7 min read
Breadmaking is one of the world’s oldest commercial activities. Bread products are the largest in the Bread & Cereal market, with an estimated total revenue of $23.8 billion by the end of 2022.
The market consists of industrial and commercial bakeries of different scales that produce almost 6.8 billion kg of bread in 2021, which amounts to an average bread consumption of 20.4 kg (44 lbs.) per person annually.
About this automation journey
The fifth installment of “Small Automation, Big Impacts” is a collaboration between ABI LTD and Hermary. ABI is a Canadian company that has been a leading brand in industrial bakery automation for 40 years. It prides itself on providing premium automation systems that bring value to customers’ bakeries. In this episode, they are sharing one of their automation journeys made possible by Hermary’s 3D machine vision.
Modern industrial bakeries rely on technologies to automate bread production, storage, quality control, and packaging. Before entering the oven, most types of dough would have to be scored.
Scoring the dough is a delicate task, traditionally performed by workers who use a specialized knife, called a lame, to make incisions across the top of the dough. Scoring adds aesthetics to the finished products — baguette scores are perhaps the most prolific. This step also helps carbon dioxide escape during the baking stage, preventing the dough from erupting into undesired shapes. To produce uniform loaves, workers must perform the task with great consistency.
The evolutions of the scoring process
Mechanical machines were implemented to alleviate labor shortages and reduce the inconsistencies associated with manual work. However, human supervision was still necessary to ensure the scoring machine lined up with the dough. There was no way for the machine to detect where each dough was. As with every piece of mechanical equipment, wear and tear over time can cause gradual changes to the machine. Equipment calibrations had to be scheduled regularly or whenever operators noticed undesirable scoring marks.
2D machine vision
2D machine vision was equipped to solve the blind mechanical process by feeding the dough positions on the conveyor belt to the incision robots. However, the system could not accurately adjust each scoring knife’s vertical distance in relation to the dough. Without high-quality height data, the blades were unable to create uniform scoring marks across the dough’s surface.
Furthermore, flour falling onto the conveyor belt impeded the system’s ability to separate target objects from their background (inadequate scene contrast), making the results prone to error. 2D image processing also requires time to parse through the pixels causing the conveyor system to slow down at this stage.
Off-the-shelf machine vision
While off-the-shelf machine vision cameras have made many industrial applications or even space travel possible, engineers at ABI found their software limiting and parameters difficult to adjust. Moreover, they could not manipulate the data to obtain satisfactory scoring results. Eventually, without adequate technical support, ABI deemed off-the-shelf, black-box devices unable to help them deliver their promises to the end users.
Small Automation, Big Impacts is a collection of industrial automation journeys made possible by 3D machine vision.
We believe sharing these stories is a way to stay connected with our partners and positively impact the manufacturing community.
This free eBook also includes a bonus chapter to help automation professionals easily decide when to use 2D and 3D machine vision.
All this is just one click away in our new eBook, Small Automation, Big Impacts.
Automating with 3D machine vision
- High-resolution point cloud data that are unaffected by the contrast with the belt color,
- Each dough’s relative position on the conveyor belt,
- Digitized representations of each bread completed with their height information, and
- Fast scan rate that does not thwart the conveyor belt speed.
Since laser triangulation scanners rely on capturing the diffused beams of an object to capture its 3D profiles, the contrast (color difference) between the target object and the conveyor belt color is irrelevant to the scanned data’s integrity. Additionally, mathematical relationships can be derived to filter out background noise or unwanted data, shortening data processing time.
Before the robotic blades make the incisions, the conveyor belt passes through a scan zone with an SL-1880 scanner directly above it. As each piece of dough passes through the laser line, the scanner generates high-resolution XY point cloud data. The system engineers designed proprietary algorithms to process the point clouds and translate the data into the robot’s coordinate system. The scoring system can thus move the blade into the exact position where the dough is.
Hermary’s SL-1880 can generate up to 812,500 points per second. System engineers at ABI use captured data and proprietary algorithms to calculate a correct path of incision tailored to each dough’s unique curved surface. Additionally, adjustments can be made in the software to accommodate different types of bread, giving users flexibility. For example, baguettes require diagonal cuts, and boules (round bread) typically sport a cross on top.
To accommodate wider conveyor systems, multiple SL-1880 scanners can also be installed to help speed up production. The SL-1880 machine vision scanner is designed to scan up to 1,250 scans per second, enabling the robotic blades to make an average of three incisions per second on baguette dough. Not only is this a significant efficiency and consistency improvement, but the accuracy and precision rates are also estimated to be 100%. The fast scan rate is a future-proofing feature that gives system engineers room for adjustments to match faster conveyor speeds.
The journey continues
In McKinsey’s 2022 Technology Trends Outlook report, Industrialized Machine Learning (ML) is poised to change many manufacturing sectors. The use of this technology will provide a better understanding of what needs to be done for a specific action or situation, which in turn will give better insights to manufacturers on how to change their operations and maintenance schedules.
Savvy system integrators have started using 3D point cloud data in machine learning for better scene segmentation and object identification. The American Bakers Association also encourages production facilities to adopt artificial intelligence to improve sustainability and advance the industry. For ABI, they also see machine learning spearheading equipment longevity and increasing production while preserving the artisanal touch.
Big impacts of small automation
Industry: Industrial bakeries
- Eliminate humans working near hot baking equipment
- Reduce human fatigue and risk of ergonomic injuries
- Accurate height information in milliseconds
- Consistent score marks
- Accurate incision paths
- Reliable scan data
- Flexible scoring patterns
- Upskilling workers to perform higher value-added tasks
- High-speed scanners are future-proofed to accommodate faster conveyor speeds to come
- Data to be used for machine learning
- Vision-guided robots to help with more baking automation
- Minimize waste from unqualified products
- Increased system performance
- Outstanding customer satisfaction
- Proprietary system design
- Inimitable competitive advantages