Weld inspection is a quality control requirement in structural manufacturing: pressure vessels, pipelines, rail structures, automotive frames, shipbuilding, aerospace components. Two inspection technologies dominate: X-ray radiography (traditional) and AI-powered visual plus ultrasonic inspection (emerging). In China manufacturing, both coexist — but the balance is shifting. This article compares them on cost, accuracy, throughput, and application fit.
X-Ray Radiography: The Established Standard
X-ray produces a 2D image of the weld cross-section, revealing internal defects as density variations. It is the reference method for code-compliant inspection under ASME, AWS, EN, and GB standards. Detection rate for internal volumetric defects larger than 1.5 mm: greater than 95%. Limitation: X-ray cannot reliably detect planar cracks parallel to the beam axis. Radiation safety requirements add operational overhead: exclusion zones, personal dosimetry, licensed operators, radioactive source management for gamma ray variants. Cost per meter (China 2024): pipeline X-ray $15-40/m, structural steel X-ray $8-25/m.
AI Visual Inspection: What It Can and Cannot Do
AI visual inspection uses high-resolution cameras and trained neural networks to detect surface-breaking defects: cracks, undercut, overlap, spatter, and profile deviations. Detection rate on surface defects larger than 0.3 mm: greater than 97% when properly trained. Speed: 10-30 meters per minute — 5-10x faster than manual visual inspection, 20-50x faster than X-ray for equivalent weld length. Cost per meter after amortization: $1-5. Critical limitation: AI visual inspection detects surface and near-surface defects only. Internal defects — the most dangerous category for pressure-containing applications — are invisible to cameras. For welds where internal quality is the primary requirement, AI visual inspection supplements but cannot replace X-ray or ultrasonic methods.
AI-Powered Ultrasonic: The Gap Filler
Combining AI with phased array ultrasonic (PAUT) probes closes the internal defect gap. The PAUT probe sweeps an ultrasonic beam through the weld volume; AI algorithms interpret reflected signal patterns to classify defect type, size, and location. This combination outperforms traditional manual PAUT interpretation: AI eliminates operator fatigue errors, maintains consistent sensitivity across a full shift, and produces structured digital records automatically. In China pressure vessel and pipeline sectors, AI-enhanced PAUT systems from domestic manufacturers (Zhongke Yanhuang, Beijing Tianmai) are replacing traditional PAUT at premium inspection facilities.
Application Matching
Use X-ray when code compliance requires radiographic records or the inspection is a one-time acceptance test. Use AI visual inspection when surface defect detection at high throughput is the goal or per-unit X-ray cost is prohibitive. Use AI-enhanced PAUT when internal defect detection is required at production throughput rates, radiation licensing is not available, or continuous monitoring of a production weld line is the application.
For inspection automation context: industrial inspection equipment and innovations in China inspection landscape.
Sources
- China Nondestructive Testing Society: Annual technical review 2024
- ESM China: AI inspection market China 2024-2025
- ASME BPVC Section V: Nondestructive Examination standards reference
