten inert gas (TIG) welding of aluminium–magnesium alloy plates, where tests found transient and residual strains located within the HAZ. Their view was that, once detected, distortion would be actively decreased to ensure the process is corrected in situ.
3. Acoustic Emission Sensors Applied to Monitoring Weld Quality In Situ
Ladislav et al.
[20] carried out a feasibility study on acoustic signals for the on-line monitoring of the MAG welding process using the short circuit (dip transfer) mode. Microphones and PZT sensors were used to detect the responsive behaviour of any weld defects by using different inserted steel materials. The study concluded that AE sensors can be used for detecting short circuits. The measured signals from the PZT sensor showed complexity and it was not possible to find any direct relation with welding current as the distance of the microphone from the weld surface induced a delay when reading the results.
AE produced by metal droplet transfer gives a very strong signal for monitoring. From measuring metal drop transfer using AE signatures, the authors of
[9] were able to measure the following phenomena: flow of the molten pool, microstructure phase changes, liberation of internal tension, dilatations, and plastic deformation. The AE study opens up the possibilities of locating defects within the welded joint during the cooling phase where microstructural phase changes and plastic deformations were identified.
The measurements found in
[20] of the arc acoustic waveforms and welding currents within the short circuit transfer mode were completed using two arc voltage settings: 19 and 21 Volts. Airborne acoustic emission measurements were completed by a microphone and a contact PZT sensor positioned at a fixed distance of 350 mm from the welding arc. During the tests, peaks in acoustic pressure were observed by the airborne microphone (time domain) and correlated with measured current peaks of short circuits. The variations in the signals provided by the PZT sensor during welding, however, were imperceptible in the time domain and were difficult to correlate to those changes in welding current. Their conclusions were that the arc behaviour can be monitored to detect weld quality and that the soundwaves from metal droplet transfer are easier to detect than those of arc re-ignition and produce the most perceptible acoustic patterns. Arc re-ignition is an important facet to monitor during welding, as arc extinguishment can lead to material irregularities. This, however, is difficult to distinguish from other phenomena.
Cayo and Absi
[21] examined acoustic signals emitted from three welding transfer modes; the microphone position was of 150 mm from the weld pool. The study was based on analysing sound waves resultant from changes within the electric arc and impact of the molten drops into the weld pool. Nevertheless, they have found that the acoustic signals of the impacted droplets in the spray transference mode are difficult to perceive due to the smaller sizes and reduced specific droplet mass during this mode of transfer. Their study showed that the sound pressure can be a good online indicator for globular transfer, the intermediate mode between dip and spray. For the short circuit transfer mode, the variation in the sound pressure is not very clear and the authors could not correlate them with the defected welds.
Karlsson
[22] investigated monitoring acoustic patterns emitted from weld cooling with no defect. Such monitoring was carried out to detect crack initiations, growth, and their locations by analysing the acoustic signals emitted during the material transformation. The monitoring of the airborne acoustic emissions for several welds was conducted in the range of 5–15 s after finishing the weld. Furthermore, Karlsson’s work stated that “emitted spikes and noise have frequencies around 150–200 kHz indicating crack growth in the material” and the study again concluded the theory of an increase in AE by the initiation and propagation of cracks within cooling welds.
Cayo and Alfaro
[23] made the comparative analysis in the time and frequency domain with the acoustical pressure generated by the electric arc to determine which of the two analysis methods was better to evaluate the stability in the MAG welding process. They concluded that the integrity of a weld can be disturbed through the introduction of grease contaminants and that the welding arc becomes unstable when passing over an unclean surface, resulting in an acoustic pressure decrease, but with increased variations in its AE. Moreover, the analysis of AE in the time domain is more helpful in distinguishing the variations in the acoustic pressures than in the frequency domain.
Wadley and Scruby
[24] measured the static component of displacement using conventional strain gauge arrangements. This enabled the development of a reasonable understanding in linking AE outputs to the microstructure’s static mechanical properties. The method allows for the detection to locate the site of active defects, especially those that grow under service loads. This technique, however, is insensitive to flaws that are not critically stressed. It should also be noted that some processes of deformation/fracture are undetectable (such as ductile fracture). Typical AE data for the steels show that carbon concentration clearly has a strong effect upon the AE activity.
Interestingly, these researchers
[24] noticed that, during plastic deformation, the duplex microstructure gave the most AE with medium carbon bainite and austenite phases being the only other structures giving detectable emission. For a slow-cooled ferrite–pearlite structure, the AE was associated with the nucleation of slip bands. This research concluded that intermediate cooling rates resulted in a dual-phase microstructure consisting of ferrite and austenite (α/γ) in low carbon steel. Subsequent straining generated very intense acoustic emissions, believed to be associated with retained austenite slip occurring from the diffusionless martensitic transformation. However, when an intergranular crack advances in such steels, this is the most readily detectable fracture process.
Sharma et al.
[25] used the following incidents to promote welding defects: cotton thread for inclusion, grease for porosity, and metallic wire inserted in different parts to promote cracking. All five specimens had the following AE parameters recorded: amplitude, duration, count, RMS, energy, and event all vs. time. Their observations identified that an 85–100 dB signal was recorded for normal welding and 120–125 ms was recorded as the standard duration time for all welding. There was no identification of defects during the “welding only” detection phase, but during the cooling phase the following imperfections were identified: porosity, shrinkage, and cracks. Radiographic results verified these defects in the weld and corroborated the AE results. It was also noticed that if a change occurred in the range from 80 to 100 amps, an increase in hardness resulted.
