In a recent paper published in the journal Aerospace, researchers explored a non-contact approach to detect multiple failures in avionic systems. They introduced a novel non-destructive algorithm for diagnosing failures in printed circuit boards (PCBs) in modern aircraft, supporting advancements in control and electrification.
This algorithm leverages advanced technology to improve the reliability and efficiency of PCB diagnostics, particularly in the aerospace field. It combines a self-attention mechanism with an adaptive neural network to enhance diagnostic accuracy. The goal was to address the complexity of multi-failure detection in PCBs, which is important for ensuring the safety and performance of avionic systems.
Advancement in the Aerospace Industry
The integration of aerospace and electronic information technologies has driven advancements in avionic systems. Modern aircraft now rely heavily on complex electronic modules for essential functions like engine control, communication, and servo coordination. The shift towards More Electric Aircraft (MEA) aims to replace traditional hydraulic and pneumatic systems with electrical ones, reducing energy loss and improving efficiency. This transition highlights the need for reliable electronic systems.
PCBs, especially those used in radio frequency (RF) applications, are crucial for high-frequency tasks but present diagnostic challenges due to their sensitivity to electromagnetic interference and complex structures. Effective diagnostic methods are essential to maintain system stability.
While traditional contact-based diagnostics can be effective, they can damage sensitive components and introduce parasitic effects that may distort measurements. This has increased interest in non-contact diagnostic methods that reduce these risks while ensuring accuracy.
Introduction of Novel Diagnostic Algorithm
This paper developed a fast, accurate, and non-destructive algorithm for diagnosing multiple failures in complex PCBs used in aerospace applications. The algorithm includes three main components, PCB feature extraction, feature enhancement, and model training.
To begin, the researchers employed spatial dimension reduction to convert two-dimensional (2D) spatial data into a one-dimensional (1D) format. This transformed data was then combined with frequency data to create a unique 2D spatial/frequency coordinate system.
For feature extraction, the authors used convolutional neural networks (CNNs), specifically a residual network (ResNet) architecture. They integrated a self-attention mechanism (SAM) with an adaptive convolutional graph neural network (AGCN) to reduce overfitting and improve feature analysis. This combination effectively mapped extracted features to multiple failure types, using a binary labeling system where each bit denotes a specific failure.
The experimental setup included a dual-phase amplitude boosting circuit (DPAB) to simulate various PCB failure modes. This design allowed comprehensive testing of the algorithm. Furthermore, a comprehensive data collection strategy was implemented using near-field scanning techniques to gather electromagnetic data from the PCB under different operational scenarios.
Key Findings and Insights
The outcomes showed that the proposed diagnostic technique performed well, especially with multiple simultaneous failures. The algorithm achieved high scores across various metrics: overall precision (99.08%), per-class precision (98.50%), overall recall (98.78%), per-class recall (98.01%), overall F1 score (98.93%), and per-class F1 score (98.25%).
The study found that using ResNet with 34 layers (ResNet34) as the feature extraction network provided an ideal balance between complexity and diagnostic accuracy. Additionally, integrating the AGCN further boosted the model's performance, enhancing diagnostic accuracy. The authors also highlighted the advantages of using electromagnetic data for PCB diagnostics, noting its broader applicability over traditional methods.
This non-contact approach improved detection efficiency while reducing risks associated with contact-based techniques, making it particularly effective for high-frequency and low-noise circuits. Furthermore, the researchers identified that the algorithm effectively handles complex multi-failure scenarios, demonstrating its robustness in real-world applications.
Applications
The presented diagnostic technique has significant implications for the aerospace industry, where dependable and rapid PCB failure diagnosis is critical for the safety and performance of avionic systems. Its non-contact design is particularly valuable in environments where direct contact with PCBs is difficult or undesirable.
The algorithm’s high accuracy and efficiency make it suitable for other high-reliability applications, such as satellite communications and space probes. It can diagnose failures across various electronic systems, including communication modules, radar systems, and other essential components in aircraft and spacecraft, making it ideal where physical contact is impractical or could risk damage.
Beyond aerospace, this diagnostic approach can benefit industries reliant on electronic systems, including telecommunications, automotive, and consumer electronics. Its ability to detect multiple failures in complex PCBs enhances maintenance efficiency and minimizes downtime. Additionally, this technology could drive advancements in automated diagnostics, contributing to the development of smart systems capable of self-monitoring and self-repair.
Conclusion and Future Directions
In summary, the novel diagnostic algorithm effectively addresses multi-failure diagnosis in complex electronic systems. It can potentially revolutionize how PCBs are diagnosed in aerospace and other high-reliability sectors.
Future work could expand the algorithm’s capabilities by incorporating multimodal feature fusion techniques, further improving its diagnostic accuracy and versatility. Overall, this research contributes to the reliability of aerospace systems and sets a foundation for future advancements in electronic diagnostics.
Journal Reference
Liu, C.; Ferlauto, M.; Yuan, H. A Non-Contact AI-Based Approach to Multi-Failure Detection in Avionic Systems. Aerospace 2024, 11, 864. DOI: 10.3390/aerospace11110864, https://www.mdpi.com/2226-4310/11/11/864
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.