Case Study | How 3D Vision Powers Intelligent Depalletizing of Refractory Bricks

In refractory material production, kiln car depalletizing serves as a vital connecting link between kiln firing and downstream processes. Traditional manual depalletizing not only entails high labor intensity and dusty, harsh working conditions, but also frequently causes incorrect placement orientations and unstable sorting efficiency when handling special-shaped refractory bricks with tapered ends.

Recently, Transfer Technology has developed a dedicated 3D vision intelligent depalletizing system for a leading domestic refractory material enterprise. Featuring high-precision recognition and stable group picking capabilities, the system effectively solves core pain points in depalletizing multi-specification and special-shaped workpieces, enabling intelligent upgrading of automated production lines.

1. Project Pain Points

The on-site depalletizing process covers more than 20 types of refractory bricks. Most bricks feature a 10–30 mm taper (tapered ends) and must be placed in a strict specified orientation to meet the assembly requirements of downstream processes.

Two sizes of kiln cars are adopted on site, forming diverse and irregular stack configurations. Each picking cycle requires simultaneous handling of 4 to 6 bricks for downstream transmission. The traditional manual operation suffers from unstable cycle efficiency and frequent orientation judgment errors during long-duration work.

Meanwhile, the high-temperature and dusty working environment brings severe labor shortages and rising occupational health management costs for enterprises. Achieving precise recognition and stable depalletizing of tapered special-shaped bricks has become the core demand of the project.

2. Solution

Transfer Technology adopts a one-camera-one-robot system architecture, with a single 3D intelligent camera matching one industrial robot and one independent depalletizing station. The compact layout fits perfectly with the existing kiln car conveying lines without major on-site renovation.

• The camera is installed in a top-view configuration at a height of 3.27 meters from the stack bottom, with a field of view fully covering the maximum stack dimension of 1200 mm × 1000 mm × 1000 mm.

• The vision system communicates with the robot via TCP/IP Socket protocol, outputting real-time 3D coordinates and orientation data of bricks to guide the robot in accurate group picking and standardized placement.

• Powered by self-developed vision software, the system supports automatic brick type switching. It rapidly switches recognition models according to upper computer signals without manual intervention or production downtime.

Tailored for complex on-site working conditions, the system is equipped with comprehensive anomaly detection logic. It triggers real-time alarms for incorrect brick orientations and missing bricks in stacks to avoid wrong picking and missing picking. For bricks with a taper below 10 mm, stable and accurate recognition can be realized with consistent incoming orientations.

3. Core Performance

±2 mm Recognition Accuracy & Reliable Orientation Positioning

The system achieves a recognition accuracy of ±2 mm for refractory bricks, with an overall vision-guided depalletizing accuracy rate and brick orientation recognition accuracy rate both reaching ≥99.9%. For special-shaped bricks with tapered ends, high-precision orientation judgment guides robots to place bricks in the standard direction, completely eliminating rework caused by manual judgment errors. The millimeter-level positioning accuracy ensures consistent picking postures for each brick during group picking, greatly improving overall picking stability.

4-Second Ultra-Fast Recognition, Matching Production Line Rhythm

The full workflow of 3D point cloud acquisition, algorithm recognition, and grasping coordinate output is completed within 4 seconds. The system supports one-scan batch coordinate output for entire brick rows, adapting flexibly to group picking modes of 4, 5, or 6 bricks. Eliminating repeated single-brick calculation, it perfectly matches the row-based stacking characteristics of kiln car loads and maximizes single-cycle picking efficiency.

Wide Working Condition Adaptability & Proactive Anomaly Early Warning

The system delivers strong adaptability to normal incoming material fluctuations. It maintains stable recognition performance under working conditions with brick gaps ≥5 mm, in-row brick misalignment ≤40 mm, and product damage rate ≤1%. Built-in stack verification logic enables active anomaly detection for missing bricks and incorrect orientations, intercepting defective stacks in advance to reduce invalid picking and material transfer errors.

After project implementation, the depalletizing station realizes fully automatic unmanned operation, drastically reducing labor intensity and staff exposure to high-temperature and dusty environments. The automatic multi-brick-type switching capability meets flexible production demands, eliminating downtime for model replacement and significantly improving production switching efficiency.

The core of automation upgrading in the refractory industry lies in the stable adaptation to special-shaped workpieces and complex stacked scenarios. Transfer Technology’s 3D vision intelligent depalletizing system relies on mature algorithm capabilities and rich engineering experience to provide replicable and reliable automated solutions for depalletizing and palletizing scenarios in refractory, building materials, and other industries, helping manufacturing enterprises achieve flexible, efficient, and stable intelligent production.