Innovative quality inspection methods for CFRP primary structural parts

Informazioni relative al progetto

SonicScan

ID dell’accordo di sovvenzione: 831830

DOI

10.3030/831830

Progetto chiuso

Data della firma CE 11 Marzo 2019

Data di avvio 1 Aprile 2019

Data di completamento 30 Settembre 2021

Finanziato da

• SOCIETAL CHALLENGES - Smart, Green And Integrated Transport

Costo totale

€ 825 377,50

Contributo UE

€ 722 109,00

Coordinato da

PROFACTOR GMBH
Austria

Periodic Reporting for period 2 - SonicScan (Innovative quality inspection methods for CFRP primary structural parts)

Periodo di rendicontazione: 2020-10-01 al 2021-09-30

Sintesi del contesto e degli obiettivi generali del progetto

The SonicScan project aimed at developing NDT methods based on ultrasonic testing that are suitable for primary structural parts. The main challenge is the compact shape of the parts and their high thickness. Existing technologies are typically limited to large shell-like components. To address this problem the project built upon the sampling phased array technology that allows the tomographic inspection of parts and to combine it with a robotic handling system to move the sensor across the part. Particular emphasis was be put on the model-based, automatic planning of the robot's inspection path to ensure that the whole volume of the part is inspected. This approach was based on methods developed for surface inspection and was extended to volumetric inspection. Data analysis for automatic defect detection, segmentation and classification was developed using recent deep learning methods.
The results of the project included:
(1) An automatic method for inspection path planning to ensure full coverage of the whole component, by using an approximated model of the physics of the inspection process.
(2) Enhanced data analysis for volumetric inspection, including 3D reconstruction and machine learning methods for defect detection.

The methods were successfully tested on sample parts, including a large landing gear component. Defect detection was possible in all areas within the physical limits of the inspection technology and the inspection of a large, complex component took just a few minutes.

Lavoro eseguito dall’inizio del progetto fino alla fine del periodo coperto dalla relazione e principali risultati finora ottenuti

Work in the project was following two parallel strands that were merged in the demonstration. One part of the work was dealing with the inspection technology itself, including developments for the 3D reconstruction of volume models and data analysis using machine learning based on data coming from a phased array sensor. The other part of the work dealt with the automation of the inspection on a robot, including the motion planning to obtain collision-free paths that (whenever physically possible) covered the whole volume of the part.

The results were integrated into two robotic workcells. A larger robotic workcell for the scanning of a whole landing gear component was set up at Profactor. This workcell was mainly used to demonstrate the capabilities of the motion planning and to estimate the inspection time that is needed. The second, smaller robot was set up at project partner ACS and used the phased array technology for inspection. In this workcell a smaller section of the landing gear component was tested, with a particular focus on assessing defect detection in different areas of the parts.

Results have been presented and published at the SAMPE conference and in a special issue of Applied Sciences on a Control and Motion Planning in Industrial Applications. The project will also be presented at the JEC World fair in Paris in March 2022.

Progressi oltre lo stato dell’arte e potenziale impatto previsto (incluso l’impatto socioeconomico e le implicazioni sociali più ampie del progetto fino ad ora)

Technologies developed in the SonicScan project enable the fast ultrasonic inspection of parts of complex shape. Previous methods were mainly dealing with large, thin-walled, shell-like components that are essentially two-dimensional. This was achieved through progress in automatic planning methods that use a model-based approach for phased array inspection and enable motion planning for a robotic scanning system. Specific strategies have been used to deal with areas of high curvature, such as along corners of the part.
The second main progress that SonicScan provided was the use of deep learning methods for defect detection in 3D ultrasonic data. Particularly it could be demonstrated that a high detection accuracy can be achieved, even if the deep learning methods are only trained on data synthesized from simulations.

Aside from the direct impact on the efficiency and flexibility of ultrasonic inspection systems, the project also has indirect impact on the use of structural carbon fiber parts on the aerospace industry. Part designs in this domain need to take into account that the part has to be tested. Additional capabilities of inspection systems thus increase the range of design options. This will enhance the use of lightweight composite parts and will make designs more efficient in terms of strength and weight, thus contributing to CO2 savings.

SonicScan phased array scanning process with live visualization of ultrasound data

Volumetric reconstruction of test part

Simulation of robot motion for scan motion planning