Why is Process Safety a Critical Consideration and How Can it be Implemented to Ensure Safe Scale-Up?
We will discuss how thermal and pressure hazards can be identified, assessed, and mitigated at each stage of your product development, and will see how designing safety into your development process can deliver confidence in ensuring a safe and efficient scale-up to manufacture.
Why Process Safety?
From the synthesis of a new API to the development of a novel polymer, chemical reactions, and the products they form, are front and center in the world in which we live. One of the biggest challenges facing those trying to bring these products to market is scale-up: taking small, lab-based synthesis through pilot plant, to full-scale manufacture.
It is well recognized that there are numerous safety risks involved in scaling up chemical processes. Central to these are the changes in heat-loss behavior which differ with scale:
- Many reactions are exothermic. In a lab-based environment, due to the small scale, this heat can be safely dissipated to the surrounding environment, but on scale-up the reaction may require additional active temperature management to ensure safe reaction conditions;
- Components in the reaction may become unstable under certain operating conditions, including the simple act of storing large volumes, which can cause additional hazards that require management.
The risks associated with scale-up, have potentially hazardous, even fatal, consequences if they are not addressed. Thus, it is essential that thermal and pressure hazards are considered, identified, assessed, and mitigated at each stage of product development (Figure 1). By designing safety into the product development from the earliest stages – a Safety by Design, if you like – one can feel confident that the hazards associated with scale-up to manufacturing can be successfully addressed.
Figure 1: Safety considerations at each stage of product development
Process Safety in the Development Workflow
Broadly speaking, process safety scale-up follows 3 main stages:
- Screening of potential hazards
- Evaluation of the main reactions, including possible unintended side reactions
- Evaluation of “what if” scenarios and subsequently implementing mitigation and control strategies
These steps naturally map on to the product development funnel, as illustrated in the schematic of Figure 2.
It follows that at the early stages of development there will be many candidate molecules and reaction processes being tested across a range of metrics before being taken any further along the development funnel. From a safety perspective, the raw materials that go into synthesizing these molecules will need to be screened for potential hazards, along with testing specific synthetic routes for these molecules.
Figure 2: Product development funnel
Once a target molecule has been identified, and its synthetic route determined, the process by which it is synthesized must be developed and optimized, with a view to manufacture at scale. In terms of safety, the desired, main reaction needs to be assessed and understood, alongside identifying potential additional thermal and pressure hazards as a result of the operating conditions. An example of these hazards could be undesired side products that are more unstable than the desired product.
There is then an opportunity to evaluate different operating conditions to assess the impact they have on these hazards and optimize the conditions accordingly. Finally, with the operating conditions optimized and the potential hazards identified, it is time to assess the magnitude of these hazards prior to full scale-up.
The transition from one stage to the next requires specific analysis, which will now be considered in detail.
Discovery into Process Development
Quick hazard screening of raw materials at the front end of development allows earlier decisions to be made on how to develop a desired reaction and the synthetic route.
It is important to know early on in a product development if the raw materials are likely to have thermal stability issues, including production of possible non-condensable gases which could pose an explosion risk on scale up.
Key questions to answer are:
- What is the thermal decomposition profile of the raw material? In particular, does the material have a ‘low’ onset temperature? If the desired synthesis reaction reaches this onset temperature via an initial exothermic event, this could cause the reactant to further increase in temperature, causing a secondary thermal runway and forming a vicious cycle of ever-increasing energy release.
- Is there a rapid rate of pressure change during the exothermic event? This would indicate that the reactant could pose an explosion hazard during a thermal runaway. The most hazardous reactions to scale-up are the ones that involve the production of non-condensable gas. This is because it is not the increase in temperature that poses the greatest risk, but rather the explosive nature of a rapid increase in pressure.
Micro-scale calorimetry is typically used to carry out a quick screening assessment on the thermal properties of the raw material, with Differential Scanning Calorimetry (DSC) being the typical method. However, this method does not provide critical information on the rate of pressure change, which is vital for assessing the possible explosion hazard. It also suffers from the challenge that representative samples can’t always be obtained. For raw, pure materials this is less of an issue, particularly when one only needs an approximate assessment on whether to proceed further with the material, but it becomes more significant as one moves to screening reaction mixes and further into process development, and rapid screening of larger samples sizes is required.
