Latest / Mechanical Engineering Made Simple / (#84) How Mixing Failures Kill Life-Saving Drugs

Transcript

(#84) How Mixing Failures Kill Life-Saving Drugs · Mechanical Engineering Made Simple

• 0:00
  Welcome back to Mechanical Engineering Made Simple.
• 0:02
  Today we are going to talk about something that sounds, well,
• 0:05
  incredibly mundane on the surface, but is actually
• 0:08
  responsible for the difference between a life saving medicine
• 0:11
  and a VAT of toxic sludge we are talking about mixing.
• 0:16
  It is deceptive, isn't it? When most people think of
• 0:18
  mixing, they think of, you know, a KitchenAid mixer or stirring
• 0:21
  milk into coffee. It feels intuitive.
• 0:23
  But in the Pharmaceutical industry, mixing is often the
• 0:27
  hidden killer. It is the silent variable that
• 0:29
  can ruin years of research and millions of dollars.
• 0:33
  Exactly. And we're not just talking about
• 0:34
  stirring things faster. We're diving deep into the
• 0:37
  engineering reality of how drugs are actually made.
• 0:40
  Based on the work of Paul Midler and Son.
• 0:43
  The stakes here are just incredibly high.
• 0:46
  It turns out the difference between a successful drug and a
• 0:48
  total failure can come down to the geometry of a turbine blade.
• 0:51
  That's right. And understand why it's so high
• 0:54
  stakes. We have to look at the the
• 0:56
  mission of these facilities. If you look at petrochemicals,
• 0:59
  oil and gas, they build these massive plants designed to do
• 1:02
  one thing continuously, 24/7. Right, they're optimized for a
• 1:06
  single steady state. Exactly, you turn it on and it
• 1:10
  runs for years. But pharma is different.
• 1:12
  They use what are called multi purpose batch reactors.
• 1:15
  Which sounds efficient from a business perspective. 1 factory,
• 1:19
  many products, but I assume that creates a massive engineering
• 1:23
  headache. You're asking one machine to be
• 1:24
  good at everything. Precisely.
• 1:26
  You have this one vessel, this one tank, and on Monday it needs
• 1:29
  to heat a reaction to boiling. On Tuesday it needs to distill a
• 1:33
  solvent, on Wednesday it needs to crystallize a solid product.
• 1:36
  That has to be a Jack of all trades.
• 1:38
  And making one machine do all of that without ruining the
• 1:42
  delicate chemistry inside, that is the core challenge.
• 1:46
  So let's start with the equipment itself.
• 1:48
  When I picture a chemical reactor, I just picture a big
• 1:51
  stainless steel tank. Standard stuff.
• 1:53
  But according to the source material, that's not always the
• 1:55
  case. In pharma, in fact, steel is
• 1:57
  often the enemy. No, often it can't be steel.
• 2:00
  In fine chemicals, you're frequently dealing with highly
• 2:04
  corrosive acids. Like hot hydrochloric acid?
• 2:06
  Exactly. That would eat right through 316
• 2:10
  stainless steel in a matter of days.
• 2:13
  You'd have leaks, contamination, structural failure.
• 2:16
  So the industry standard is the glass lined vessel.
• 2:19
  Which sounds incredibly fragile for an industrial environment.
• 2:23
  I mean, we're talking about thousands of liters of liquid
• 2:26
  swirling around. It's surprisingly robust
• 2:29
  chemically, but it introduces this massive mechanical paradox.
• 2:33
  These vessels can range from 80 liters up to 20,000 liters.
• 2:36
  20,000 liters. Imagine a 20,000 litre glass
• 2:39
  bottle encased in steel. The problem is, as an engineer,
• 2:44
  how do you fabricate the internals?
• 2:45
  You cannot Weld glass. Right.
• 2:47
  If you have a standard steel tank, you can Weld baffles to
• 2:50
  the wall, you can Weld complex sharp impeller shapes to the
• 2:53
  shaft. But with glass lined steel,
• 2:55
  you're stuck. You can't just spark up a tick
• 2:57
  torch. You are severely limited.
• 2:59
  The source material explains that because you have to coat
• 3:01
  the metal agitator in glass, you can't have sharp edges.
• 3:04
  Everything has to be rounded. Why is that?
• 3:06
  If you have a sharp 90° angle, the glass coating will crack due
• 3:10
  to surface tension during the firing process.
• 3:13
  So traditional pharmaceutical mixers, what we call retreat
• 3:17
  blade impellers, have these soft rounded geometries.
• 3:20
  Which changes the fluid dynamics completely.
