Do you want our news directly to your feed? Follow us on LinkedIn!
Anna Stiwne steps in as new COO at our parent company
This week, we had the privilege of hosting Anna Stiwne, who stepped in as COO at our parent company, Svevik Industri AB, in February.
“It feels great to work with Mabema and all the other fantastic companies in the Svevik group. Mabema is in an incredibly exciting growth phase with great potential ahead,” says Anna during her visit.
Warm welcome Anna!
Two new GPV systems set to be installed during Q2
Our GPV is a fully automated system based on image processing, designed to measure the volume of timber piles transported on trucks from the forests to sawmills or pulp mills. Sweden stands strong as one of the world’s largest and most significant exporters of timber products, alongside global giants such as Canada and the United States of America.
Our eight GPV systems accurately measure over 6 million m3sub*, and with forthcoming installations we are predicting to nearly double this figure within just a few years.
*The term m3sub means ”solid cubic meter under bark” specifying the volume of timber excluding bark and tops.
Vision with Precision – Automated Ammunition Inspection
Vision with Precision – Automated Ammunition Inspection
We have been delivering quality inspection systems to a wide range of customers within the manufacturing industry for over 15 years, ensuring the quality of everything from cinnamon buns to brake discs. We have been inspecting the quality of ammunition since 2008. Today we have automated ammunition inspection machines installed at customers in both Scandinavia and the United States.
From AAI 1.0 to AAI 3.0
In 2016, we received a request from a Scandinavian ammunition manufacturer looking to automate their quality inspection process. And simply, this is where the product we call AAI, or Automated Ammunition Inspection was born. As the first product, it worked well, but there was of course room for improvement. Since the parts slid though the machine, one of the major challenges was the varied friction. It made it difficult to achieve consistent results for certain inspections since we had limited control over the position of the parts within the machine.
A little over a year later, at the end of 2017, we reengineered the mechanics of the machine and AAI 2.0 was introduced to the market. Stops were introduced along the way, for example by image capture and sorting, to gain better control over the whereabouts of the parts. We knew that one solution to gain full control over the parts was to reduce the throughput rate, essentially making the machine slower, though this would not meet the customer’s process requirements. AAI 2.0 machines were sold to manufacturers in Sweden, Norway, Finland, and a few in the United States and most of them are still in operation.
The Crucial Evolution of Part Transportation in AAI 3.0
Today’s AAI, which we refer to as 3.0, features four inspection stations, and we can identify over 30 different types of defects through 360° inspection. One of the significant mechanical upgrades we aimed for when continuing to develop the AAI was how the parts travel within the machine. If they travel sideways instead of lengthwise, the machine can operate at a slower pace without altering the throughput rate. In the production of AAI 3.0 the former transport method relying on gravity and friction was replaced with a conveyor belt where the parts move sideways at an unchanged throughput rate. Unlike the sliding transport method, the conveyor belt gives us full control of the parts whereabouts in the machine.
We have also worked extensively to enable customers to operate and customize the machine for different caliber sizes as smoothly and easily as possible, and we have, and still are, making great progress in that regard. Currently, the most significant challenge is in the software, involving setting the sorting level the customer wants for their inspection. Sorting level encompasses how severe a defect can be before the part is sorted out as not acceptable. Different customers have highly individual opinions on what constitutes an acceptable part, making it difficult to establish a standard level. At the moment, we customize the sorting levels for each customer to find suitable settings with the possibility to adjust them later on.
More and more companies in the manufacturing industry are venturing into automating their quality inspections, and this includes ammunition manufacturers. In the past, all types of quality inspection in ammunition manufacturing were done manually by factory workers. Usually, it involved rotating parts passing by on a conveyor belt, which the worker would examine to determine whether they met quality standards or not. Manual quality inspection works quite well actually, but the assessment can never be entirely objective. At Mabema we believe that automated processes are the future and can guarantee that through automated quality inspection there is consistent and objective evaluation, regardless of the circumstances.
The Machine Guide
The parts enter the machine through a feeder and are then placed one by one on the conveyor belt by a rotating wheel. The parts travel through the machine at a controlled throughput rate of 120 parts per minute. Different types of lighting and cameras are used at the inspection stations depending on the type of defect the station is intended to detect. Technically, we are looking for deviations on a reasonably uniform surface. Since we rotate the part 360° at the final inspection station, only one camera is needed to capture an image of the part’s entire surface. We obtain the surface image by using a line scan camera that captures only one pixel row at a time and then our software stitch them together to a full surface image. Although the final inspection step only technically requires one camera to obtain the surface image, we have installed both a 2D and 3D camera since some defects are more visible in one type of image.