A device such as the H.E.L Thermal Screening Unit (TSu) enables quick screening of both temperature and pressure on the same platform, allowing a quick assessment to be made on how hazardous a material is. As Figure 3 depicts, the sample temperature is ramped up until an exothermic event is triggered. Both the sample temperature and sample pressure are monitored throughout, yielding information about the onset temperature of the exothermic event, and the resultant rate of pressure change.
Figure 3: Example data from the TSu
If there is a need to characterize rapid thermal decompositions, higher resolution data on the rate of pressure change may be required. The Phi-TEC I is an adiabatic screening tool, but it can be run in a mode similar to that of a DSC and the TSu, whereby the sample temperature is ramped up to induce an exothermic event. In contrast to the basic thermal screening methods, the Phi-TEC I offers a high data-rate acquisition option, which provides the necessary precision for classifying extremely energetic exothermic events.
The key objective in Process Development is to define the conditions in which the reaction should be operated to minimize possible hazards and risk. It requires an understanding of the potential sources of thermal runaway risk, building on the preliminary screening studies carried out in Discovery, and an understanding of the chemistry of the process itself. This enables the process safety chemist to classify the criticality of the reaction, and what are safe operating conditions needed to mitigate the hazards identified.
It can be split into four main areas of investigation, the first three of which are illustrated in the reaction profile in Figure 4:
1. Desired Reaction
2. Thermal Runaway of Reaction
3. Decomposition/Second Thermal Runaway?
4. How Can the Operating Conditions be Modified to Reduce the Risks Identified?
Figure 4: Key parameters to determine for assessing the criticality of a reaction
1. Desired Reaction
Here, the key pieces of information to extract are:
- What is the energy evolved in the reaction?
- What is the subsequent rise in temperature when no additional cooling is applied to keep reaction isothermal?
- What is the cooling capacity required in the manufacturing plant to deal with this and keep the reaction isothermal?
In the following example of a semi-batch esterification reaction process, acetic anhydride was added at a 2ml/minute dosing rate over 50 minutes to methanol in a 1L reactor in the Simular, a reaction calorimeter, while the temperature was held isothermally at 30 ⁰C. Figure 5 shows the:
- Reaction power (this is essentially the rate of heat generation) [pink trace]
- Energy released [blue trace]
- Feed rate [yellow trace]
Figure 5: Esterification reaction at 30 ⁰C, 2ml/min feed rate
Under these conditions, it can be seen that the reaction power reaches a maximum of 15W. This means a cooling capacity of at least 15W is required to keep this reaction isothermal. As the reaction was performed in a 1L reactor, this can be directly scaled up for a 2000 L pilot reactor, indicating that a 30kW (30,000W) cooling capacity is required for the larger system.
Thus, reaction calorimetry enables the heat evolved during the desired reaction to be measured, which is then used to calculate the reaction power and subsequently the cooling capacity required to keep the system at the desired reaction temperature, Tp.
2. Thermal Runway of Reaction
However, what happens if there is a plant failure? What is the maximum possible temperature the desired reaction will run to if, for example, the cooling used to maintain an isothermal reaction fails? The key questions to answer in this area of investigation are:
- what is the Maximum Temperature of Synthesis Reaction (MTSR)?
- what is the Time to Maximum Rate (TMRr)?
Determining these parameters, which are illustrated schematically in Figure 6, enables one to assess whether there will be enough time and sufficient emergency cooling capacity to deal with the increase in temperature.
Figure 6: Schematic illustrating the key parameters of the thermal runaway of reaction
Critical to determining these parameters is an understanding of the kinetics of the desired reaction. In a kinetics-controlled reaction there is a risk of reagent accumulation if the reaction is too slow. Effectively, this means that there is potential energy stored within the reactor, which could be released adiabatically in the event of a cooling failure. This scenario is illustrated schematically in Figure 7. This could lead to a thermal runaway if not dealt with, and perhaps trigger undesired reactions or decompositions and further thermal runaways.