• 3:23
  It's like trying to row a boat with a spoon instead of a
• 3:26
  paddle. You don't get that same bite.
• 3:28
  But there's a detail in this text about assembly that I found
• 3:31
  fascinating. It's how they actually get the
• 3:33
  blades into the tank. Ah yes, the cryoloc and alkaloc
• 3:37
  systems. This is a brilliant workaround
• 3:40
  for this small hole big part problem.
• 3:42
  Right, because the tank has a small man way at the top, but
• 3:45
  the impeller needs to be big to be effective.
• 3:47
  Too big to fit through the hole. In a metal tank you just bolt
• 3:50
  the blades on inside, but you can't do that with glass, can
• 3:54
  you? Exactly.
• 3:55
  Bolting creates stress points and stress cracks glass.
• 3:57
  So manufacturers develop this cryogenic shrink fitting
• 4:00
  technology. They make the drive shaft and
• 4:02
  the impeller is 2 separate glass line pieces.
• 4:05
  They lower the shaft into the tank, then they pump liquid
• 4:09
  nitrogen inside the hollow core of that shaft.
• 4:12
  So they're freezing the metal from the inside out.
• 4:14
  Yes, the extreme cold causes the metal inside the shaft to
• 4:18
  contract. It shrinks just enough, we're
• 4:21
  talking millimeters. Then while it's shrunken, they
• 4:24
  slide the impeller hub over the shaft.
• 4:27
  They stop the nitrogen flow, the shaft warms up, expands, and it
• 4:31
  creates this incredibly tight, seamless interference fit.
• 4:34
  That is just cool. You're using thermal contraction
• 4:37
  to assemble a glass machine inside a sealed tank.
• 4:41
  No bolts, no gaskets. It's brilliant engineering, but
• 4:45
  it highlights the constraints. And there's another massive
• 4:47
  trade off with glass. Glass is a ceramic.
• 4:51
  It's a terrible conductor of heat.
• 4:53
  Which is bad news if you have a reaction that produces a a lot
• 4:55
  of heat, right? You can.
• 4:56
  Get that nightmare for process control.
• 4:58
  The source knows the heat transfer coefficient is
• 5:00
  significantly lower than steel. Plus you have a thermal shock.
• 5:03
  Laurent the maximum temperature difference you can have between
• 5:06
  the coolant jacket on the outside and the liquid inside is
• 5:09
  only about 125°C. Wait, so if the reactor gets too
• 5:13
  hot you can't just blast the jacket with ice cold fluid to
• 5:16
  cool it down? If you do, you shatter the glass
• 5:18
  lining. It's like pouring boiling water
• 5:20
  into a cold glass cup. So you have to be gentle.
• 5:23
  Very gentle. This forces you to run highly
• 5:26
  exothermic reactions much slower than you otherwise would just to
• 5:30
  manage the heat. You are throttled by the glass.
• 5:33
  So you have rounded blades that don't mix aggressively and a
• 5:36
  tank that acts like an insulated thermos when you actually want
• 5:39
  to cool things down. That sets the stage for what I
• 5:42
  want to talk about next, the actual chemistry.
• 5:45
  Because if the mixing is compromised, the chemistry
• 5:48
  starts to do strange things. Let's get into the kinetics.
• 5:51
  This is where it gets really interesting for anyone
• 5:53
  interested in why scale up is so hard.
• 5:55
  The source material calls this mixing sensitive reactions.
• 5:59
  Now, to a layperson, mixing liquid sounds easy.
• 6:02
  I pour milk into my coffee. It mixes done.
• 6:04
  Why is this hard? Because in your coffee cup you
• 6:07
  don't have a chemical war going on in pharma.
• 6:10
  You have to distinguish between macro mixing, which is just
• 6:13
  moving the bulk fluid around, and micro mixing, which happens
• 6:15
  at the molecular scale. And the problem arises when the
• 6:18
  mixing speed is slower than the chemical reaction speed,
• 6:21
  correct? Exactly this is the concept of
• 6:23
  competing time scales. Let's look at the competitive
• 6:26
  consecutive reaction scenario described in the text.
• 6:29
  This is a classic trap. Imagine you have region A in the
• 6:32
  tank and you're adding region B. You want A+B to react and form
• 6:37
  product R That's your drug. OK, A+B makes R simple enough.
• 6:41
  But there's a catch. Product R is also reactive.
• 6:45
  If there's a lot of region B hanging around, product R will
• 6:48
  react with it to form substance S and substance S is waste.
