Optimizing Industrial Bioreactor Design for Biofuel Production

[00:00:01]okay so good morning everybody it's me again okay uh I'm going to talk about our CIO project number seven which is closely related to wp7 from Sun and uh well the the subject of this talk is also related to the to yesterday's talk so I'm present you again this this slide which is the the the the evolved processing diagram that we are proposing for for sugar cane processing industry but actually it applies to to uh other biomass sources okay so remember we were we were we are proposing a pre-treatment followed by hydrolization processes in order to free up sugars so it involves bioreactors here and also uh a rea reor for enzyme production which can be on site or somewhere else and we also proposing to recycle nutrients from ven by processing uh this product uh through a photobi reactor okay so uh I'm going to talk about efficient large scale uh bioreactors which are applicable here here uh and also in enzyme Hydro Iz ation and production so now we're focusing I want to focus on the development of high efficiency industrial bi reactors and in particular I want to show and explain some upscaling analysis I cannot avoid to show you some equations but don't bother with the language it's very simple we have three uh conservation equations Mass momentum and energy uh again don't don't bother by by the by the language if you're not used to because uh what s what is stated here is very simple mass is to be conserved this is an equilibrium of forces this is acceleration these are gravity forces pressure forces and visous forces so it all all has to balance and also energy equation says that the the inteno rate of uh energy variation is uh equal to the energy transported by the flow and also energy that is coming by by conduction in inside the the flow okay so these equations uh are very simple and uh but they are sufficient to show you what I want to show uh well they they presume that you have an evisit flow and compressible and uh uh this is happening inside the bio reactor okay so the idea to see how how Dimension affects the the problem is to transform these dimensional equations into non-dimensional ones by uh substituting these variables velocity temperature pressure by uh the corresponding non-dimensional ones so if if you have a generic variable a it has a dimension it's can be in meters or me/ second so the idea is to replace uh a by a non-dimensional it's non-dimensional version which is calculated by dividing it by a reference Dimension capital A for instance okay so if we do that for instance if we we we elect uh capital u as a reference velocity and we create this one which is the corresponding non-dimensional version if you do that I'm coming back one slide if we replace that in all these equations we're going to get these new equations here okay uh well the indexes here uh refer to the common Einstein's notation but what's interesting to show is that mass conservation equation is is the same momentum equation well let me let me highlight here momentum equation you're going to get uh this uh new numbers new new variables or new non-dimensional numbers fruit number you get raino number here in front of of the viscous uh term and in the energy equation you get Str number R notes and penal number so you have this uh equations that states Mass conservation momentum conservation and energy conservation but uh and in non-dimensional variables so you don't have a problem with the scale but you have this non-dimensional number which are def find it here okay and they actually determine the behavior at the scale of your problem so you have for instance true number which is a quotient between transient effects and inertia effects so it's important if you have if you have uh transient heat transfer I'm coming back one slide here so it it affects only the energy conservation equation okay you have Ray noes number which is a quent between inertia terms and viscosity terms it affects I'm coming back one slide the momentum equation and also the energy equation and you have fruit number which is important if you have a free surface in your problem okay and uh because it states a balance between inertia and gravity effects and the print note number uh which is a cent between momentum diffusion and thermal diffusion and it's intrinsic to the fluid okay it depends only on thermophysical properties of the fluid uh so these these numbers determine the behavior at the scale that you that you want to study your problem and the the scale parameter here is D which can be the diameter of your photo of your reactor or whatever okay so an important consequence of this is that similar geometries May respond completely different to the same boundary conditions depending on these non-dimensional numbers so you may have exactly the same flow rates and so on and your system may be completely different I'm going to give you a very simple example an isothermal flow past the circular cylinder so I'm coming back some slides here okay so it's isothermal so you don't have the energy equation you don't need to enforce it so you have only mass and momentum and what we going to vary is uh the ROM's number okay so we're going to vary the diameter of the of the cylinder so by doing that I'm varing actually the Ron's number so it's only the viscosity term the momentum equation that that's being varied okay at very uh low rain Nots number you have the flow from the from from left to right and you have a nice uh wake behind the cylinder what's being uh plot here is the the angular velocity of the flow so uh blue in dark blue is the flow is almost uh has no rotation and what is uh white and red you have some angular velocity of the flow so if this is uh a photobi reactor for instance and remember that when you have uh vorticity you have sheer stresses are very important so if uh this is a photobi reactor any cell that passes through these red regions may may may die because they will be they will be torn apart by these regions okay so by simply increasing the rain not number to to 110 so it's the same uh same geometry same velocity here everything is the same except that we changed the diameter so we changed the rain notes number now you can see that uh inertia terms of balancing uh Vis viscous dissipation so the flow wants to oscillate but you have viscous dissipation so it's a kind of a dynamic equilibrium like this and you have this uh oscillation this is exactly why a flag oscillates in the wind okay and uh again if you have cells within this this medium well they are going to have a hard time only in this uh regions here okay where you have high high vertices and by increasing a little