Laser Ultrasonics 2010 Laser-based ultrasonic-emission sensor for in-process monitoring during high-speed laser welding B. Pouet, A. Wartelle and S. Breugnot Bossa Nova Technologies, Venice, CA 90291, USA S. Ream Edison Welding Institute, Columbus, Ohio 43221, USA
Motivation Fuel cells Attractive energy alternative Efficient, quiet, non-polluting Run on pure hydrogen, methanol, diesel, other hydrocarbons Applications: Commercial & Military (Electronics / residential / Car & truck Auxillary / Automotive / Heavy Vehicle / Marine / industrial.) Manufacturing challenges One generic challenge that appears in many fuel cell system designs: To join thin stainless-steel sheet for variety of components (bi-polar plates, recuperators, reformers, cassettes and other heat exchangers) Biggest opportunity for fuel cell is in the automotive sector.
Motivation Fuel Cell for Automotive Sector A fuel cell vehicle includes ~400m long weld - 400 bi-polar plates - 1 meter of laser weld / bi-polar plate Very high-speed welding needed to achieve high production rate and cost target Manufacturing Challenges Zero tolerance for defect (lack of fusion / lack of penetration) Post-process inspection is not possible (Slows production rate & Increases cost) In-process inspection is needed
Laser Welding Highlights Narrow weld seam Minimum heat affected zone Little metallurgic effect on material Little distortion No filler material required Non-contact and no-wear High process speed CO 2 laser weld @ 10m/min Nice but too wide & too slow! Fiber Laser @ 800mm/s Single mode CW Fiber Laser Best for welding of thin metal sheet Demonstration using a 600W Single mode CW Ytterbium fiber lasers (l=1070nm) at 800mm/s
Laser Welding Thermal Conduction Welding Penetration Welding
Laser Ultrasonic in-process inspection Laser Ultrasonic inspection In-process Inspection limited by the repetition rate of generation laser In-processing monitoring of welding quality by monitoring the AE During the welding process, the weld vicinity is subject to high level of strain leading to localized strong elastic and non-elastic behavior of the material that is associated with continuous and/or rapid release of elastic energy: Acoustic Emission (AE). Welding Laser acts as the ultrasonic source Using a laser-ultrasonic sensor to follow the welding laser and to listen to the ultrasonic noise emanating from the weld
Ultrasonic Emission (UE) Airborne Acoustic Emission Acoustic Emission (AE) emanates from the weld pool as the generated vapor displaces the ambient air Detected by microphone (<100kHz) Not used for assessing weld quality Mostly used for controlling the welding condition (laser focal height) In-Solid Acoustic Emission Emanates from the weld Buried source, HF >100kHz Propagate in the plate Carries information about the internal process Well suited for In-process inspection
Detection of Ultrasonic Emission Sensor Requirements Direct detection of the ultrasonic surface displacement Transverse surface motion up to 1m/s Unprepared surface High sensitivity (sufficient for single shot measurement) Small footprint Broadband detection [20kHz to 2MHz and higher] Ability to measure very near the weld and on top of the weld molten pool. Must be able to be integrated with the welding Laser Fiberized Random-Quadrature Multi-channel Interferometer
Speckle processing Random-Quadrature Multi-Channel Interferometer Use of a detector array instead of single-element detector to sum all contributions and increase sensitivity EQUIVALENT TO MANY SINGLE-SPECKLE INTERFEROMETERS IN PARALLEL Signal beam Probe beam Multi-detector Absolute amplitude demodulation Signal out Sample Reference beam Interference principle used in Quartet Optical path difference Speckle pattern Undesired signals from object motion are filtered out electronically No-stabilization required: Quadrature is achieved via the random distribution of speckle phases
Laser Ultrasonic Sensor Fiberized Random-Quadrature Multi-channel Interferometer
Laser welding prototype platform Welding system - 600W Single mode CW Ytterbium fiber laser (l=1070nm) - Focal spot size = 19mm - Shielding gas injected through coaxial nozzle - Sample is fixed - Laser beam position controlled by XYZ translation - Thin sheet welding demonstrated at 800mm/s Sensor integration - Mounted with laser welding head. - Sensor follows the welding laser - Constant offset during welding/measurement (distance between laser welding & detection spots) - Detection can be positioned near or on top of the weld
Laser Ultrasonic Sensor- Setup Welding Laser Optical Head Stand-off = 10cm Clamp Sample to weld To Demodulator Demodulator
Signal Detection & Processing Computer & Acquisition Card Multi-channel detector High-Pass Filter - To reject the background noise - 20KHz / 200kHz / 1MHz To correlate with visual/destructive inspection - Signal Processing - RMS / sliding window Display
Test Samples SAMPLES - Stainless steel sheets (2) - Thickness = 100mm - Sample length = 10cm Tab weld Tab INDUCED DEFECTS - To introduce a small gap between sheets: - Small tab (100mm thin & 5mm wide) - Small wire - To introduce contaminant between sheets - Paint, silicone Peeled sample Wire weld WELDING PARAMETERS - Welding length: 70mm - Welding speed for test: 100mm/s & 200mm/s * weld * Sensor demonstrated at 3m/s.
