2018 / Data analytics
Welding quality prediction
We collected welding current and voltage data and analyzed how process conditions relate to weld quality.
Technologies Machine learning, process optimization
Save image
From welding quality prediction to AI that connects manufacturing work.
Explore the technologies and projects developed on the factory floor.
Select “Related project” on a card to view its illustration below.
The TO-BE column on the right is a research direction, not shipped work. After zooming in, drag the diagram to explore. Each project’s implementation status is listed below.
Explore our manufacturing AI projects in chronological order.
These AI-generated illustrations are based on company presentations and project materials. Equipment and screens are illustrative examples of how the systems work.
2018 / Data analytics
We collected welding current and voltage data and analyzed how process conditions relate to weld quality.
Technologies Machine learning, process optimization
Save image2019 / Vision inspection
We built a vision inspection system for a worker-friendly pilot smart factory. It captures parts and compares their detected positions with reference data.
Technologies Object detection, transfer learning
Save image2020 to 2022 / Vision inspection
A robot moves to preset inspection positions to capture images. AI detects parts, compares them with reference data, and stores the results with supporting images.
Technologies Robot control, object detection, reference image comparison
Save image2020 to 2021 / Vision inspection
Circuit boards are photographed under controlled lighting. The system compares them with reference images to locate visual defects.
Technologies Reference image comparison, anomaly detection
Save image2020 to 2021 / Vision and text recognition
The system checks package seals and contents for foreign material, reads printed expiry dates, and records the inspection results.
Technologies Image analysis, optical character recognition (OCR)
Save image2020 to 2022 / Text recognition
Cameras capture the displays of inspection equipment and meters that lack a communication interface. The system reads the numbers and saves them in the production management system.
Technologies Cameras, optical character recognition (OCR), equipment data collection
Save image2020 to 2022 / Workflow automation
We researched automation that collects and classifies orders received by email and document, then connects them to production tasks and reporting.
Technologies Robotic process automation (RPA), OCR, document classification
Save image2022 to 2023 / Vision inspection
The system captures poultry carcasses during production, assesses quality attributes, and stores grades and statistics. Image records extend quality checks beyond sample inspections.
Technologies Object detection, anomaly detection, body-region feature analysis
Save image2022 to 2024 / Vision research
We explored methods for reconstructing a 3D object from images taken at multiple viewpoints. This work began with weld inspection and informed later research into visual understanding.
Technologies NeRF, Instant-NGP, DUSt3R
Save image2022 / Training data
We collected and organized AI training data for applications including weld quality and robotic grasping. The work includes quality checks and preparation of training and evaluation datasets.
Technologies Data labeling, data quality verification, training dataset development
Save image2023 / Data analytics
We validated an analytics workflow that extracts features from equipment data and connects prediction models with explanations of their results.
Technologies Fast Fourier transform (FFT), boosting models, AutoML, SHAP
Save image2024 / Vision inspection
We extended inspection to large frames by aligning camera positions with reference images and checking complex structures for missing parts.
Technologies Image segmentation, feature matching, reference image comparison
Save image2024 to 2026 / Manufacturing AI research
This research reads equipment manuals and communication documents to draft connection settings and data interpretations. Settings are checked against the source documents and tested with actual data collection.
Technologies Document recognition, language models, data acquisition integration
Save image2025 / Document AI
The system reads items and specifications from development requests, drawings, and specification sheets. It drafts a bill of materials (BOM) and process information for a reviewer to check against the originals before saving.
Technologies Document and image understanding, structured output, human approval
Save imageWe aim for a smart factory where AI leads manufacturing across product design and development, production planning, materials and inventory, quality, equipment, and cost management. AI understands factory-wide goals and conditions, decides and plans the work, and carries it out through business systems, equipment, and robots. Our TO-BE is a factory that continually improves by feeding execution results into its next decisions and plans.
Let’s review your inspection needs, existing equipment, and repetitive tasks.
Discuss manufacturing AIBack to technology history