AI-Powered Weld Quality Assessment Automating Plastic Pipe Inspection

This patent introduces a revolutionary computer-implemented system for assessing plastic pipe welds using artificial intelligence and ultrasonic testing. Rather than waiting hours or days for manual inspection, this technology provides immediate on-site assessment of weld quality, significantly reducing delays and costs in pipeline construction.

The Problem

Traditional plastic pipe weld inspection relies on manual ultrasonic testing followed by visual assessment at remote test centers. When an inspector detects a potential defect, they must send the ultrasonic scan data to an external facility where human experts perform visual inspection. This process introduces significant delays – assessment results often take hours or days to return to the construction site. If a weld fails inspection, crews must return to the site to produce a replacement weld and conduct additional testing, making the entire process extremely time-consuming and expensive.

The existing workflow also depends on subjective human judgment and visual interpretation, which can vary between inspectors and may miss subtle defects. Construction companies face productivity losses and increased labor costs while waiting for inspection results. The decentralized inspection process creates bottlenecks that slow down project timelines across the pipeline installation industry.

What This Invention Does

This invention presents a fully automated, AI-driven weld assessment system that performs real-time analysis of ultrasonic test data at the construction site. The system captures ultrasonic scan data from plastic pipe welds and uses machine learning algorithms (including neural networks trained on historical weld data) to automatically assess whether the weld meets quality standards.

Instead of waiting for manual inspection, inspectors can receive immediate pass-or-fail results on site. The neural network processes complex ultrasonic patterns and defect signatures that would require hours of expert analysis, delivering consistent, objective assessments in seconds. This automation eliminates the need to transport data to remote test centers and wait for human evaluation.

Key Features

• Computer-Implemented Analysis. The system captures raw ultrasonic scan data and processes it through trained neural networks to classify welds as acceptable or requiring rework, all without human intervention.
• Real-Time On-Site Assessment. Inspection results are available immediately at the construction site rather than after hours or days, enabling rapid decision-making about weld quality.
• Machine Learning Optimization. The neural network is trained on historical ultrasonic data from thousands of welds, learning to recognize defect patterns and assess weld quality with high accuracy.
• Reduced Manual Intervention. The system eliminates the need for inspectors to send samples to external test centers and wait for human assessment, dramatically reducing inspection cycle time.
• Cost Efficiency. By accelerating the inspection process and reducing rework requirements, the system lowers labor costs and prevents expensive project delays.

Who Is Behind It?

The patent was filed by Georg Fischer Rohrleitungssysteme AG, a Germany-based organization. The invention was created by Riccardo Barbone. The patent application was represented by FB Rice Pty Ltd, Sydney, Australia.

The patent traces its priority to 3 September 2024 (EP), establishing the earliest claim date for this technology.

Why It Matters

This technology addresses a critical bottleneck in modern construction and pipeline installation. As manufacturing complexity increases and quality standards become more stringent, the ability to perform immediate on-site assessment of critical welds saves both time and money. The shift from manual visual inspection to AI-powered analysis represents a significant advancement in industrial quality control. This patent represents important innovation in its field, with potential applications that could improve safety, efficiency, or functionality across multiple industries.

Related Concepts

Ultrasonic testing is a widely used non-destructive testing technique that propagates high-frequency sound waves through a material to detect internal flaws, voids, and inclusions. It is extensively applied in pipeline construction to verify weld integrity without cutting or destroying the component under inspection.

Artificial neural networks are machine learning models loosely inspired by biological neural structures. Trained on large labelled datasets, they can learn to recognise complex patterns – such as those found in ultrasonic weld scans – and classify new samples with high accuracy, enabling the automation of quality assessment tasks that previously required skilled human interpretation.

---

AU 2025223879 was published in the Australian Official Journal of Patents on 19 March 2026 and is open for public inspection. Patent applications represent inventions that are sought to be protected and do not necessarily reflect commercially available products.