Advancing Timber Measurement with AI
Originally introduced in 2015, Mabma’s GPV system uses advanced camera-laser technology to precisely measure the volume of timber piles as they journey from the dense Swedish forests to sawmills and pulp mills. Our GPV system is at the starting line of an AI-driven transformation with unmatched precision, efficiency, and quality in sight.
Only three percent of Sweden’s 41 million hectares of land is made of urban areas, and a mere seven percent is dedicated to agriculture. A further 22 percent falls under the quite uninformative classification of “Other land”. This leaves a staggering 68 percent of Sweden’s land area covered by dense forests, spanning nearly 28 million hectares¹ — an area that could be hard to imagine.
Sweden stands strong as one of the world’s largest and most significant exporters of timber products, alongside global giants such as Canada and the United States of America². According to gathered data from the Swedish Forest Agency, the net volume (m³sub)∗, referring to the harvested volume of tree trunks utilized either wholly or partially in sawmills or pulp mills, was an impressive 77.1 million m³sub in 2021³. Within this extensive amount of timber harvest, our eight GPV systems accurately measure over 6 million m³sub, and with forthcoming installations we are predicting to nearly double this figure within a few years. Mabema’s GPV is a fully automated system based on image processing, designed to measure the volume of timber stacks transported on trucks from the forests to sawmills or pulp mills. Our first GPV system was put into operation back in 2015.
The measurement process includes six camera-laser pairs, a color line scan camera, and five standard color cameras thoroughly positioned within a 30-meter-long tent. Upon arrival the truck driver registers their piles via a screen positioned in front of the tent. Subsequently, the truck passes through the tent at a controlled speed of maximum 5 kilometers per hour. It is while the truck transports the piles through the tent that the volume measurement takes place. The six camera-laser pairs collectively capture an astonishing 3000 images per second which are then stitched together to a comprehensive 3D image by the GPV system software.
The GPV system provides measurement results that include solid wood volume over and under bark (including height, length, and width) and diameter distribution within the piles. Diameter distribution refers to determining how many of the logs in the timber piles fall within the pre-defined diameter ranges. This information allows sawmills and pulp mills to ascertain the optimal use of the timber in the pile. Logs within specific diameter ranges may be earmarked for distinct processing purposes, for instance logs within a smaller diameter range are not suitable for wide plank production.
Since the initial installation in 2015, our GPV system has undergone a few important updates. Among these is the implementation of color cameras that ensured an entirely new level of accessibility. Images captured by these cameras are sent to a Biometria remote center, precluding the need for manual evaluations of the timber stacks by personnel on-site. While the core volume measurement technology has remained rather unaltered, new 3D cameras with a larger field of view are presently being integrated into the system with the first installation set to occur at Setra’s facility in Skinnskatteberg during the fall. The new more modern 3D cameras will increase the accuracy of our volume measurement and result in better images for the manual controls.
To uphold the reliability of our measurement outcomes, random quality tests are undertaken by Biometria. These involve manual measurement of each log within the stack to yield precise volume and diameter distribution data. Impressively, when comparing the results of the random quality tests with the GPV system’s outcomes, there is a mean error of 0-1 percent. Our GPV system work 24/7 with very few stops and at the biggest sites we measure over 10.000 stacks every month.
Last spring, we initiated a project to research whether we can further enhance our GPV system with the help of AI. After an exciting spring and summer, AI consultant and Postdoc Micael Karlberg and our GPV team, have been able to produce results with greater accuracy than before. Specifically, during this project Micael has focused on exploring whether the images from the color cameras, which currently are only used by Biometria for assessment, can be useful when determining diameter distribution in the GPV system.
The future paves the way for AI to be implemented into more parts of the GPV system. Among the frontiers under consideration for future development projects lies the assessment of timber components designated as waste. In the forestry industry, waste refers to those parts of logs that are unsuitable for processing by sawmills or pulp mills, owing to factors such as irregular roots, log rot, unfavorable shape, and other similar considerations. Implementing this function to our GPV system would eliminate the need of manual work for personnel at Biometria and would lead to the first ever fully automated system for volume measurement of timber piles.
“There is no definite end to industry automation in general. You could basically automate endlessly.” says Micael Karlberg. At Mabema we believe the relationship between AI technology and the timber industry in Sweden will continue to evolve and with GPV, we want to be at the forefront of that evolution.
¹(SCB, 2023)
∗The term m3sub meaning ”solid cubic meter under bark” specifying the volume of timber excluding bark and tops.
Is AI really a game-changer?
Artificial Intelligence (AI) has become a buzzword in almost every industry, including manufacturing. With the potential to improve efficiency, reduce costs, and enhance product quality, it’s no wonder that more and more industrial companies are investing in AI. And, as AI continues to evolve, it is evident that the manufacturing industry will continue to undergo significant changes. However, is AI really a game-changer?