further the rain's number the behavior is completely different now you see that uh you have these vertices being being uh released and everyone with high high verticity so you cells are dying here actually okay so remember this is exactly the same problem what's being changed is the scale of the problem by changing the diameter of the problem so if I have a small photobio reactor represented by the first example cells will will have a good time in this small photobi reactor and if I increase the scale I have uh regions very dangerous region where cells may die within them okay so this is uh this is what happens when you increase the scale of the problem and this actually happens at every scale this is a satellite picture from uh from the islands of Cape V okay you see these wakes behind the island so it really actually the Jupiter's spot is also a a kind of a Vortex like this okay so we we did some studies in a we did some we elected a test case which we wanted to to develop to increase uh its efficiency we took a we trying to develop continuous uh continuous fermentor they are actually used in practice uh they are very simple to operate they're very uh inexpensive to build but they are still very inefficient at these scales those ones are have a slightly different configuration than the the previous one I showed the input is over here then the flow circulates within them then it goes up and the output of it's behind the the the the surface here and goes to the other one usually you have two or uh three or four reactors like this in which you can control uh you can impose different fermentation conditions different temperatures in order to to increase um increase efficiency but remember fermenters uh continues or the batch versions are still the the the least efficient processes in the overall process so increasing their efficiency has a very important impact in the overall uh efficiency of the conversion problem and what we are trying to what we did in this uh test case we try to assess the the residence time distribution the concept of residence time is important because well it's defined by what we call an ideal or hypothetical residence time which is determined by the time necessary for the biochemical reactor to biochemical reaction to take place okay so it's not uh it's imposed by Nature let's say and what we do usually we simply calculate in a hypothetical I I would say say or theoretical residence Time by dividing the bioreactor volume by the inut input flow rates and one of the problem is that this residence time has nothing to do with the actual even with the average residence time I'm going to show you because uh depending on each flow line that a particle follows you have a specific residence time okay so uh consider this example here you have a flow line that that passes right over the the the the guy here and if you calculate this specific residence time over this line is you just uh perform this integration okay you calculate the distance divided by the velocity velocity is always uh tangential to the to the to the flow line okay so what you have here is actually a resident time distribution not a single resonance time and that applies also to a to a reactor uh and so our idea was to use a cfd uh platform to to make this uh upscaling studies and in this particular case we wanted to to optimize the residence time okay so just some examples here uh this is a 10 cubic M continuous reactor input is here and the the output is uh is over here okay so the flow SS around this Central tube here and you can see the the flow lines and you can imagine that Associated to this flow lines you have uh a residance Time distribution which I'm going to show you just in a few slides okay just by increasing the volume of this reactor you see that now you have a the same J the same uh the same flow rate or or the same theoretical residence time but you see that the stream lines are completely different completely completely different just by changing the volume of the the reactor and you can figure out that the corresponding residence timee distribution is very different okay so what can happen when we calculate these residents uh time distribution you see things like this first of all here let's say it's your theoretical residance time calculated by dividing the volume by the input flow rate okay here you have your actual average res residence time which is an average determined by this dist distribution here okay and you can see things like this you have short circuits those are particles that take a a stream light that goes straight to the to the output so it spends uh the particles here spends less time than the necessary time for the biochemical reaction to take place okay and you also have this region here in which the the corresponding particles stays longer than the necessary time okay so this effect and this effect actually uh contri Utes to the det deterioration of the performance of the the the device okay you also have and you may measure this uh these things you can calculate how how the distribution is spread around the the average value and you can also calculate the difference between the actual average residence time and the theoretical residence time because when you when you design your reactor in the first place you have this time here which is determined by the necessary uh time for the reaction to take place so this is our designing machine or upscaling machine it's a cfd platform I show you this transparency yesterday okay so we we have these geometry parameters we have a mash generator and we have a cfd solver so the cfd solves uh MH momentum and energy equations and you have from this solution you can calculate your optimization parameters which can be as I showed you the difference between the residence time the actual residence time and the average uh in the theoretical residance time but you can also uh calculate if you're designing a photobi reactor you can calculate the sheer stress and so on so then by by analyzing this parameters you can uh propose corrections to the to the geometry parameters and you change your your reactor and the process go on until you have an Optimum design okay so I'm going to show you some uh results due to ai ai was at Cambridge last year so this is just an example just to show you the the the influence of the input angle okay so you have uh the height here is 120 CM it it's fixed just to to show you the influence of the input angle you have uh at 90° okay 90° is the input angle here you have this uh purple distribution you see a huge uh short circuit and so that's it this is the worst distribution oh uh an important thing is that the actual residance time is completely different