• 6:52
  It's junk. So it's a race, you want A&B to
• 6:55
  find each other, but you need to stop B from then attacking the
• 6:57
  product you just made. Correct.
• 6:59
  Ideally you want B to be consumed by A instantly, but if
• 7:02
  you're mixing is slow, when you pour B into the tank, it doesn't
• 7:05
  disperse instantly. It creates a hotspot, a
• 7:08
  localized cloud near the feed pipe, where the concentration of
• 7:11
  B is extremely high. And inside that cloud, B eats up
• 7:14
  all the local A makes R, and then because there's still tons
• 7:18
  of B left, it immediately eats the R to make S.
• 7:21
  Precisely. You end up destroying your
• 7:23
  product the moment you make it simply because you didn't spread
• 7:25
  the Regent be out fast enough. And the source points out this
• 7:28
  is the classic scale up trap in the lab.
• 7:31
  In a one liter beaker you can mix things almost instantly.
• 7:34
  Right, but then you move to a 10,000 liter tank.
• 7:37
  The circulation time, the time it takes for fluid to loop
• 7:41
  around the tank, might be 20 or 30 seconds.
• 7:44
  In chemical terms, that's an eternity.
• 7:47
  The damage is already done. And we aren't talking about
• 7:49
  losing just a little bit of product.
• 7:51
  We are talking about the 0.1% rule.
• 7:54
  This is the regulatory hammer in the pharmaceutical world.
• 7:57
  If an impurity level rises above 0.1%, you often have to register
• 8:02
  that impurity as a new chemical entity with regulators.
• 8:05
  Which means toxicity studies paperwork, massive delays.
• 8:09
  It's not just paperwork, though. The source mentions these
• 8:12
  impurities can physically wreck the process downstream.
• 8:15
  A tiny amount of byproduct can change the surface tension,
• 8:19
  causing it to foam uncontrollably, or it can turn
• 8:22
  into a gel that makes filtration impossible.
• 8:25
  You might have a tank full of product, but if you can't filter
• 8:27
  it, you have to dump the whole batch.
• 8:29
  So 0.1% is the difference between shaking the drug and
• 8:33
  shutting down the plant. It absolutely is.
• 8:35
  Let's ground this in a real world story.
• 8:37
  The text breaks down the gas wars involved in synthesizing an
• 8:41
  antibiotic called sufoxetin. This story really shows what
• 8:45
  happens when mechanical mixing just fails.
• 8:47
  This is a fascinating case study because it shows a war on three
• 8:50
  fronts, chemistry, mechanics, and regulation.
• 8:54
  So in making sufhoxitin, there's a step where they react to
• 8:57
  intermediates, but this reaction produces hydrochloric acid, HCL
• 9:01
  as a byproduct. And as we mentioned, HCL is
• 9:04
  nasty, but here it wasn't just corrosion.
• 9:06
  Right. No.
• 9:06
  Here it was acting as a catalyst for decomposition.
• 9:09
  In this specific reaction. If the concentration of HCL hits
• 9:13
  just 0.1 molar, it destroys the product.
• 9:17
  OK, so the obvious solution is to remove the HCL.
• 9:19
  You can't remove it all. If you remove it completely, the
• 9:23
  reaction stops. Needs a tiny amount of acid to
• 9:26
  proceed. The source says the Goldilocks
• 9:28
  zone was exactly 0.004 molar. That's an incredibly tight
• 9:36
  window. It is so.
• 9:37
  The engineers first tried mechanical solutions.
• 9:40
  They tried distillation, boiling off the HCL gas as it formed.
• 9:43
  Too slow. Way too slow.
• 9:46
  Mass transfer from liquid to gas is limited by surface area.
• 9:50
  By the time they boiled it off, the concentration in the liquid
• 9:52
  had already spiked and killed the product.
• 9:54
  Oh, what did they do? They turn to molecular sieves.
• 9:58
  These are solid materials with tiny pores that trap specific
• 10:01
  molecules. They act like a sponge for acid.
• 10:04
  But the source makes a point of the physical form of these
• 10:07
  sieves. They couldn't just use standard
• 10:08
  pellets. No pellets were too slow.
• 10:10
  The diffusion of the acid into the center of a pellet took too
• 10:13
  long. It's the same problem again,
• 10:15
  mass transfer limitations. So they had to use powdered
• 10:18
  sieves 1 to 4 micrometers in size to get more surface area.
• 10:21
  But even that wasn't enough, was it?
• 10:23
  No, it wasn't. They hit the absolute limit of
• 10:26
  what physical mass transfer could do, mechanical mixing and
• 10:29
  physical absorption, which is too slow.
