New Extraction Technologies Lining Up to Be Game-Changers: Page 4 of 4

May 15, 2020
Abstract / Synopsis: 

This first “Tech Innovations” column examines the development of various extraction technologies. Extracting from botanicals is an age-old process that dates back hundreds of years, including in processing for cannabis. But now there is more interest in what extractions can do, how they can be accomplished, and what new processes or equipment can be brought to bear. This column focuses on defining extractions, looks at new technologies, including goals and objectives, what can be done with artificial intelligence, and more. We also take a look at other extraction products, techniques, and developments as discussed with various extraction experts who are working both in laboratories and with extraction equipment manufacturers throughout the country.

Andy Joseph, president at Apeks Supercritical, makers of CO2 plant oil extractions systems, is an industrial engineer with five extraction-related patents. Apeks Supercritical was founded in 2001, and started building extraction equipment back in the early 2000s, focusing on botanical oils and flavors. “That got attention from the cannabis industry in 2009,” said Joseph.

“Automation to me is an expected, it’s not an option,” he said. “The cannabis industry early on was the exact opposite. It was a black market environment. The idea of automating something to them was just another way for the cops to get involved.”

Extraction processing without automation presented problems, such as lack of consistency, and differences in outcomes from operator to operator. But automation has its drawbacks as well. “Automation can only take you so far,” Joseph says. “It can automate the ‘process.’ But it doesn’t let the operator see inside the equipment.”

He explained how AI could work in the extraction process: “The CO2 equipment operates at high pressures, 1000 to 5000 psi in a batch-type configuration. We are talking botanical oils,” said Joseph. “So we got a Mother Nature element going. And Mother Nature is never the same.”

Operators program their automated control system to run an extraction at a certain parameter set. “But the best operator with the most consistent material he can find still doesn’t know when the extraction is maximized, when to shift the parameters of the extraction to get the target cannabinoids or molecules that he or she is looking for,” said Joseph. “That is where AI starts to come in.”

With AI, the operator would have the ability to look inside the extraction in real time, with sensors that can detect the differences between molecular weights as the extraction is going on. “Now the operator can say ‘Hey I am ready to switch from the terpenes, the volatile compounds, to cannabinoids.’ The machine says ‘Hey it’s time for me to switch.’ And then it does so automatically. That is the next level of automation, or what we call AI.”

Joseph said that the AI extraction process that they are developing is based on “refraction” technology, which is a trademarked process that denotes a combination of the words “reflection” and “fraction.”

In the process, the operator shoots infrared light into and through the stream of CO2 and the oils mixed in it. The operator then reads the reflection of that light, and crunches the numbers to find out what they are seeing. Fractioning refers to the separation of different molecular weights, such as fractioning off the terpenes.

“We can use refraction technology to drive the decision as to when we stop fractioning terpenes, for example, and start fractioning different cannabinoids and different molecular weights,” said Joseph.

One advantage of this process is for the artisanal cannabis manufacturer. Grabbing and controlling specific molecule weights can make the artisanal cannabis manufacturer more efficient. “This technology gives vision into the equipment while it is operating so the operator can make smart and efficient decisions to create that artisanal product that they want to do,” said Joseph.

Getting the AI process to work will take more time. One issue is understanding the THC molecule in the CO2 process. “As the density of the CO2 changes with changes in pressure, so does the environment around that THC molecule,” said Joseph. “It’s not that it is modifying or changing the molecule itself, but everything around it is changing. It looks different because it reflects light differently. So understanding what it looks like at different temperatures and pressures is part of the calibration process we have to go through in order to implement this technology.”

He said that they are in the beta phase of working out details of the process, and have done trials on materials at different locations to prove that the technology works. Now they have to get the beta piece of equipment into an operator’s hands, which they expect to do sometime later this year. “The problem with beta is that it takes a boatload of work to be able to crunch the numbers we get during the process,” he said. “We are in this calibration phase where we have proven it works, but proven it works over time in different environmental characteristics, like changes in plant material from California to Maine for instance, are calibration sets that we are looking for. We need to prove the technology in different environments to be able to prove that it works consistently no matter where we put it.”

Deeper Dive into AI?

Is machine learning part of this new AI processing?

“We don’t have the technology yet,” said Joseph. “Where the machine says ‘Hey I did this and here are the results and now I am going to change what I do.’ That is a fairly complex process. We are still in the infancy of applying that. But absolutely, that is the direction we are going with this technology.”

MacKay said that if you do principal component analysis, and you do chemometrics, which is a simple part of AI, then yes, the machine can learn. “Machine learning can work as long as it always sees the same product. You have to make sure what your moisture is. You have to have that information that AI can use as a functional variable and then see what happens as it’s impacted.”

He said that the people who are going to “knock it out of the park” in extraction technology are the people who have spent the time looking at the entire process.

“There are places where they do not touch the cannabis from the time it comes in from the truck, to the time it’s an oil ready to be placed in a product,” said MacKay. “It never gets touched. It goes from one mode to another. It comes in and comes out the other side. It’s like Willy Wonka.”


Product reviews such as the present work would not be possible without the contributions and cooperation of the manufacturers that responded to the Cannabis Science and Technology survey. Their effort is greatly appreciated. Although Cannabis Science and Technology has made every attempt to include every submission, it is possible that some have been missed. If there have been omissions or if you want to be sure to be included in the 2021 new extraction coverage, please contact Megan L’Heureux, Editor-in-Chief, Cannabis Science and Technology, at [email protected].