Contents
A stirred reactor is a pot with a lid until the reaction starts giving off heat. From then on it is a heat exchanger with a very awkward shape, and the jacket around it decides how fast the plant can run. The scale-up series showed that the 10,000 L reactor R-101 has so little wall for its volume that a one-hour laboratory reaction has to be stretched to about three hours. This part of the sizing series looks at what can be done about that: which kinds of jacket exist, how to calculate what each can remove, and when a jacket is not enough.
Step 1: state the duty
A reactor jacket usually has three jobs in every batch: heat the charge to reaction temperature, remove the heat of reaction, and cool the product down afterwards. The energy balance in From Pilot Data to a Mass and Energy Balance sets them out for R-101. The one that fixes the jacket is the heat of reaction, because it cannot be made to wait.
R-101 holds a 10,000 kg batch that releases 150 kJ/kg, or 1.5 GJ in all. One reactant is added over 3 hours and reacts almost as it arrives, so the heat appears at an average rate of:
\[Q_{\text{gen}} = \dfrac{10{,}000\ \text{kg} \times 150\ \text{kJ/kg}}{3\ \text{h} \times 3{,}600\ \text{s/h}} = 139\ \text{kW}\]
The batch is held at 60 °C. Cooling water enters the jacket at 25 °C. The jacket must remove at least 139 kW across that temperature difference, with something in hand.
Step 2: know the jacket types
There are three common ways to wrap a heating or cooling fluid around a vessel, and two ways to add area when the wall is not enough.
| Type | How it is built | Strengths | Limits |
|---|---|---|---|
| Plain (conventional) jacket | A second shell around the vessel, leaving a gap of 50 to 100 mm. A spiral baffle or inlet nozzles are often added to make the fluid swirl | Simple and cheap. Ideal for steam, which needs no velocity. The usual choice on glass-lined vessels | Liquids move slowly, so heat transfer on the jacket side is poor. The jacket pressure pushes inward on the vessel wall, which limits the pressure and thickens the shell |
| Dimple jacket | A thin outer sheet pressed into a pattern of dimples and spot-welded to the vessel wall | Light and economical on large vessels. The dimples stir the fluid and stiffen the wall | Higher pressure drop. Narrow passages foul and are hard to clean. Moderate pressure only |
| Half-pipe coil jacket | Pipe split lengthwise and welded in a spiral around the shell | High velocity and good heat transfer with liquids. Holds high pressure, so it suits hot oil and high-pressure steam | A great deal of welding, so it costs more. Not practical on glass-lined vessels |
| Internal coil | A coil of pipe inside the vessel, in the liquid | Adds area without a bigger vessel, and has no thick vessel wall in the way | Hard to clean, takes up volume and gets in the way of mixing. A leak goes into the batch |
| External loop | The batch is pumped through a heat exchanger outside the vessel and back | The area is no longer limited by the size of the vessel | Needs a pump and more hold-up outside the reactor. Unsuitable for slurries and shear-sensitive products |
Step 3: measure the area
The jacket only works where there is liquid on the other side of the wall. With the liquid depth equal to the diameter, as in the scale-up series, R-101 is 2.34 m across and the wetted wall is:
\[A = \pi\, D\, H = \pi \times 2.34 \times 2.34 = 17.1\ \text{m}^2\]
That is 1.7 m² for each cubic metre of batch. The 100 L pilot reactor had about 8 m² per cubic metre. A jacket on the bottom head would add some area. A jacket above the liquid level adds none, which matters early in a fed batch when the vessel is only part full.
Step 4: build the overall coefficient
Heat leaving the batch has to cross four resistances in a row: the film of liquid on the inside of the wall, the wall itself, any fouling, and the film of cooling water on the jacket side. The overall coefficient \(U\) is the reciprocal of their sum:
\[\dfrac{1}{U} = \dfrac{1}{h_{\text{batch}}} + \dfrac{t_{\text{wall}}}{k_{\text{wall}}} + R_{\text{fouling}} + \dfrac{1}{h_{\text{jacket}}}\]
For R-101 the batch side coefficient is taken as 800 W/m²·K, typical of a thin organic liquid with a well-sized agitator. The wall is 12 mm of stainless steel with a conductivity of 16 W/m·K. Fouling on both sides together is 0.00033 m²·K/W. Only the last term depends on the jacket. With cooling water, typical jacket side coefficients are about 1,000 W/m²·K for a plain jacket with a spiral baffle, 2,500 for a dimple jacket and 5,000 for a half-pipe coil, because the water moves at roughly 0.3, 0.9 and 1.6 m/s. For the plain jacket:
\[\dfrac{1}{U} = \dfrac{1}{800} + \dfrac{0.012}{16} + 0.00033 + \dfrac{1}{1{,}000} = 0.00333 \quad\Rightarrow\quad U = 300\ \text{W/m}^2\!\cdot\!\text{K}\]
Figure 1 is the lesson of this step. Going from a plain jacket to a half-pipe coil makes the jacket side five times better, yet \(U\) rises only from 300 to 395 W/m²·K, about 32%. Once the jacket side is good, the batch film and the wall are in charge, and money spent on the jacket buys little. On a glass-lined reactor the glass and the thick steel behind it dominate even more, and the jacket type hardly matters.