Figure 7: The risk of accumulated reagent
Reaction calorimetry can also be used to determine the kinetics of the reaction, which enables the level of reagent in the reactor at any given time to be predicted. Returning to the previous example in Figure 5, it can be seen that although the reagent feed stops at 110 minutes, the energy output from the reaction continues for a further 240 minutes. The rate of heat generation does not drop to zero – signaling the end of the reaction – until the 350-minute mark. This indicates that there is unconsumed reagent in the reactor at the point the feed rate stops, which then continues to react until equilibrium is reached. Turning to the energy released, we see that only 50% (34.6 kJ) of the total energy released over the course of the reaction was released at the point the feed rate stops. We can extrapolate from this that only 50% of the reagent has been consumed at the point the feed rate stops.
The true extent of the issue posed by accumulated reagent depends at what point the plant failure occurs. Usually, the worst-case scenario is if the failure occurs just before or after the feed is stopped, as this represents the point at which the concentration of accumulated reagent is at its highest. In this particular case, this would result in the release of 34.6 kJ of energy (assuming no other reaction occurred) from the remaining unreacted reagent. From this, and a knowledge of the thermal mass of the reactants and reactor, we can calculate the temperature rise for this worst-case scenario, via the equation depicted below:
where is the energy released by the reaction of the accumulated reagent, , and is the thermal mass of the reactants and reactor. In this case, the temperature rise is determined to be 106 ⁰C.
As the temperature of the reaction (Tp) was being held isothermally at 30 ⁰C before the cooling failure, we can determine that the maximum temperature attained, or the Maximum Temperature of Synthesis Reaction (MTSR), is 136 ⁰C under these conditions (30 oC isothermal hold, plus 106 oC of temperature increase). This is summarised by the equation below:
where represents the adiabatic temperature rise for reagent accumulation, . This calculation represents the worst-case scenario for the MTSR, as it assumes that there are no heat losses from the system (i.e. it is under adiabatic conditions).
The speed of the runaway (Time to Maximum Rate – TMR) can only be estimated from reaction calorimetry, based on the heat release rate and the level of accumulated reagent. The pressure released from this thermal runaway can, similarly, only be estimated. At this stage of development, these estimates are sufficient. However, when moving to scale-up, and in particular in those cases where the calculations suggest the runaway is particularly exothermic, it is important that the magnitude of the worst-case scenario is accurately quantified.
3. Decomposition/Secondary Thermal Runaway
The next information required is whether a decomposition or secondary reaction might be triggered by the MTSR, which could lead to a secondary thermal runaway. A secondary thermal runaway is likely to occur if MTSR is greater than the onset temperature of a component in reaction mix (Td). This scenario is illustrated in Figure 8.
As previously described, MTSR is derived from reaction calorimetry, while the onset temperatures of materials are determined using thermal screening tools, such as the Thermal Screening Unit (TSu).
A more accurate value can be determined using adiabatic calorimetry, but at this stage it is often considered sufficient to use a simpler thermal screening tool. The TSu enables representative samples of the reaction mix to be screened and reliable onset temperatures determined for the reaction intermediates and products. Temporal thermal stability studies also need to take place on the reaction waste stream, which the TSu also facilitates with isothermal stability screening. Just as it is important to identify hazards during the reaction itself, it is also important to understand if there are any longer-term stability issues in the waste products formed, so that these issues can also be mitigated.
Figure 8: A decomposition or secondary reaction is triggered if MTSR is greater than Td
If it becomes apparent that there is a risk of a secondary thermal runaway occurring, it is necessary to then determine what the adiabatic temperature rise of the decomposition/secondary reaction is, and the TMRd. This allows an assessment to be made as to whether there will be sufficient time and enough emergency cooling capacity to deal with the increase in temperature.
As this reaction involves a secondary thermal runaway, it is inherently a hazardous process to scale-up. Therefore, although these parameters can be estimated from micro-calorimetry, it is advisable to use adiabatic calorimetry to get an accurate assessment.
The data collected enables the criticality of the reaction to be classified, as depicted in Figure 9
Figure 9: Classification of the criticality of a reaction
The schematic references the key reaction parameters introduced in Figure 8:
- the isothermal temperature of the reaction (Tp)
- the decomposition temperature of a component in the reaction mix (Td)
- the Maximum Temperature of the Synthesis Reaction (MTSR)
It also introduces an additional parameter, the boiling point of the solvent or active chemicals used (BP), as this is another factor that needs to be taken into account when considering the overall safety of a reaction. The relative order of these parameters in terms of temperature determine the criticality of the reaction.
