Executive
Summary:
Welding and fabrication industries rely heavily on Non-Destructive
Testing (NDT) film inspection to ensure structural integrity and safety.
Traditional manual inspection is slow, subjective, and prone to human
error—leading to missed defects, inconsistent reporting, and increased
downtime. Dotcom’s AiFab, an AI-powered radiographic film interpretation
software running on the RK3588 edge platform, transforms this process
with precision and speed.
AiFab leverages advanced image enhancement, defect
detection algorithms to highlight the smallest anomalies—cracks, porosity,
slag, inclusions—often invisible to the human eye. With secure cloud
integration, automated reporting, and centralized data management, AiFab
reduces inspection time, improves safety, and ensures consistent quality
validation across industries such as oil & gas, aerospace, automotive, and
power.
Challenges
: Engineering an Integrated and
Automated NDT X-ray detection platform:
Building a Reliable and Scalable NDT fault detection system overcomes these hurdles.
- Manual interpretation of NDT films is time-consuming, error-prone, and inconsistent
- Difficulty in detecting fine defects like porosity, slag, and cracks at early stages
- Lack of standardized reporting and traceability
- Increased downtime and safety risks due to missed or late detection
The Solution : Smart Vision AI for Radiographic Film Analysis:
Dotcom’s AiFab – A plug-and-play AI-enabled inspection system designed for Non-Destructive Testing (NDT) radiographic films, combining advanced image processing with machine learning for fast, accurate, and consistent defect detection. Running on the high-performance RK3588 edge platform, AiFab eliminates manual errors and streamlines the entire inspection workflow.
Core components include:
- Imaging & Processing Unit: Captures high-resolution NDT film images and applies intelligent filters (X-Filter, contrast, gamma correction) to highlight anomalies such as cracks, porosity, and inclusions.
- Local AI Engine: Performs on-device defect detection, enabling inspectors to achieve reliable and repeatable results without internet dependency.
- Expert Interface: A user-friendly desktop application that supports OCR for film text recognition and manual annotations, allowing inspectors to verify or add missed defects before finalizing reports.
How It Works?
- NDT films are scanned using the AiFab imaging system.
- Images undergo ML-based analysis for tungsten, porosity, slag, and crack.
- X-Filter enhances image quality and accuracy.
- Results are securely stored in AWS S3 Bucket for centralized access.
- An application provides defect visualization, annotation, and approval workflows.
Key Takeaways
- AiFab is a scalable, future ready, AI-first solution for NDT film analysis.
- It addresses both operational challenges and business goals—efficiency, safety, and compliance.
- This solution is future-ready for Industry 4.0 transformation and adaptation.
Why Choose AiFab built by Dotcom IoT?
- High-Performance Edge Processing – Powered by the Rockchip RK3588, AiFab performs real-time defect detection and film analysis directly on the device.
- Consistent Accuracy & Reliability – AI-driven detection ensures repeatable results, minimizing human error and variability.
- Seamless Workflow & Reporting – Integrated OCR, manual annotation, and automated report generation streamline the entire inspection process.
- Future-Ready with Continuous Learning – Cloud-based retraining pipeline ensures AiFab evolves with every inspection, adapting to new defect patterns over time.
Technical Architecture of AiFab:
AiFab is built on a hybrid edge–cloud architecture, combining high-performance on-device AI processing with cloud-assisted model retraining. The system ensures real-time defect detection and reporting locally, while continuously improving accuracy through feedback-driven learning.
Device
Layer (AiFab Inspection Unit):
- The Processor (SoC) used is Rockchip RK3588, 8-core (quad Cortex-A76 + quad Cortex-A55), Mali-G610 GPU, 6 TOPS NPU for on-device AI inference.
- It has Memory & Storage up to 32GB LPDDR5 RAM, onboard eMMC + NVMe SSD (via M.2 slot) for high-volume film storage.
- It comes with dual MIPI-CSI camera interfaces; supports high-resolution X-ray film capture with real-time image adjustments (brightness, contrast, gamma).
- There is Application GUI Local desktop interface for:
- LED control, live capture, and
transforms (rotate, flip).
- Manual annotation for Expert,
film detail entry, and defect tagging.
- On-device ML inference for tungsten,
porosity, slag, inclusions.
- Automated report generation and
defect statistics table.
- This provides Connectivity USB 3.1, Gigabit Ethernet, Wi-Fi 6, Bluetooth 5.2 for local data transfer or cloud sync.
- The Power feature comes with USB Type-C PD / DC input with active cooling support for continuous operation.
Software Layer:
- Image Acquisition Module: Captures and preprocesses X-ray films from the imaging unit.
- Preprocessing Module: Applies enhancement filters (contrast, gamma correction, X-Filter) to highlight anomalies.
- AI Engine: Runs defect detection locally on the ultrafast RK3588 NPU.
- User Interaction: Inspectors can manually annotate defects, add film metadata, and approve results directly in the GUI.
Communication Layer:
- Local Data Handling: Images, analysis results, and reports are stored securely on the device.
- Cloud Sync (AWS): Only images and annotations are periodically uploaded for model retraining.
- Retraining Loop: Updated AI models, once trained in the cloud, can be deployed back to AiFab devices for continuous improvement.
Data Storage Layer:
- Local Storage: High-speed NVMe/eMMC stores inspection history, images, and reports.
- Cloud Storage: AWS S3 for long-term archiving of annotated images and training datasets.
Results & Benefits:
|
Benefits |
Impact |
|
50% reduction in inspection time |
Faster film evaluation compared to manual methods |
|
Enhanced accuracy &
consistency |
Reliable detection of cracks, porosity, slag, and
inclusions |
|
Improved worker safety &
reduced downtime |
Early detection prevents failures and costly delays |
|
Centralized data management |
Streamlined audits, compliance, and traceability |
Key Features:
- Multi-format image support: AiFab accepts various image formats (JPG, PNG, JPEG), making it easy to upload and analyze films captured from different imaging devices without extra conversions.
- One-click X-Filter: A specialized enhancement filter that automatically adjusts contrast, brightness, and sharpness, helping inspectors quickly highlight even the finest defect details.
- Automated report generation: Generates comprehensive inspection reports in a few steps, with customizable templates tailored for specific industry or customer requirements.
- Complete NDT film management: Tracks every inspection, preserves operation history, and organizes films systematically for traceability, audits, and compliance.
- OCR-enabled text recognition: Automatically reads identifiers, labels, and annotations from NDT films, reducing manual entry and minimizing the chances of human error.
- Expert manual annotation tools: Allows inspectors to manually add, edit, or correct text and defect markers missed by AI or OCR, ensuring both flexibility and accuracy.
Future Scope:
- Real-time inspection via IoT-enabled imaging sensors, enabling continuous monitoring during fabrication.
- Expansion to ultrasonic and CT-based NDT workflows, covering multi-modal defect detection
- Predictive analytics for maintenance forecasting, reducing unplanned downtime.
- Edge-to-cloud federated learning pipelines to continuously improve AI models without exposing sensitive data.
- Automated defect severity grading to assist inspectors in prioritizing critical issues.
- Integration with enterprise MES/ERP systems for seamless production quality tracking.
- AI-driven adaptive image enhancement that learns optimal filters per defect type.
- Multi-device synchronization for centralized fleet management across factories.
"Turning X-Ray Films into Clear Answers — Smarter, Faster,
and More Reliable NDT with AI."