• 10:32
  So they had to introduce a homogeneous scavenger.
• 10:35
  A homogeneous scavenger? What's that?
• 10:37
  They added a chemical trimethyl silly metal carbamate that
• 10:42
  reacts instantly with HCL in the solution.
• 10:45
  So they essentially hired a chemical mercenary to neutralize
• 10:49
  the acid the microsecond it was formed.
• 10:51
  Exactly. It bypassed the mixing problem
• 10:53
  entirely because the scavenger was dissolved in the liquid.
• 10:56
  It was right there, no mixing required.
• 10:58
  And the take away there is crucial.
• 11:00
  Sometimes you cannot mix your way out of a problem.
• 11:03
  You have to change the chemistry to survive the equipment.
• 11:06
  That is the lesson. Don't bang your head against the
• 11:08
  wall trying to design A better impeller if the physics of
• 11:11
  diffusion won't Did You win change the game.
• 11:13
  Speaking of limitations, let's move to another section.
• 11:16
  The expert calls this the most difficult operation to scale up.
• 11:20
  We're moving from liquids to solids crystallization.
• 11:23
  Crystallization is a nightmare because the requirements are
• 11:26
  completely contradictory to mix chemicals as we just discussed,
• 11:30
  you usually want high energy, you want turbulence.
• 11:34
  Right, smash it all together. High turbulence, good mixing.
• 11:37
  But to grow a crystal, you need gentleness.
• 11:40
  A crystal. It is an ordered lattice.
• 11:43
  It takes time. If you hit a growing crystal
• 11:45
  with a high speed blade, you shatter it.
• 11:48
  You break the lattice. It's like trying to build a
• 11:50
  House of Cards inside a tornado. That is a perfect analogy and
• 11:54
  the source gives a great example involving a side chain for an
• 11:56
  antibiotic. It worked fine in the lab, Nice
• 11:59
  big crystals and they moved to the pilot plant and disaster
• 12:01
  struck. They got what the text calls a
• 12:04
  bimodal distribution. Which is engineer speak for.
• 12:08
  We have some giant chunks and a massive amount of dust.
• 12:11
  They call the dust fines. And fines are the enemy of
• 12:13
  manufacturing. Why?
• 12:15
  Fines clog the filters. Instead of liquid flowing
• 12:18
  through a pile of rocks, it's trying to flow through mud.
• 12:20
  It can take days to filter a single batch.
• 12:23
  Plus, fines trap impurities. The process just ground to a
• 12:27
  halt. So they investigated what went
• 12:29
  wrong. It turned out to be a double
• 12:31
  failure. First, the reaction was
• 12:33
  happening too fast. It was causing super saturation
• 12:36
  to spike too quickly. High super saturation triggers
• 12:40
  nucleation, which means instead of growing existing crystals,
• 12:44
  the liquid just spawns millions of tiny new ones.
• 12:47
  So that creates the dust, yes. And on top of that, the impeller
• 12:52
  they were using was spinning too fast, smashing the crystals that
• 12:55
  did manage to grow. So they had to slow everything
• 12:58
  down. Drastically.
• 12:59
  They extended the time they spent adding reagents from 2
• 13:02
  hours to six hours. This kept the Super saturation
• 13:05
  low, allowing the crystals to grow slowly.
• 13:07
  But they also changed the hardware.
• 13:09
  This is where we talk about the Ekato Intermeg.
• 13:11
  Yes, the intermeg is a specialized impeller.
• 13:14
  Unlike a standard turbine that pushes everything out with high
• 13:16
  force, the intermeg consists of multiple stages that create a
• 13:20
  chaotic up and down flow. It provides excellent
• 13:23
  circulation, keeping solids floating, but with very low
• 13:26
  shear. It doesn't act like a blender.
• 13:28
  It gently lifts rather than chops.
• 13:30
  Exactly, and this leads to a critical lesson about scaling
• 13:34
  up. There's a standard rule called
• 13:36
  equal power per unit volume. It says you have a big tank.
• 13:39
  You should put in the same watts per liter as you did in the
• 13:42
  small tank. That sounds logical.
• 13:44
  Preserve the energy density. For mixing liquids it usually
• 13:47
  works. For crystallization it is
• 13:49
  dangerous because as you scale up the tank, if you keep power
• 13:53
  per volume constant, the tip speed of your mixer blade
• 13:57
  increases. The tip speed being how fast the
• 14:00
  outer edge of the blade is moving.