- Scenario (1): the thermal order of the parameters is Tp<MTSR<BP<Td. This represents the least hazardous reaction scenario and no special measures are required as the maximum temperature of the synthesis reaction (MTSR) does not exceed that of the boiling point (BP) of the solvent, nor the decomposition temperature (Td).
- Scenario (2): as with Scenario (1), the MTSR does not exceed Td, although in this case, the boiling temperature of the solvent is higher than Td (Tp<MTSR< Td< BP). This means that, unlike Scenario (1), the boiling point will not act as a limit on the maximum temperature the reaction will attain, and heat accumulation conditions should be avoided.
- Scenario (3): the MTSR is higher than the boiling point, but less than Td (Tp< BP< MTSR<Td). This means that risk reducing measures are required, but that the boiling point of the solvent could be used as a safety mechanism for the reaction.
- Scenario (4): the MTSR is greater than Td, which means that there is a risk that the decomposition could be triggered However the boiling point of the solvent could be used as a limiting mechanism to prevent the decomposition reaction being triggered, as BP is lower than Td (Tp< BP<Td < MTSR). In this case, risk reducing measures are required.
- Scenario (5): as with Scenario (4), the MTSR is greater than Td, which means that there is a risk that the decomposition could be triggered. However, the boiling point of the solvent is also above that of Td, so it can’t be used as a safety barrier (Tp <Td < MTSR< BP). In this case, emergency control measures are also required, which may include redesigning the process
4. How Can The Operating Conditions Be Modified To Reduce The Risks Identified?
In the case of Scenarios (4) & (5) previously discussed, it is necessary to consider if the MTSR and TMRr can be reduced by altering the conditions of the reaction to reduce the levels of accumulated reagent.
There are a number of ways this could be addressed, such as:
- Increasing the temperature of the reaction
- Increasing the pressure of the reaction
- Increasing the agitation of the reaction mix
- Adding a catalyst
All of these methods will increase the rate of reaction, and thus increase the rate of consumption of the reagent. However, it must be balanced against the subsequent increase in the rate of heat evolved and so an assessment on whether the cooling capacity of the plant is able to contain this increase in heat.
Finally, reducing the feed rate of the reactants could also be considered: this will slow the total reaction time, but it will reduce the rate of reagent accumulation as it effectively allows more time for the reaction to occur and the rate-limiting reagent to be consumed before it accumulates to hazardous levels.
It is desirable to explore the impact of varying these different factors on the reaction. The Simular enables users to experiment in this way, to create a design of experiments that allows the optimum operating conditions to reduce the thermal risk to be identified.
Expanding on the example introduced earlier, in an additional study both the feed rate and process temperature were varied to understand the impact these factors would have on the level of accumulated reagent.
Reducing The Feed Rate
The effect of reducing the feed rate from 2g/min to 1g/min, whilst keeping the temperature of the process at 30 ⁰C, is illustrated in Figure 10. It reduces the accumulation of reagent to 32% at the point at which the feed stops, and it reduces the maximum reaction power from 15W to 8W. This reduction in maximum reaction power is critical, as it significantly reduces the cooling capacity required to keep the reaction isothermal and under thermal control.
Figure 10: Variation of reaction power with reagent feed rate
Increasing The Temperature
The level of accumulated reagent can be reduced further by increasing the temperature from 30 ⁰C to 40 ⁰C, whilst maintaining the slower feed rate of 1g/min. The impact of this is depicted in Figure 11. This reduces the level of reagent accumulated to 22 % at the point at which the feed is stopped, but it does result in an increase in the maximum reaction power from 8W to 10W.
Figure 11: Variation of reaction power with temperature
Cooling Capacity Considerations
Increasing the reaction temperature poses an interesting question: whilst the increase in temperature will result in less accumulated reagent, and thus a lower MTSR, should cooling fail and a thermal runaway occur, it does require a larger cooling capacity to deal with the faster rate of heat evolved. As before, if it is assumed that the 1L lab-based reactor is initially scaling up to a 2000 L reactor, the 10W reaction power for the reaction run at 40 ⁰C equates to a 20 kW cooling capacity requirement on scale-up, versus the 16 kW required for the same reaction run at 30 ⁰C. Therefore, although the 1g/min feed rate with 40 ⁰C process temperature is the desired combination from a safety point of view in terms of reagent accumulation, it may be that a lower temperature is required if the cooling capacity of the pilot plant is <20kW.
