★ ★ ★ ★
★ ★ ★ ★
★ ★ ★ ★
Active - PM 0.0 / 36

UltrasonaAIc

AI-based ultrasonic testing for
real-time defect detection in NDT

An ERDF/JTF NRW funded research project developing an AI-powered software system for automated, explainable evaluation of ultrasonic inspections – from simulation to productive deployment in industry.

1.99 M EUR
Project Volume
85 %
ERDF Funding Rate
36 Months
Duration
3 Partners
Consortium
90 % less
Misclassifications
Challenges

Why does NDT need AI?

Skills Shortage

The shortage of qualified NDT personnel is growing. AI must act as an assistance system to close the gap and effectively support less experienced staff.

Data Growth

The advancing digitalisation of inspection processes generates huge volumes of data that can no longer be economically evaluated manually.

Misclassifications

Classic threshold-based methods fail with small defects and poor signal-to-noise ratios. A complete characterisation of defect type, size and location is not possible.

High Inspection Costs

Costly manual evaluation and radiographic alternatives with radiation protection requirements are expensive, especially for SMEs without in-house AI expertise.

Innovation

Six innovative approaches –
one scalable system

UltrasonAIc combines AI development, Sim2Real transfer, explainable AI and user-centred design for the first time in a complete inspection system for industry.

01

AI Automation Level 2-3

For the first time, ultrasonic testing is elevated to automation level 2 and 3. AI models detect defects with up to 75 % smaller detectable flaw sizes and up to 90 % fewer misclassifications. Multiple components can be evaluated in parallel and in real time.

02

Multimodal AI Models

Conventional UT, PAUT and TOFD are processed simultaneously by a single AI model for the first time. Since the techniques complement each other in defect detection, the combined evaluation opens new application fields and can replace expensive radiographic methods.

03

Sim2Real Transfer with GAN

Realistic simulation data is generated from CAD models. Generative Adversarial Networks then adapt it to real measurement conditions, efficiently closing the Sim2Real gap without the costly production of test specimens with artificial defects.

04

Explainable AI (XAI)

Under the EU AI Act, human oversight of high-risk AI systems is mandatory. Our system implements user-controlled explainable AI: inspectors receive transparent decision justifications, can intervene, and retain full control at all times.

05

No-Code Inspection Platform

The complete software framework covers data input, simulation, AI training and real-time inference. Users without AI expertise can independently configure new inspection scenarios. SMEs in particular gain access to productive AI in NDT for the first time.

06

New Training Concepts

Role-specific training content for data collectors, evaluators and inspection supervisors, aligned with ISO 9712 and ISO TS 25107. DGZfP integrates these into nationwide training programmes with over 1,000 participants annually.

deeplify BAM DGZfP
Consortium

Three Partners. One Vision.

An interdisciplinary consortium of industry, science and education.

P1 - Bochum, NRW
deeplify GmbH
deeplify GmbH, Bochum

AI startup for non-destructive testing systems. Consortium leader with expertise in developing No-Code AI platforms for industrial quality inspection.

Budget
954,780 EUR
Foerderquote
80 %
Stunden
11,089 h
P2 - Berlin
BAM
Federal Institute for Materials Research and Testing, Berlin

Over 15 years of experience in UT simulation, Probability of Detection and human factors in NDT.

Budget
894,458 EUR
Foerderquote
90 %
Stunden
--
P3 - Dortmund, NRW
DGZfP
DGZfP Training and Education GmbH, Dortmund

Europe's largest provider of NDT training and certification. Centre of excellence for PAUT/TOFD in Dortmund.

Budget
137,636 EUR
Foerderquote
90 %
Stunden
--
Budget distribution – Total: 1,986,838 EUR
deeplify GmbH – 954,780 EUR
BAM – 894,458 EUR
DGZfP – 137,636 EUR
Exploitation

From research
to market

The project outcome addresses four customer segments and diverse industrial application fields – cross-sector and scalable.