Very similar to AE and strain gauges is the use of fibre Bragg grating (FBG), wherein strain released from thermal variations can be extracted from the waveform travelling through optic fibre waveguide. When variation is subjected to the FBG in the form of strain, this waveform is changed, and such variations measured. FBGs were used to determine material differences during welding by Rodriguez-Cobo et al.
[26].
The discussed investigations reinforce the applicability of using both airborne and contact methods of AE for the determination of weld surface and volumetric integrity both in situ or off-line.
As there are many thermal and kinematic events occurring during the weld cycle, it is important to deliberately force or seed the correct defect for detection. Out of all the deliberate defect tests, it is thought that the use different material inserts promotes cracks more readily than other techniques. The experimentation of the presented work will follow the ideas presented by
[9]. Such tests will investigate alternative, but more compliant, insert materials to closely obtain more expansive realistic datasets wherein both time and frequency domains will show distinguishing phenomena. By using advanced PAC wideband AE sensors, it is possible to acquire a higher-resolution recording, displaying the suitability for such technologies, which was not displayed by
[9]. Other authors also concluded that it was not possible to detect ductile fracture mechanisms as well as flaws that were stressed sufficiently to detect initial crack growth; such sensor enhancements should now ensure this is possible where both elastic and plastic material phenomena are identified during scratch tests
[27]. The setup of using an AE contact sensor to distinguish crack initiation/propagations with such a setup using different insertable inserts has not been seen before in the literature.
In terms of the specific digital signal processing techniques to show separation between the two sources of extracted acoustic date, the following discussions are necessary.
He and Li’s work
[28] used time–frequency analysis in the form of continuous wavelet transform (CWT) in order to compress the AE signal and use it as a signature fingerprint to discriminate between a good/bad MIG weld. Principal Component Analysis (PCA) are used to determine the most significant compressed–transformed data of the CWT-applied AE data. A support vector machine is then used to provide automated classifications between a good/bad weld. PCA used as the CWT AE information can produce computationally heavy data signatures on their own. This is perhaps the closest work completed when compared with the discussed work here, where short-time Fourier transform (STFT) is used a method to differentiate such phenomena. STFT was chosen as CWT was considered more computationally expensive and needing techniques such as PCA to display signal differences. STFTs have been used by the authors before in previous work
[27] with the setup experiment providing definite phenomena from material interactions. STFT is perfectly adequate to describe such a focused phenomenon of interest, because converting from CWT to the time–frequency domain can result in a loss of substantial information.
Basantes-Defaz et al.’s
[29] research also discusses very similar work to what is discussed here. However, instead of using a non-contact microphone and a contact acoustic emission sensor (AE), it looks at using two contact AE sensors as airborne sensors, which is certainly novel, because the use of contact dedicated to non-contact, airborne audio acoustic activity is very unorthodox. Surely, using dedicated airborne microphones as in the research presented here is more appropriate and sensitive to airborne change phenomena? Both contact and non-contact microphones have been used, where the former monitors phenomena in the ultrasonic frequency range and the latter monitors phenomena in the audio frequency range. Nevertheless, the airborne acoustic signature is obtained in
[29] and used in a qualitative manner to represent the depth of penetration of the weld where different signatures result in different achieved depths of weld penetration. The two AE sensors work independent of each other as they both have different characteristics; one being specifically used for a low frequency range, and the other a wide-band frequency range. In addition, the dB noise threshold for sensitivity is set differently for each sensor. The third sensor, however, is configured in a conventional manner on a track to move along at the same rate as the weld torch. The sensor in this case is a contact ultrasonic pulse echo system and based on time-of-flight measurements, which can discriminate between different material characteristics through a change in boundaries based on different recorded penetration energies.
This work not only verifies the setup with the work being discussed here, it also discusses that such technologies play a vital role in the detection of welding anomalies. Finally, the work discussed by
[29] concentrates specifically on anomaly identification, namely burn through. Further work will look at using machine learning technologies to provide automatic detection. The work concentrates on anomalies, namely weld quality and crack detection. Future work will also look at automated discrimination through the use of machine learning technologies.
Interrogating Basentes-Defaz et al.’s research work
[29] revealed that the data collected by these AE sensors also provided information on possible superficial discontinuities or defects in the weld metal. If a superficial indentation or defect is displayed, both the ASL scan and the AE absolute energy plot showed the exact location of this surface discontinuity, which was located at 85 mm from the start of the weld. The sudden signal burst that occurred at the exact location of the superficial indication was determined, confirming that, where a non-uniform condition or weld defect appears, this appears as a sudden surge in the AE absolute energy and therefore a key indicator in determining the signature and potential characteristic for automatic detection.
There are a few sources, such as Madhvacharyula et al.
[30], Kanungo et al.
[31], and Kale et al.
[32], which have started reporting machine learning techniques applied to detected AE sources that can indicate welding defects. Moreover, Kanungo et al.’s work
[31] used cluster k-means analysis to show the more significant features from several AE parameters, namely peak amplitude, kurtosis, energy, and the number of counts. As there are several dimensions here and cluster k-means analysis requires the most significant features, principal component analysis is used.
AE signals used to distinguish cracks from porosity would look at short-burst high-amplitude data vs. shorter decay time and lower amplitude, respectively, as discussed by Roca et al.
[33]. These differences were stored within an artificial neural network to give a computer model of the gas–metal arc welding (GMAW) process. This research is perhaps the closest research discussed in the paper where several AE parameters are used to distinguish different crack phenomena. However, no machining learning work has been used within this research work, as the focus is purely on connecting the signal analysis phenomena with the physical material analysis phenomena.
This entry is adapted from the peer-reviewed paper 10.3390/acoustics5030043