Ideally, a process should be as close to a dose-controlled reaction as possible: this means the reactant is consumed as rapidly as it is added and means that in the event of a cooling or agitation failure, it is usually sufficient to just switch off the feed to prevent a thermal runaway.
However, in cases where accumulation can’t be avoided, it is necessary to undertake a more detailed analysis of the risk.
During Process Development, thermal screening tools are used to identify thermal hazards, such as the onset of secondary reactions or decompositions of components in the reaction mix, should a thermal runaway of reaction occur. If modifying the reaction conditions were unable to completely remove the risk of this occurring, on moving to scale-up it is necessary to obtain more detailed and accurate information on the magnitude of the thermal runaway risks using adiabatic calorimetry.
Why Adiabatic Calorimetry?
Large scale reactors, like those found in manufacturing and pilot plants, effectively behave adiabatically – i.e. very little heat is lost to the surroundings. This is in contrast to the smaller vessels used in laboratories, where heat loss is much more significant. This poses a hazard on scale-up, as that heat will be retained within the system – at best requiring more plant cooling and at worst triggering a thermal runaway. This is due to the decrease in surface area to volume ratio in larger vessels, which provides less area for radiated heat dispersal.
Adiabatic calorimeters mimic these large-scale vessels in small scale systems. An adiabatic calorimeter will compensate for the heat losses with the application of carefully modelled and controlled system temperature increases.
There are two types of adiabatic calorimeter which find use at different stages of scale up. The more common, classic “ARC”-style adiabatic calorimeter, such as the Phi-TEC I, is used earlier in the scale up process as an additional reaction screening tool. The test cells used normally have a larger thermal mass than the reactants they contain and are described as having a high Phi (Φ) factor. A Phi factor is defined as:
As a result, the data from these high Phi-factor tests cells needs to be manipulated before the results can be scaled up.
However, the second type, such as the Phi-TEC II, uses low thermal mass test cells. This allows the Phi-TEC II to simulate plant-scale adiabatic runaways and generate data that can be used to define the safety measures that will need to be put in place on a manufacturing scale. As such, it tends to be used at the final stages of scale up.
The classic “ARC”-style adiabatic calorimeter sits somewhere in between the basic screening tools used earlier in process development, and the advanced, low Phi-factor adiabatic calorimeter used in final scale up. Accordingly, experiments undertaken with the Phi-TEC I deliver more information about the thermal hazards than those conducted on a TSu. Likewise, the Phi-TEC II delivers even richer process information. The Phi-TEC I is often used when the initial screening tests identify that the reaction being scaled up is highly exothermic and/or carries significant thermal risk, and thus more accurate information is required on the potential thermal runaways.
The Phi-TEC I allows the process safety expert to carry out a more detailed sample evaluation by studying the exotherm under adiabatic conditions, thus delivering a more accurate determination of the:
- Adiabatic temperature change of the thermal runaway of the reaction, ΔTad,r (and with it, a calculation of the MTSR);
- Time to Maximum Rate (TMRr) of the thermal runaway of the reaction;
- Onset temperature of any decompositions or side reactions (Td);
- Adiabatic temperature rise of any secondary thermal runaways (ΔTad,d);
- Time to Maximum Rate of the secondary thermal runaway (TMRd).
These parameters are illustrated schematically in Figure 12.
Figure 12: Schematic illustrating the key parameters for characterizing a thermal runaway
By providing a direct measurement of the sample temperature, coupled with a rapid heater response time to thermal changes, the Phi-TEC I ensures that exothermic events are fully tracked, and true adiabatic conditions are maintained.
Scale-up: Pilot Plant to Manufacturing
Before final scale-up to manufacturing, it is necessary to fully understand what safety control factors are required to mitigate the worst-case scenarios. The ability to fully simulate thermal runaway risks under manufacturing conditions in the lab is a valuable way of de-risking the process and ensuring the appropriate safety measures are planned and implemented, prior to manufacturing.