Manufacturing Companies

Inline quality inspection in welding production enables immediate feedback to the process. Pseudo-scrap decreases by up to 90 %, and up to 10 % more components can be reworked and recycled.

Inspection Service Providers

Up to 90 % more efficient evaluation, reduced staffing requirements and higher reliability. Scalable to new test parts and defect types without complex reprogramming.

Infrastructure Operators

Pipelines, bridges and refineries: combined with robotics, drones and IoT-enabled devices, the system enables demand-driven maintenance and extended asset lifetimes.

Rail Transport

Automated UT inspection of wheelset axles and rail infrastructure increases availability in railway operations and directly contributes to public safety.

Science and Standardisation

BAM and DGZfP publish results Open Access and feed them into ongoing standards revisions (ISO TS 25107, DIN, CEN).

Sustainability

Less scrap, fewer energy-intensive repairs and a faster transition to hydrogen infrastructure as a contribution to climate targets.

★★★★★ EUROPEAN ADDED VALUE

Strengthening the EU single market through digital industrial technology

Data sovereignty remains fully in Germany. Economic exploitation takes place directly within the EU single market and strengthens Europe's digital sovereignty in AI-supported inspection technologies.

EU single market strengthening
Data sovereignty in Germany
Circular economy contribution
Energy efficiency and climate targets
FAQ

Frequently Asked Questions

Answers to the most common questions about the UltrasonAIc project.

UltrasonAIc develops an AI-powered software system for fully automated evaluation of ultrasonic testing (UT). The goal is to elevate the automation level in NDT to Level 2 and 3, with significantly higher reliability and without the previously high manual effort.

The project is funded under the ERDF/JTF NRW 2021-2027 programme through the call Industrie.IN.NRW. The total volume is 1,986,838 EUR, of which 1,692,676 EUR (85 %) is funded from EU and state funds.

The consortium consists of deeplify GmbH (Bochum) as consortium leader, the Federal Institute for Materials Research and Testing (BAM, Berlin), and DGZfP Training and Education GmbH (Dortmund).

Since real test specimens with defined defects are difficult to obtain, simulation data is generated from CAD models. Generative Adversarial Networks then adapt this data to real measurement conditions, creating a large training dataset cost-effectively.

The project focuses on three ultrasonic techniques: conventional pulse-echo technique, Phased Array UT (PAUT) and Time-of-Flight Diffraction UT (TOFD). A key novelty is the simultaneous processing of all three modalities by a single AI model.

Under the EU AI Act, high-risk AI systems must be subject to human oversight. Explainable AI ensures that the basis of every decision is transparent, users can intervene, and acceptance of the technology in daily practice is secured.

The system is applicable across industries: welding manufacturing, inspection service providers, infrastructure monitoring (pipelines, bridges), rail transport, drone and robotics deployments, and research. SMEs without AI expertise benefit particularly from the No-Code platform.

DGZfP identifies the further training needs of NDT personnel and develops role-specific training concepts for data collectors, evaluators and inspection supervisors. These are integrated into nationwide programmes and may feed into the revision of ISO TS 25107.

Three main deliverables: (1) a validated software prototype for AI-supported UT evaluation, (2) scientific publications and an industrial guideline, and (3) updated training concepts for NDT personnel. Milestone M3 (prototype tested with application partners) is planned for PM 30.

As an associated partner you can provide real UT data or trial the prototype. As an NDT inspector you can participate in user studies on human-machine interaction. Interested companies, research institutions or individuals can contact the project team at deeplify GmbH directly.

Contact

Interested in the project?
We look forward to connecting.

For questions about the research project, cooperation opportunities or participation in user studies, please contact the project team.

info@deeplify.de deeplify.de
EFRE JTF NRW
Co-funded by the European Union
European Regional Development Fund · NRW 2021–2027 · Innovation und Forschung · Industrie.IN.NRW