Given the potential to increase efficiency, reduce costs, and improve product quality it is not surprising that more and more industrial companies are realizing that AI is, in fact, the future. In addition, according to a 2023 study conducted by Deloitte1, companies that have invested in emerging technologies were able to adapt better to the challenges brought on by the COVID-19 pandemic.
Artificial Intelligence (AI) has already begun to revolutionize the manufacturing industry in various ways. For machine vision purposes the development of AI, and particularly deep learning, has progressed rapidly over the last few years. Several machine vision software suppliers now have their own deep learning implementation, either as stand-alone software or as a supplement to their standard machine vision software. Meanwhile, there has also been a significant improvement of development environments, such as Google’s Tensorflow. Even though AI has been used within machine vision for some years now, many possibilities are yet to be explored.
When AI Meets Machine Vision
“Initially, we hoped that AI would be the solution to all of our problems.”, says Anders Wiklund, Business Area Manager at Mabema. We have followed the development of AI for machine vision purposes for a long time and about five years ago, we decided to really try to adopt and use the technique. Although the expectations were rather high due to some good demonstrations and tests, we did not really know what to expect.
Despite the fact that we have worked on some projects solely based on AI technology, the majority of machine vision systems we deliver to our customers uses traditional machine vision technologies only or with AI as supplement. Our machine vision systems are, among other things, used for classification, quality control and text reading in various industries such as food, automotive, and construction. For us, some of the biggest challenges, when it comes to AI, have been extensive annotation and establishing effective communication between our machine vision systems and the Python-based AI systems.
“We are always looking to solve our customers’ problems in the most efficient way possible and in some projects, traditional machine vision technology has worked better than AI.”, says Anders.
Navigating the Future of AI
During the last five years we have explored the many possibilities of using AI in machine vision. Now, we have decided to further strengthen our team with a new AI-expert, Micael Karlberg. He has both worked and did his postdoctoral degree within the field of AI and machine vision.
Micael sees the future with AI as a bit overwhelming. As technology evolves, one could say that we find ourselves in different technological eras.
”I’m not sure where we are on that journey, whether we’re at the beginning, in the middle, or at the end. If we’re at the beginning, a lot of things will change in the future.”, says Micael. He believes the development of AI relies on implementation and application, but that theories and principles will essentially be the same.
Considering the progress that AI has made in recent years, we at Mabema are confident that we will continue to use AI in our solutions. We look forward to further exploring how we can implement AI to become even better suppliers of state-of-the-art machine vision solutions. However, we do not expect to completely replace traditional machine vision technology with AI, but rather to use it as a complement to an even greater extent.
So, while AI may not be the answer to all problems, it does have great potential to be a game-changer.
The Future of Industrial Automation
These are challenging times, with great concern around the world. The last three years we had to experience the COVID-19 pandemic and in the beginning of last year a devastating war broke out in Ukraine. Unfortunately, these are challenges we must continue to confront in 2023. Despite the current situation in the world, there are big business opportunities. Sales of robotic automation have reached unprecedented levels with companies like Mabema capitalizing on the trend by increasing the number of cameras per sold robot. During 2022 we have carried out numerous replacements of outdated vision systems for customers seeking greater machine availability, a trend we think will continue in 2023.
As we move forward into the new year, we have the ambition to maintain our momentum and set new sales records. Simultaneously, we aim to accelerate the development of our product FINDER, which will feature additional smart functions and new cameras to better meet our customers’ needs. As the market grows it will be more and more important to put the customer in the center of product development.
While market uncertainty and an anticipated recession in traditional industries may have a negative impact, sectors such as energy and electrification are performing better than ever before. Regardless of the industry, it is now more crucial than ever to optimize production costs through increased automation.
We look forward to continued successful projects with our partners and customers throughout 2023.
Text: Peter Andersson, Head of Robot Vision
“There was no business idea.” – Get to know our CEO & Founder Magnus Lundberg
In 2001, with partially finished Master studies and his elite career in canoeing behind him, Magnus received a call about a summer job. A summer job that would turn out to be the starting point for Mabema.
Canoeing, drive, and determination.
When I asked Magnus about the peak moments of his canoeing career, he humbly claimed that he feels he had few to speak of. But to succeed in canoeing takes strength and will with grueling training and the lack of shortcuts. Despite his self-assessment, he had earned several National Cup medals, a Nordic Cup gold, an unofficial team World Cup gold, and he had the opportunity to participate in several world cups. A quick Google search revealed that Magnus was even mentioned in the Växjö canoe club’s history book six times – which I think is an achievement in itself.