Step 5: calculate the cooling capacity
The heat a jacket removes is the overall coefficient times the area times the mean temperature difference:
\[Q = U\, A\, \Delta T_{\text{mean}}\]
The water warms as it goes round, so the mean is a log mean, as for the heater in Part 2. With the batch at 60 °C and water entering at 25 °C and leaving at 34.5 °C, the differences at the two ends are 35 K and 25.5 K and the log mean is 30 K. For the plain jacket:
\[Q = 300 \times 17.1 \times 30 = 154{,}200\ \text{W} \approx 154\ \text{kW}\]
That is the 154 kW quoted in the mass and energy balance, against the 139 kW the reaction releases, so about 10% of the capacity is in hand. Carrying that heat away with a 9.5 K rise takes about 14 m³/h of cooling water. The same calculation for each jacket, with the same 14 m³/h of water, gives the table below. The coefficients are in W/m²·K.
| Jacket on R-101 | Jacket side | Overall U | Capacity, kW | Addition, h |
|---|---|---|---|---|
| Plain jacket | 1,000 | 300 | 154 | 2.7 |
| Dimple jacket | 2,500 | 366 | 182 | 2.3 |
| Half-pipe coil jacket | 5,000 | 395 | 194 | 2.2 |
| Plain jacket plus internal coil | 249 | 1.7 | ||
| Half-pipe jacket plus internal coil | 279 | 1.5 |
The last column is the 1.5 GJ of reaction heat divided by the capacity: the shortest time over which the reactant can be added without the batch warming up. The better jackets shorten the addition from 2.7 to about 2.2 hours. The gain is a little less than the rise in \(U\), because water that picks up more heat leaves warmer, at 36.9 °C from the half-pipe jacket, and that trims the mean temperature difference. It is worth having, but it is not a different plant.
Step 6: add area when the wall is not enough
The larger gains come from more area. An internal coil of 2-inch pipe, 9 turns at 1.6 m diameter, adds 8.6 m². It also works better per square metre than the jacket, because its wall is only 3.9 mm thick and water runs fast inside it: about 493 W/m²·K. With the same 14 m³/h of water shared between jacket and coil, the coil adds 95 kW, which takes the plain-jacketed reactor to 249 kW and the addition down to 1.7 hours. More water would raise it further.
The price is paid elsewhere. A coil is hard to clean between products, it disturbs the flow pattern the agitator was chosen for, and a pinhole puts cooling water into the batch. Where those matter, the alternative is an external loop: the batch is pumped through a heat exchanger whose area can be whatever the duty needs. Either way, the choice between jacket, coil and loop is made at the design basis stage, because it changes the vessel, the agitator and the layout.
Step 7: check the worst case
The capacity has to be compared with what happens when things go wrong, not only with the design rate. If the whole charge were added in one hour, as in the laboratory procedure, the heat would arrive at 417 kW. The plain jacket removes 154 kW. The difference starts warming the batch at about 47 K an hour, and it heads for the boiling point of 110 °C. No jacket in the table would hold that rate. The protection is the addition rate itself, a high-temperature trip on the feed, and the analysis in What Happens If the Cooling Fails.
Two other cases belong in the check: cooling water at its summer temperature, which shrinks the temperature difference, and the start of the addition, when the level is low and only part of the jacket is doing anything.
Figure 2 shows R-101 cut in half, with the batch circulating inside, coloured by temperature, and cooling water running round the jacket. Switch the jacket type and watch three things: the speed of the water in the jacket, the orange arrows that show how much heat the wall can pass, and the chart of temperature through the wall. Add an internal coil, and change the addition time, the cooling water and the agitator speed. Watch whether the jacket can hold the batch at 60 °C or the temperature climbs away.
This browser cannot show the 3D model. The controls below still run the calculation.
Design point: the reaction releases 139 kW and the plain jacket can remove 154 kW at 60 °C, so the temperature controller holds the batch with about 10% in hand.
The data sheet
| Item | R-101 jacket, base case |
|---|---|
| Vessel | 10,000 L batch, 2.34 m diameter, stainless steel, 12 mm wall |
| Jacket type | Plain jacket with spiral baffle, on the cylindrical wall |
| Heat transfer area (wetted) | 17.1 m² |
| Batch temperature | 60 °C |
| Cooling water in / out | 25 / 34.5 °C, 14 m³/h |
| Overall coefficient | 300 W/m²·K |
| Heat of reaction to remove | 139 kW average over a 3-hour addition |
| Cooling capacity | 154 kW |
Common mistakes
- Carrying the pilot coefficient to full scale. The wall is thicker, the agitator turns more slowly and the jacket is a different design. Build \(U\) from its parts.
- Counting jacket that is not wetted. Area above the liquid level does almost nothing.
- Upgrading the jacket when the batch side controls. Check which resistance is largest before paying for half-pipe.
- Sizing on the average. Heat release is rarely even through the addition. Calorimetry gives the peak.
- Forgetting the jacket's own pressure. A plain jacket puts external pressure on the vessel wall, and that can decide the shell thickness.
Key points
- The duty that sizes a reactor jacket is the heat of reaction at its highest rate.
- Plain, dimple and half-pipe jackets differ mainly in the jacket side coefficient, cost and pressure rating.
- The overall coefficient is set by the largest resistance. On R-101 that is the batch side, so a better jacket raises \(U\) by only about 32%.
- \(Q = U A\, \Delta T_{\text{mean}}\), with the wetted area only.
- When the wall is not enough, add area with an internal coil or an external loop, and control the heat release with the addition rate.
Earlier in this series: Part 1 sized the pump, Part 2 the heater, Part 3 the feed tank, Part 4 the control valve and Part 5 the lines. Next: Part 7 sizes the agitator that keeps the liquid moving past this jacket.
The plant, fluid properties and equipment in this series are an illustrative example. The calculations are preliminary sizing estimates. Final sizing needs vendor data, the applicable codes and a qualified engineer.
