Optical sorting technology helps material recovery facilities (MRFs) reduce their reliance on manpower to sort various materials. They are masters at sorting efficiently and quickly.
Yet, as do human sorters, optical sorters have some limitations. For instance, an optical sorter is a pro at sorting polyethylene terephthalate (PET) from high-density polyethylene (HDPE). Yet, the optical sorter starts to run into problems when sorting nonfood-grade HDPE from food-grade HDPE.
“Most mixed rigids are made of HDPE, and mixed rigid plastic has a much lower value than what pure HDPE has in a container form,” says Thomas Brooks, director of technology development at Bulk Handling Systems (BHS), Eugene, Oregon. BHS is the parent company of both optical sorter provider National Recovery Technologies (NRT) and the artificial intelligence and robotics technology Max-AI. Brooks adds that traditional optical sorters are not able differentiate between these two materials.
However, this limitation and others can be overcome when optical sorters are paired with artificial intelligence (AI) technology.
Daniel Bender, technical manager at Germany-based Tomra Sorting GmbH, says all optical sorters have a very basic level of AI, or machine learning, using near-infrared (NIR) technology and cameras. He says AI takes that machine learning a step further by offering deep-learning skills and the ability to analyze more data.
“AI is very much a tool that elevates optical sorters and provides a way to sort more specifically,” adds Mark Baybutt, vice president of product at Denver-based Amp Robotics. “AI complements traditional optical sorters in that it allows it to select materials in a more refined way.”
Going back to the example of sorting mixed rigids from HDPE bottles, Brooks says AI can solve that problem for optical sorters. “[Optical sorters] purely see what the chemical composition is but not other things, and that’s where AI sees a huge dividend,” he says, adding that pairing AI with optical sorters can help create new opportunities.
How it works
Coupling optical sorting with AI provides MRFs a new level of flexibility. Both technologies have their specialties. Optical sorters are speedy and handle basic sorting but can miss some objects; AI and robotics technologies are slower, but they are great at quality control applications and providing additional data on the material stream.
Overall, optical sorters do an adequate job with sorting, according to sources, but incorporating AI takes that sorting to the next level. Certain sorting tasks can be challenging for the standard NIR technology on optical sorters.
“NIR measures the absorption of the different materials in the NIR wavelength spectrums,” says Felix Hottenstein, sales director at Nashville, Tennessee-based MSS Inc., a division of San Diego-based CP Group. “It can tell different plastics and materials apart. But then the selling point of AI is that it functions like the human brain. [AI] knows if something is a PET thermoform or a PET bottle or if it’s a round lid versus another round object, just like human sorters on the line.
“And AI can go to a much more granular level with things where it can add more data if required.”
Brooks of BHS says using these two tools in tandem can enable MRF operators to be more specific about what they sort.
“If you think about an optical sorter being able to discern between reds, blues and greens, then the artificial intelligence can look within those colors and determine the shape,” he says. “This is a circle and it’s red, or this is a square and it’s red. You’re able to combine the tools together in a way that you only end up with blue circles, red squares, etc. That’s what AI provides to the optical sorter.”
Bender of Tomra says the result of pairing these two technologies is higher purity and higher recovery of wanted materials.
Where it works
MRF operators need to be strategic about where they apply AI and robotics to get the most out of the investment.
Baybutt of Amp Robotics says one of the best applications for AI when paired with optical sorters is on the quality control line.
“Optical sorters can sort the majority of your material, but material on various streams necessitate humans to perform quality control. That’s where we see AI helps—predominantly to assist on quality control lines to improve efficiency.”
Adding AI to a quality control position can help MRFs reduce their headcounts, as well. Brooks says one BHS customer was able to reduce its headcount on the quality control line by two people by adding AI.
“In that application, we combined NIR and artificial intelligence to sort HDPE from mixed rigid plastics,” he says. “In this case, we were able to add the AI to the sorter and basically reduce the headcount needed to postsort the material stream. We were able to reduce their headcount by two people. So, going from three people to one basically to quality control that line was mostly possible because NIR was coupled with AI—it could make a complex decision and provide a manpower savings.”
Bender adds that it’s important to apply AI on tasks that are “superhuman”—the tasks that can’t be performed by a manual sorter or a person on the quality control line. For example, he says it’s tough for a human to distinguish certain types of plastics from one another. AI’s deep-learning capabilities can process this kind of information and sort as well as a human or better in most cases.
When it doesn’t make sense
Some operations may be tempted to jump quickly to embrace AI and robotics, but manufacturers caution operators to curb their enthusiasm and analyze the pros and cons before adding AI to help improve sorting.
Most manufacturers agree: AI doesn’t solve all of a MRF’s problems.
Alex Wolf, director of technology at Norwalk, Connecticut-based Van Dyk Recycling Solutions, says AI might not be worth the investment if improved quality is not a concern for the MRF.
“If you are happy with the sorted product that comes off your optical sorter, considering that it is not overrun, then adding AI won’t bring you any benefit on sorting performance,” he says.
Also, Bender of Tomra and Brooks at BHS say AI doesn’t make sense if it’s only used for a single, specialized task.
Weighing the cost of adding AI is another important consideration. Hottenstein says it’s smart to invest in AI if it enables the MRF to reduce manpower and save money that way. However, he continues, it makes less economic sense to invest in this technology if AI only provides a slight boost in statistical data on material streams with no real improvements in sorting. “Like with any equipment, you have to investigate and evaluate each potential application to see if the advantages and potential added values outweigh the investment and monthly maintenance costs,” he says.
Tips and takeaways
Cost considerations aside, AI typically helps MRF operations to step up their sorting game, leading to cleaner end products.
It also doesn’t require much maintenance, making it easy to manage.
“Maintenance on AI is not much different than it is for optical sorters,” Wolf of Van Dyk says. “At some point, it may need trained or updated on new material that shows up in the stream, but that is typically done in conjunction with a service provider.”
Material streams are constantly evolving at MRFs, and AI enables these plants to see data in real-time and respond quickly to those changes.
“AI at its core allows for a fast response to changes,” Baybutt says.
With the COVID-19 pandemic, Brooks adds that there have been odd changes to material streams at MRFs, such as an influx in rubber gloves, masks and personal protective equipment (PPE). He says AI can recognize these contaminants without much training.
“AI and robotic sorters can recognize [PPE] as contamination and remove those,” he says. “They don’t require much training to do that; they are in a mindset of saying, ‘That’s not supposed to be he here.’”
Also, the COVID-19 pandemic has forced some MRFs to reduce their headcounts, and Baybutt says AI and robotics can help fill the roles of some missing personnel.
“There are many facilities that have embraced these technologies and have showcased that they work and add to their operations,” he says. “We’re seeing an upswing in this technology. I think if customers are hesitant, it may be due to a lack of familiarity. But the technology is not too complicated—it’s deployed and it’s working.”