• 14:01
  Right. And tip speed is what breaks
• 14:02
  crystals is the impact velocity. So in this case, if they had
• 14:06
  followed the textbook rule, they would have kept destroying their
• 14:09
  product. They actually had to use less
• 14:11
  power per volume to save it. This brings us to the final
• 14:14
  dilemma, the Jack of all trades problem.
• 14:16
  We mentioned at the start that these plants use multi purpose
• 14:19
  reactors. This is the economic reality.
• 14:22
  You have one vessel on Monday, you might be doing a
• 14:25
  hydrogenation. Which involves gas.
• 14:27
  Right, Hydrogen gas. To get gas to react with liquid,
• 14:32
  you need to chop the bubbles up into tiny spheres.
• 14:34
  You need high shear. You need to spin that impeller
• 14:37
  fast, chopping violently. But then Tuesday comes around
• 14:40
  and you're doing that crystallization we just talked
• 14:42
  about. Where high shear destroys the
• 14:44
  product so you have the same motor and the same shaft needing
• 14:48
  to do 2 physically opposite things.
• 14:50
  This is why the source emphasizes variable speed drives
• 14:53
  or VS DS. It sounds simple, just a knob to
• 14:57
  turn the speed down, but historically industrial motors
• 15:00
  were often fixed speed. Adding the VSD is capital
• 15:03
  intensive, but without it you're dead in the water.
• 15:06
  You can't just pick 1 speed. And there's a legal trap here,
• 15:09
  too, called the Biobatch. This is crucial when you're
• 15:12
  getting a drug approved by the FDA.
• 15:14
  You run a specific production batch called the BIO batch.
• 15:17
  This is the batch used for the pivotal clinical trials.
• 15:19
  The one that proves the drug actually works in humans, right?
• 15:22
  Once that bio batch is approved, the FDA essentially says OK, you
• 15:26
  can sell this drug but you have to make it exactly the same way.
• 15:29
  You are legally locked into those physical attributes.
• 15:32
  So if your bio batch had a certain mixing profile, you have
• 15:35
  to replicate that at full scale forever.
• 15:38
  Yes, if you optimize your bio batch using a specific mixing
• 15:41
  speed that gave you a certain crystal size, you must be able
• 15:45
  to hit that target in the 20,000 liter tank.
• 15:48
  Yeah, if your big tank can't match the biobatch parameters,
• 15:51
  you don't have a product, you have a regulatory failure.
• 15:54
  That's terrifying. You are just designing for
• 15:56
  physics. You are designing for
• 15:58
  compliance. The constraints of the vessel
• 16:00
  define the legality of the drug. It connects back to the very
• 16:03
  beginning. The mixer is not just a stirrer.
• 16:05
  It is the device that determines purity, yield and legal
• 16:09
  compliance. It is the heart of the process.
• 16:11
  So let's unpack this. All we've learned that mixing is
• 16:14
  this incredible balancing act. You're fighting corrosion with
• 16:17
  glass, which limits your shape and heat transfer.
• 16:19
  You're fighting chemistry that wants to eat your products, so
• 16:22
  you need to mix fast. But you're fighting crystal
• 16:24
  breakage, so you need to mix slow.
• 16:26
  And you are doing it all in a giant tank Where you can't even
• 16:29
  see what is happening. You rely entirely on
• 16:31
  calculations and sampling. It gives you a new appreciation
• 16:34
  for that little pill in the medicine cabinet.
• 16:37
  It certainly does. Yeah.
• 16:38
  And it raises A provocative question.
• 16:40
  When we look at the history of drug development, we see many
• 16:44
  drugs that failed in phase three trials or were cancelled because
• 16:48
  they couldn't be manufactured reliably.
• 16:50
  The unscalable drugs. We have to wonder how many life
• 16:53
  saving drugs failed, not because the chemistry was wrong, not
• 16:57
  because the molecule wouldn't have cured the disease, but
• 17:00
  simply because we couldn't build a mixer that worked at 20,000
• 17:03
  liters. How many cures are lost in that
• 17:06
  scale up gap? That is a heavy thought to end
• 17:10
  on. It really shows that mechanical
• 17:12
  engineering and chemical engineering are two sides of the
• 17:14
  same coin. You can't have the cure without
• 17:16
  the tank. Absolutely.
• 17:18
  Well, on that note, we are going to wrap up this deep dive.
• 17:20
  Thank you for listening. And if you haven't already, get
• 17:22
  over to our YouTube channel Mechanical Engineering Made
• 17:25
  Simple, where we post hands on how to videos, visual
• 17:28
  breakdowns, and practical engineering walkthroughs.
• 17:30
  You can actually use a family that solves together, stays
• 17:33
  together.