The use of an adiabatic calorimeter that utilizes low thermal mass (i.e. low Phi factor) test cells, such as the Phi-TEC II, means that very little of the heat produced during the reaction and runaway is consumed in warming up the test cell. It is this that enables the conditions of a large-scale plant to be emulated, with the adiabatic temperature change of the runway (ΔTad,d), the maximum temperature attained (Tend), the time to maximum rate of decomposition (TMRd) and the rate of pressure increase, representative of what would be expected to occur during a plant scale incident. Unlike adiabatic calorimeters with a high Phi factor, the reaction rate is not tempered by the high thermal mass of the test cell, and so little to no manipulation of the data is required to compensate for heat losses to the test cell and no knowledge on how the reaction proceeded or what the products of the reaction are is needed.
Comparing Reaction Scenarios
The Phi-TEC II also enables different reaction scenarios to be simulated and compared. In the example introduced earlier, the worst-case scenario that would release the most energy is the reaction of equimolar amounts of the alcohol and anhydride. Less severe scenarios can also be simulated, which in this case would involve reducing the relative amount of anhydride – effectively corresponding to reducing the amount of accumulated reagent.
Comparing the temperature results, it can be seen from the left-hand plot in Figure 13, that halving the amount of anhydride not only reduces the maximum temperature from over 200 ⁰C to 170 ⁰C, but also the time taken to reach that point. This change is further highlighted by considering the plot of the rate of temperature against temperature (right-hand plot in Figure 13), where the maximum rate reached drops from 4000 ⁰C/min to just over 1600 ⁰C/min when the amount of anhydride is halved. Reducing the level of anhydride to a quarter that of the alcohol increases this effect further.
As these reactions were carried out using a Phi-TEC II, on a low Phi-factor test cell and under adiabatic conditions, the parameters describing the exothermic event (MTSR, TMR, ΔTad, Tend) are true representations of what would be expected to be observed in a plant-scale incident.
Figure 13: Left – Temperature data from different alcohol/anhydride ratios at 30 ⁰C; Right – Rate of temperature rise data resulting from different alcohol/anhydride ratios at 30 ⁰C
As Figure 14 illustrates, a similar effect can be seen with the pressure data: when the level of anhydride is reduced, the maximum pressure reached is reduced and the rate of pressure increase is slowed.
Figure 14: Pressure data resulting from different alcohol/anhydride ratios at 30 ⁰C
The data illustrates that for the worst-case scenario, it barely takes 3 minutes (after the chemicals come together) for the reaction to runaway to completion, and in the process:
- reaching a temperature of approximately 200 ⁰C;
- a pressure of 16 bar;
- with the rate of temperature increase peaking at 4000 ⁰C/min.
In contrast, reducing the anhydride amount to a quarter that of the alcohol, the incident takes 12 minutes to reach worst conditions, which are now reduced to:
- a maximum temperature of 125 ⁰C;
- a maximum pressure of 8 bar;
- a peak rate of temperature change of 150 ⁰C/min.
Subsequently, this information enables the correct combination of reactor type and protection features to be selected, which could include:
- Emergency & evaporative cooling
- Controlled depressurization
- Vent sizing
Thus, the ability to directly compare the impact of different operating scenarios enables the necessary safety controls to be selected upfront for the manufacturing plant.
Solutions in Process Safety
It is clear that in order to ensure safe scale-up of chemical processes, consideration for Process Safety is essential throughout the product development funnel, from Discovery through to Scale-Up.
Discovery is characterized by the need for a quick assessment on whether to take a candidate molecule forward for development. Hazard screening enables rapid decisions to be made on whether a material is safe to progress with, based on the hazards identified. In Process Development, it is important to understand how hazardous the selected synthesis reaction itself is, and whether the plant cooling capacity is sufficient to mitigate these potential hazards. It is also important during this stage to explore how the synthesis reaction conditions can be modified to reduce the identified hazards. Hazard screening once again has a role to play, alongside reaction calorimetry for delivering understanding of the reaction itself. The progression into scale up is typified by an increasing focus on the true magnitude of the hazards identified, so that the plant safety design can be optimized accordingly. Adiabatic testing and HAZOP assessments come to the fore here. These considerations are summarized on the product development funnel in Figure 15.
Making safety an integral component of product development – by using a variety of tools to consider, identify, assess and then mitigate the potential hazards at each stage of the process – ensures confidence in delivering a successful and efficient route to market.
Figure 15: H.E.L solutions in Process Safety & Scale-Up