compared with the hypothetical one and by changing the angle you see that these distributions they collapse into uh an Optimum one here so the the the cfd platform is changing automatically the the the geometry of the the reactor in order to enforce a thinner residence time distribution okay this ideas can also be applied to a photobi reactor in which you have to enhance light propagation so we started by by considering this prospective configuration it's a tube photo reactor it's very good for for illumination because you have a huge interfacial area but you have a poor interfacial area for for gas uh exchanges because well the area that you you have is is the section here and the section here so it's very poor and as a consequence you have a poor temperature control actually we uh considering this configuration it's a airlift by reactor something like this you have a bubble colume here light is being is entering the the the volume by the external surface here okay and you have a flow like this induced by the bubbles so what's interesting is that you have an excellent interfacial area for gas exchanges for photosynthetic O2 renewal and for CO2 Supply okay because it's related to interfacial areas of the bubbles you can also control the dark light photo period which is important for for for cell growth okay but in this case you have a poor interfacial area for external illumination at Large Scale because if you increase the scale your your volume increases more rapidly than the the external surface so you have uh comparatively uh less elimination area so what we are considering is uh slightly different configuration we start by by this what we call culture cell we have illumination pole at the center around which there's a a bubble colum so you still have this Vortex here so you you still have the possibility of controlling the photo period but the upscaling process is is done like this you have several of this culture modules here uh arranged in uh Rings like this okay so we did some light propagation studies this is the beer Lambert law for a cylindrical geometry okay or is the distance from a specific point to the light source okay and uh well the the total light intensity at a given position is given by the sum of all the light sources I'm coming back on the slide so if you are at a specific position here you have to uh add all the light sources which are at different at different positions okay so we can plot the light intensities what we did here we just calculated light intensity for for a single illumination pole which is this red curve here okay and for the proposed geometry we calculate the light intensity along this line here it's r0 and along this line here it's R 30 denoted by by this lines here so the red one is for a single illumination pole the blue line is along this horizontal axis here and the red line is the illumination along this line here okay and you see that uh if if we consider the single illumination pole case and if you consider that the above 90 or 80% the the of the light intensity you have photo inhibition here so the cells want to protect themselves so they don't grow and above uh 40% you have not enough light for for the cells to grow so the the useful region here for the single elimination region is uh indicated by this red rectangle here okay so it's it's the region between 80% and 40% if we do the same for for the horizontal line here you see that the the the useful region it's much it's much bigger okay from from this position to this position okay now here you have photo inhibition and along the the horizontal line the r zero region you have a much bigger uh uh light intensities or a much bigger useful region okay so if you consider now that you have uh algae within the the the broth the culture broth and you can calculate the the the light intensities by uh considering the light attenuation due to the the the water algae mixture and due to the bubbles okay to the bubbles so this is this this formula corrects for the the presence of of algae and Bubbles within the culture broth and what we did to determine how uh how these things work we did we used our cfd I know if I have to click here yeah maybe and what we did we calculate the void fraction at this position here and we adjusted an average void fraction curve which can be the by this formula so we adjusted these parameters this one this one and this one and then we coming back one slide we introduced them in this formulas here okay here and here we can calculate the the actual light intensities this is this is this red curve here and also the the average bubble vo fraction the red curve and the average light plus out V fraction here okay and we can calculate those same production regions okay so you see that very close to the light source you have photo inhibition so you have a small uh useful smaller useful region far away from the illumination you don't have enough light and you have this uh useful let let's say useful region for cell growth okay so what we did we put all these uh calculations into our designing machine that previous uh slide that I showed you where you have the the cfd platform and the optimization method and we we we designed a photobi reactor uh based on this on this procedure what we're now doing we are building um a photoo reactor a 10 cuic meter photobi reactor in in order to to verify to validate all these these calculations okay it's also going to be uh instrumented we're going to have light sensors and so on and we want to to test all this in order to see if we we we can obtain High algae concentration at large scales because well in if you look at the literature you see that high very high concentrations are something around 5 G per liter 10 G per liter and uh but very small reactors we want to achieve this is our goal this concentrations at these scales okay so conclusions and perspectives uh an optimal design method was successfully developed uh that is an iterative procedure based on parametic modeling of the flow so we did this optimization for fermentor and also for phot rea reactor and we with uh transfer of Technology uh already started and we intend now as I said to build a laboratory SL pilot scale small plan to test and we also want to do a larger one possibly a 100 cubic meter photobi reactor multiphase flow equations are already been introduced at this moment and we want to actually uh Implement a multiobjective optimization procedure by optim by optimizing at the same time she stresses segmentation the difference between the actual and the hypothetical residence time and so on so we need to implement a Paro Frontier search method in that design machine okay so that's it thank you very much