With a diameter of 45 mm, the ball Jonas Bisgaard is holding could easily pass for a golf ball. But a small twist at the centre of the ball separates the two halves and reveals the electronic inner workings.
“We found the best size by experimenting. The ball has to be just the right size to flow freely in a bioindustry production tank. It has to follow the normal flow in the tank. This gives you a realistic picture of the mixing efficiency. This information can be used to adjust the process if the mixing is inadequate,” says Jonas Bisgaard (MSc Eng). He is an industrial PhD student in the Freesense start-up-company, with DTU Chemical Engineering as his academic support base.
Freesense’s product addresses a key problem in biological production. It is difficult to ensure the same mixing efficiency in the industry’s large production tanks as in a laboratory. In industry parlance people talk of gradients arising in the tank. This means areas where key parameters such as temperature, PH value, or oxygen level differ from the target levels. If the conditions are not right, this can lead to lower production yields. Or the process might have to run for longer. Both situations cost bio companies money.
Free movement equals better measurements
Freesense was founded in 2015 and now has six employees. The company grew out of the BIOPRO network, involving a number of Danish biotech companies and participants together with DTU and the University of Copenhagen.
“It’s important that there is a good framework in Denmark for collaboration between companies and universities, and that there are opportunities for start-ups. Novozymes has been involved from the outset, where industrial needs and new technical possibilities involving microcomputers meet,” says Karin Nabizada of Novozymes regarding the company’s commitment to the BIOPRO network.
The days when you poured the various substances into a tank and hoped for the best are long gone. Using the Computational Fluid Dynamics (CFD) method and other techniques, bio companies try to calculate the fluid flows in advance.
The aim is to achieve adequate mixing without overdoing it and incurring unnecessary costs. However, precise CFD simulations are too data intensive to be used in practical production. Simplified simulations are therefore used, which must be validated by measurements.
Many companies have fixed sensors in the tanks for this purpose, but this is not an ideal solution. Not only do they require extensive maintenance, they also only read values at the given location. This led to the idea of Freesense, where the sensor floats around freely. Over time it will visit all parts of the tank.
“With their free movement around in process equipment, the sensors will be a key supplement to existing methods for evaluating large-scale production,” says Karin Nikolajsen.
Inspired by drones
On its journey around the tank, the sensor not only has to measure the desired parameters, but also note where in the tank the measurements came from.
The first information to determine the ball’s position comes from a pressure sensor. The deeper the ball is in the tank, the more liquid there will be above it—and hence the more pressure.
“Information about the depth is of great value alone, because the biggest gradients usually develop between the top and bottom of the tank,” notes Jonas Bisgaard.
But they also want to know the position of the ball horizontally. For this purpose, the company uses a combined accelerometer and gyroscope to determine the speed and angle of the ball. This solution has been taken from drone technology.
“Our sensors move around independently, like drones. We cannot control them remotely, but otherwise the problem is very similar. A drone has to constantly track its location. The same is true of our sensors,” says Jonas Bisgaard.
Measurements in ‘semi-real time’
The size of the sensors is tightly constrained. The permitted weight is proportional to the cube of the radius. In other words, a small reduction in size would mean that the ball would quickly sink to the bottom of the tank, as its density would increase.
“We have to tune both the size and weight precisely to get the desired effect. This is a major challenge in relation to biological production. For example, for a fermentation process that runs for three weeks, the density in the tank will change a lot along the way. This could prevent our sensors from floating correctly,” says Jonas Bisgaard.
The solution is that Freesense pre-adjusts the balls to the desired density. Where there can be even bigger changes in the density of the liquid, you can add more sensors with different densities.
Another challenge relates to the transmission of data. It would require too much energy and hence battery capacity if the sensor had to deliver its data through liquid in real time. Instead, the measurements are stored in the sensor’s memory. Each time the sensor detects that it is on the surface, the data packets are transmitted wirelessly to a receiver mounted on the tank. The disadvantage of the solution is that you cannot know in advance when you will receive your measurements.
“Our trials have shown that you receive the measurements often enough to be able to control production satisfactorily. We call this model ‘semi-real time’. If you feel that data is received too infrequently, you can simply add more sensors to the same tank,” note Jonas Bisgaard.
Tests under real conditions
It is also possible to adjust how frequently the sensors are active. This is just a matter of balancing how long you want the battery to last.
If you measure once a minute, the battery can last for three weeks, which is a typical production time for fermentation. You can also choose much more frequent measurements, but then the battery will only last a few days. Now that the sensor knows where it is and is able to deliver its data, Freesense has a product. The next question is what parameters from production should be measured.
“In version 1.0 we have chosen to include pressure, temperature, and PH value. The next parameter, which we are working on intensively, is the amount of dissolved oxygen in the liquid. This is an extremely interesting factor for the bioindustry. Dissolved oxygen is one of the parameters for which large gradients can occur. But there are several other parameters we are also looking at,” says the industrial PhD student.
“It’s exciting in BIOPRO that the companies trust us and allow us to test the sensors under real conditions. BIOPRO has given us a head start compared to many other start-up companies. The large companies in the network have expressed interest in the product we have now succeeded in developing,” says Jonas Bisgaard.
Next step: big data
It is not yet possible to ‘buy a pack of ten Freesense sensors’, notes Jonas Bisgaard:
“We are still at the stage of developing the sensors in cooperation with our customers. But we have a number of projects and trials underway, both in Denmark and abroad. The customers include BIOPRO members, and other companies that have heard of the product in other ways.”
As a chemical scientist, Jonas Bisgaard cannot help but think about how the new data Freesense will provide can be used:
“The sensors open up exciting possibilities. For example, visualizations that make data easy for plant operators to grasp will be very interesting. It will also be possible to use machine learning and big data to find patterns in the data, in order to optimize production.”