However, Magnus’ canoeing career ended when he missed out on the Olympic qualification in 2000. Many people told him it was too early to quit at the age of 26, claiming he was just about to hit his stride. And Magnus admits, it may have been a bit too early.
But despite ending his canoeing career, the drive and determination have stayed, from elite sports to entrepreneurship. I quote Magnus:
“Coming in second in sports is really bad and in sales it’s completely worthless.” . So yes, the competitive spirit is still there.
The starting point for Mabema.
With his canoeing career behind him, Magnus made the decision to complete his Master degree in Applied Physics and Electrical Engineering in 2001. But it was during the spring of that same year, that a call came in offering him a summer job at Forsmark nuclear power plant, where he would repair and renovate radiation-resistant underwater cameras. Said and done. The following summer, Forsmark nuclear power plant invited him back. That is when Mabema was born.
“The first year I worked for another company, but in 2002 it was not possible to solve it in any other way than starting my own business.”
Starting off as a solo venture, Magnus was the only employee at Mabema. As the business began to grow, he sought the help of a friend, former study and canoeing buddy Pelle. The duo worked together, selling their services by the hour, repairing cameras all summer long. There was no business idea. Mabema was nothing more than a well-paid summer job alongside his Master studies at Linköping University.
His master thesis was finally written, presented and approved. The year was 2004 and Magnus found himself at a crossroads. He had the opportunity to either pursue a doctorate in Medical Image Analysis or focus solely on Mabema. And there was a market need, a business opportunity for Mabema.
Full time and all in.
When Autoliv and Carbex, two big names in the industry, became Mabema’s first real customers, it was a momentous occasion. It was clear that Mabema was making progress, not at lightning speed, but it was moving forward steadily. For Magnus, the slower growth had its own set of advantages. Without investor money, there was no room for failure and every step had to be thoroughly calculated and well executed. Every move had to be right, most of the time at least.
Despite the consequences of failing, it is the major breakthroughs that Magnus can look back on and be proud of. The projects that seemed pitch black in the beginning but were eventually still solved in the end. A satisfied customer is a successful project.
For Magnus, the challenge of solving a customer’s problem is what ignites his passion. The thrill of getting to know a customer, understanding their needs and practically starting to see the solution during the customer visit is what drives him. Above all, it is about being able to clearly communicate and explain to the customers exactly what they are buying, even when the solution may be advanced and technical. This ability to connect with customers and find solutions to their problems is a great source of inspiration to Magnus.
But when it comes to the most fun thing about running and working at Mabema, he is certain:
“You can come into work on a Monday and feel a little bit off, but when there is energy flowing in the office, it spreads. I try to spread energy to others too, it’s important.”
Mabema AAI 3.0 – High quality Automated Ammunition Inspection
We proudly introduce the Mabema AAI 3.0, an improved and more superior inspection machine for ammunition. The Mabema AAI 3.0 increases manufacturers process control, quality and profitability by providing cost effective and reliable quality inspections. With a fixed throughput rate of 120 parts per minute, world leading vision technology and high-resolution cameras the Mabema AAI 3.0 carries out a full inspection of small arms cases, cartridges and bullets.
Since 2014 Mabema has delivered eight machines for automated ammunition inspection to both Swedish and international customers. Based on the feedback from our customers and updates in the machine vision industry we have developed the Mabema AAI 3.0 – a more superior machine for inspection and measuring of small arms cases, cartridges and bullets.
Our new automated inspection machine carries out high precision measurement and inspection of all critical features of the part such as the extractor groove, primer, bullet or mouth as well as both 2D- and 3D inspections of the body. Some examples of defects that the machine detects are wrinkles, scratches, dents along with various defects on the extractor groove, primer and bottom stamp. One of the key benefits of the Mabema AAI 3.0 is the surface inspection. It is carried out with line scan cameras and special LED-lights that enables the machine to measure and sort out scratches, wrinkles and especially soft dents on the body.
The Mabema AAI 3.0 has a fixed throughput rate of 120 parts per minute, which is considerably faster than the earlier versions of the machine. The parts travel through the machine on a conveyor belt which results in less stops and higher reliability than before. Defect parts are blown out after each inspection station with the help of compressed air. This makes it easier to organize the parts depending on the defect category.
This January we delivered the first Mabema AAI 3.0 to an international customer and a few weeks ago the second machine was delivered to a customer in Sweden. At the time of writing we are building the third and fourth machine which are to be delivered to both new and existing US customers.
Before the second Mabema AAI 3.0 was sent off to our customer we took the opportunity to record a video of it. The result became an introduction of our new automated ammunition inspection machine, the Mabema AAI 3.0.
Video from 2023.