Measurement Procedure - Record a 1 st pass with Welding Laser Off: To acquire background noise - Record a 2 nd pass with Welding Laser Off & calibration signal ON: To acquire calibration signal RMS / 200ms sliding window - 3 rd pass, record UE signal from welding Weld length=70mm Calibration Signal @ 240kHz Plate length=100mm RMS
Noise Sources Sensor - Detector Electronic Noise (minimized by design) - Laser Intensity Noise (rejected by differential detection scheme) - Shot noise limited detection Environment - Electromagnetic noise from the translation stage motor (pickup from the acquisition card) Solution: shielding of acquisition card Experiment - Optical noise from transverse speckle motion. No noise visible at 200mm/s - Doppler shift due to variation in stand-off distance. Not an issue: The welding laser beam has tighter stand-off distance requirement than the detection laser. Example of EM noise before shielding of acquisition card - Vibration noise (motor vibration.) Rejected if frequency below the detector High-pass filter cut-off frequency.
Results No Defect - - Sensor: [200kHz 10MHz] - Welding speed = 100mm/s - Offset between weld and sensor: 5mm - Signal strength variation (reflectivity): 50% - Background noise is low - Ultrasonic Emission burst visible when welding start & stop
Results induced defects: Gap - 1cm - Sensor: [200kHz 10MHz] - Welding speed = 100mm/s - Offset = 5mm - Sliding window = 200ms
Results Induced defect: contaminant - - Sensor: [200kHz 10MHz] - Welding speed = 200mm/s - Offset = 2mm - Sliding window = 800ms
Detection on top of Keyhole Induced defect: contaminant - Welding spot: 19mm - Detection spot =100mm - Sensor: [1MHz 20MHz] - Welding speed = 200mm/s - Sliding window = 800ms
Findings summary Tested Detector Bandwidth - Low Frequency [20kHz to 2MHz] Sensitive to background & laser Intensity noises - Medium Frequency [200kHz 10MHz] Most useful for this demonstration [200kHz-1MHz] - High Frequency [1MHz 10MHz] Used for detection on top of weld pool Detection near the weld: - Closer to weld leads to stronger UE signals - Strong UE signals clearly correlate with Lack of fusion and partial penetration defects - Sharp UE bursts caused by random impurities on the top surface getting vaporized Spatter ejection Recoil force Detection on top of the Keyhole - Location of defect corresponds to a loss in the detected signal! - Despite the 1MHz High frequency cut-off, Strong background noise visible. - Detection not reliable / too much disturbance
Conclusion Preliminary results are very promising For detection near the weld, using very simple signal processing we clearly detected Lack of fusion & partial penetration defects Some weak UE signals (slightly above the background noise) were correlated with concave weld defects (further processing needed) Detection on top of the keyhole is very noisy. Next Step Detection on the weld seam, behind the weld pool to be tested Further Signal processing to improve defect detection & characterizations Further Signal processing to reject unwanted signals (UE bursts from random impurities) Acknowledgement: This work was supported by the National Science Foundation, DMI-0740241