Using 3D machine vision and robotic automation to improve bag handling, increase throughput, and reduce product loss.
For this coffee roaster, the robot moving 150-pound bags of raw coffee beans from pallets to the conveyor had become a weak link in production. The existing system frequently tore bags, spilled product, and struggled to accurately locate each bag—contributing to approximately 100,000 pounds of lost beans every year.
Concept Systems developed a vision guided robotic workcell to improve bag handling, increase throughput, and reduce product loss.
The Problem: Ripping Open Bags
The coffee roaster relied on a robot to unload pallets containing 150-pound burlap bags of raw coffee beans and place each bag onto a conveyor leading to the roaster.
The existing robot used plier-like grippers that could tear the burlap, spilling beans and creating product loss. The system also relied on predefined positioning rather than machine vision, limiting its ability to account for variations in how bags were oriented on each pallet.
These limitations slowed the unloading process, created a production bottleneck, and contributed to significant product loss.
The Solution: A Vision Guided Workcell
Concept Systems developed a robotic workcell combining 3D machine vision, laser scanning, custom software, and a redesigned end effector.
Two SICK Ranger cameras and laser scanners create a 3D model of each pallet layer, allowing the system to determine the position and orientation of individual bags. The resulting coordinates are sent directly to the FANUC robot, allowing it to move to the appropriate pick point rather than relying on predetermined positions.
Building an Accurate 3D Model
Each pallet contains 20 bags arranged in four layers. As the robot works through the pallet, the vision system creates a new 3D model for each layer and uses laser triangulation to measure the surface contours of the bags.
Concept engineers positioned and angled the cameras and lasers to capture the entire top of the pallet while keeping the scanning equipment safely outside the robot’s operating area. The resulting gantry positions the scanning equipment approximately 13 feet above the pallet, minimizing the potential for an inadvertent robot collision.
Custom software analyzes the 3D model to identify the edges, position, orientation, and height of each visible bag. Because the cameras are calibrated to the robot’s coordinate system, the software can calculate the center and optimal pick point for each bag and send those coordinates directly to the robot.
The system calculates the pick points for all five bags in a layer in a single scan and sorts them by height. The robot begins with the highest bag and continues through the layer. Once a layer is removed, the system generates a new 3D model for the next layer. When no bags remain, the PLC signals the system to eject the empty pallet and bring in a new one.
Overcoming Glare and Environmental Challenges
Creating an accurate 3D model required Concept to account for glare from reflective objects such as pallet nails. Laser light reflecting from metal could distort the camera image and interfere with the system’s ability to identify the bags.
Concept added polarized filters to the cameras to minimize reflected glare while preserving the light returning from the burlap bags. This produced a clearer image and allowed the vision system to more reliably distinguish the bags from surrounding objects.
Designing a Better End Effector
Accurately locating each bag solved only part of the problem. Concept also needed a way for the robot to securely lift the 150-pound burlap bags without causing the tearing and product loss experienced with the previous grippers.
A custom end effector was designed with 16 points of contact and pneumatically operated tines. As the tines rotate outward from the center, they move between the burlap threads rather than tearing through the material, allowing the robot to securely lift each bag while maintaining its integrity.
The end effector uses pneumatics rather than onboard sensors or electronics, reducing the need for high-flex cabling and ruggedized sensors. Pneumatic hoses can also be replaced quickly when needed, simplifying maintenance and reducing potential downtime.
The Results
The vision guided robotic workcell removed the infeed bottleneck while improving product handling, safety, and overall system efficiency.
- Doubled bag-handling speed compared with the previous robotic system
- Increased throughput to six 150-pound bags per minute
- Reduced bag tearing and associated product loss
- Improved robotic pick accuracy through real-time 3D vision
- Eliminated the need for the robot to rely on predetermined bag positions
- Created additional infeed capacity to support future production increases
Project Details
Project Duration
About 6 Months
Team
Client: 1 engineer
Concept Systems: 3 engineers
Technology Used
SICK Ranger camera
FANUC robot reprogrammed
Laser scanner
Combine robotics, machine vision, motion, and controls to solve complex manufacturing challenges.