Phased Array Ultrasonic Testing for Crane Steel Welds

Fatigue cracks in overhead crane steel structure welds are the leading cause of structural failure in lifting appliances. Critical load-bearing welds—including main girder splice welds, end carriage-to-main girder connection welds, and trolley frame welds—are subjected to repeated cyclic loading, where crack initiation and propagation often remain invisible inside the steel structure. Conventional manual ultrasonic testing with single-crystal probes is inefficient (approximately 30 minutes per weld), relies heavily on operator experience for defect positioning, and cannot retain imaging data. The KL-WELD-PAUT system employs 64-element phased-array ultrasonic testing (PAUT) technology, with weld inspection grades and acceptance criteria referenced to GB/T 11345-2023 and the load-bearing weld quality requirements of ISO 4301 Crane Design Standard. Combined with the TOFD time-of-flight diffraction method and an AI-based defect classification model, the system delivers rapid imaging, precise quantification, and type identification of weld defects. A single weld inspection takes just 5–8 minutes, with crack height quantification accuracy of ±1 mm and a 92% defect classification accuracy.

Phased-Array Ultrasonic Testing Principles and System Architecture

The core of the Kelude KL-WELD-PAUT system is a 64-element phased-array probe (center frequency 5 MHz, element pitch 0.6 mm, total aperture 38.4 mm), paired with a multi-channel ultrasonic pulser/receiver board (64:64 parallel, transmit voltage 50–200 V adjustable, gain 0–80 dB). Phased-array technology achieves electronic focusing and beam steering by controlling the transmit delay of each element: the focusing law calculation ensures that acoustic waves from all elements arrive in phase at the target depth, producing a focal spot size of approximately 1.2 mm (at a 40 mm depth)—far superior lateral resolution compared to single-crystal probes. During inspection, the probe scans linearly along the weld length with encoder-triggered acquisition at a step interval of 0.5 mm, covering a single-pass weld width of 35 mm.

The system supports three scanning modes:

Sector Scan (S-Scan) — angular range −45° to +45° in 1° increments, ideal for fillet welds and T-joints;

Linear Scan (E-Scan) — electronic linear movement over a 35 mm aperture, suited for butt welds;

Focal Law Scan — individual focusing at each depth, optimized for thin-plate welds up to 20 mm.

Acquired data is displayed in real time as S-scan images, with each frame containing 2,048 A-scan data points. Upon completion of the scan, the system automatically generates three orthogonal projection views: C-scan top view, B-scan side view, and D-scan end view. Full-length weld scan data is saved in standard DICONDE format, supporting offline analysis and historical comparison.

Crack Sizing and Quantitative Assessment Methods

Accurate crack height quantification is performed using the TOFD (Time of Flight Diffraction) technique. When ultrasonic waves strike a crack tip, diffraction waves—rather than reflection waves—are generated, traveling from the transmitting probe to the crack tip and then to the receiving probe. By precisely measuring the diffraction wave travel time t, and using the probe center spacing 2S and sound velocity c, the crack tip depth is calculated as d=√[(ct/2)²−S²]. The time difference between the upper and lower tip diffraction signals directly corresponds to the crack height h. The KL-WELD-PAUT system features a built-in automatic TOFD algorithm that identifies upper and lower tip diffraction signals in the S-scan image, achieving crack height measurement accuracy of ±1 mm (within a thickness range of 10–50 mm).

Crack length is determined using the 6 dB drop method: as the probe moves along the weld length, when the amplitude of the reflected echo from the crack end drops to half of its maximum value (−6 dB), the probe position is recorded—the distance between the left and right −6 dB positions defines the crack length. For locating multi-layer defects in the depth direction, the system utilizes sound path and angle information from the B-scan image, applying trigonometric calculations to determine the precise coordinates of defects within the weld cross-section (depth from surface d and offset from weld centerline x). All measurement results are automatically annotated on B-scan and C-scan images, generating a comprehensive weld inspection report. Inspection standards reference GB/T 11345-2023 Non-destructive testing of welds — Ultrasonic testing — Techniques, testing levels, and assessment.

AI-Powered Defect Classification and Field Validation

The system integrates a built-in CNN-based defect classification model. Its input is a 128×128-pixel time-frequency image from the S-scan, and its output is the probability of six defect types: crack, lack of fusion, incomplete penetration, porosity, slag inclusion, and undercut. The model was trained on 3,800 TOFD-plus-metallography validated calibration datasets, achieving a classification accuracy of 92.3%. The recall rate (detection rate) for cracks reaches 95.1%, with a false alarm rate of 4.8%. The system automatically flags high-risk defects and marks their coordinates on the 3D projection view.

In a butt weld inspection of the main girder of a 32t overhead crane at a shipyard, the system detected two root lack-of-fusion indications measuring 8mm and 14mm in length (at depths of 18mm and 22mm). UT re-inspection and local grinding verification confirmed the defect size deviation was within 1.5mm. This case validates the reliability of the PAUT+AI system in real-world engineering applications. Over the following 12 months of continuous use, the system inspected 486 weld seams and detected 37 defects (9 cracks, 14 lack-of-fusion, 8 porosity, and 6 slag inclusions), with a 94.6% agreement rate against destructive verification. Compared with conventional UT spot checks, the PAUT+AI solution raised the defect detection rate from 78% to 96% and reduced the missed-detection rate from 22% to 4%, significantly lowering the risk of weld failure.

Weld Inspection Applications and Field Case Studies

The KL-WELD-PAUT system is designed for online inspection of load-bearing welds on overhead cranes, covering typical scenarios such as main girder butt welds (plate thickness 10–50mm), T-type fillet welds between the end carriage and main girder (leg size 6–16mm), trolley frame splice welds, and rail clamp welds. Based on plate thickness and weld configuration, the system automatically selects the appropriate probe setup and inspection parameters: for 10–20mm plate thickness, a 5MHz 64-element probe with TOFD combined scanning is used; for 20–40mm, a 4MHz 64-element probe with sector scanning from −60° to +60°; and for 40–50mm, dual-sided, dual-access scanning or a 2.5MHz low-frequency probe is required.

During the annual weld inspection of the main girder on a 40t portal crane at a port, the system completed PAUT scanning and AI analysis of 32 weld seams totaling 48m in just 12 hours, detecting 7 defects (2 cracks, 3 lack-of-fusion, and 2 porosity). UT re-inspection and local grinding verification showed crack sizing deviations of 0.8mm and 1.1mm (both within the design requirement of ±1.5mm), with a 100% defect classification accuracy. The port's equipment department reported that the PAUT+AI solution compressed the annual weld inspection cycle from 5 days to 1.5 days, while digital archiving enabled traceable management of weld health status. Kelude's weld inspection services have been deployed across 12 steel plants, 8 ports, and 6 manufacturing facilities, with more than 2,800 weld seams inspected to date.

Detection Technology Comparison

Frequently Asked Questions