Yi Fei,Bing Sun ,Fan Zhang ,Wei Xu ,*,Ning Shi ,Jie Jiang

1 Research Institute of Safety Engineering,SINOPEC,State Key Laboratory of Safety and Control for Chemicals,Qingdao 266071,China

2 School of Chemical Engineering,China University of Petroleum(East China),Qingdao 266000,China

Keywords:Reaction runaway Safety Chemical reactor Heat transfer Thermal risk assessment Early warning detection

A B S T R A C T Reaction runaway has longtime been an issue in chemical industry as it often leads to severe accidents if not controlled and inhibited properly.Herein we have reviewed several key considerations and procedures to prevent such phenomena,including inherently safer reactor design,thermal risk assessment and early warning detection of runaway,and pointed out that the basic principle underlying is necessary heat management and construction of resilient processes.For inherently safer reactor design,important factors such as heat removal,heat capacitance, flow behaviors and explosive behaviors have been investigated.The survey shows that heat exchanger(HEX)reactor and microreactor outperform traditional reactors.Meanwhile,we have looked into the effect of thermal risk ranking and safety operation region determining for thermal risk assessment,and the influence of runaway criteria and construction methods for early detection of reaction runaway as well.It shows that thermal risk assessment plays a key role on process design,and early warning detection system(EWDS)is preferable on prevention of reaction runaway.In the end,perspectives regarding inherently safer designs with the measures discussed above have been provided.

1.Introduction

Reaction runaway is a constant topic in chemical industry.It is well accepted that the intrinsic causes of reaction runaway is the heat accumulation during exothermic reaction processes.If the rate of heat generation is greater than the rate of heat removal,the amount of heat accumulated will raise the system temperature and further lead to the exponential increase of reaction rate,eventually resulting in a reaction runaway.Although the work of collecting data,analyzing accidents,optimizing processes,and absorbing experience has effectively reduced the probability and severity of a reaction runaway,they never completely eliminate it.The main obstacles are the variant and complex runaway mechanisms of a specific reaction,and the unpredictable external disturbers to a process.According to Barton and Nolan[1]and Saada et al.[2],the apparent causes of reaction runaway are:a)technical and physical causes(mischarging,defect in plant design,loss of process control,agitation failure,etc.)and b)human and organizational causes(operation error,management failure,poor emergency response,inadequate monitoring,etc.).Facing all these causes and chaotic scenes,the strategy of heat management becomes the most accepted route to prevent reaction runaway.As early as 1928,the Semenov's thermal explosion and runaway theory had demonstrated the concept of critical temperature,parameter sensitivity analysis and time to maximum rate under adiabatic condition(TMRad)[3].Although the kinetic model was simplified and the result was conservative,it still laid the foundation for the heat management strategies.Based on that,lots of efforts have been made on reaction process design and operation.

Inherently safer design(ISD)is to prevent human error and invalidation of facility to reduce the risk of a process by ways of minimizing,substituting,moderating,and simplifying[4].Strategies toward ISD can be divided into four categories[5]:

?Inherent— “Eliminating the hazard by using materials and process conditions which are non-hazardous.”

?Passive— “Eliminating or minimizing the hazard by process and equipment design features which reduce either the frequency or consequence of the hazard without the active functioning of any device.”

?Active—“Using controls,safety interlocks,emergency shutdown systems,mitigation devices such as sprinkler systems,or other active systems to detect potentially hazardous process deviations and to take corrective action.”

?Procedural— “Using operating procedures,administrative checks,emergency response,and other management approaches to prevent incidents,or to minimize the effects of an incident.”

Fig.1.Layer design of protection systems[6].

These four categories represent the characters of protection layers as shown in Fig.1[6].In practice,enhancing safety is a process of optimizing and balancing safety and economy[7].Building up protection system should take investment efficiency into consideration.Consequently,researchers focus more on the inner layers,and the measures belonging to the outer layers are also guided and optimized by the concept of“resilience”.

Since 1960,the focus of process safety management has evolved from technical safety to resilience as shown in Fig.2,where Visser[8],Knegtering and Pasman[9]and Jain et al.[10]contributed to the black lines,red line and the blue line,respectively.The word “resilience”is proposed to the safety area to describe“the capability to absorb and overcome unexpected,unforeseen,and unknown threatening disturbances that could otherwise result in a catastrophe”.It also refers to“the ability of the system to withstand a major disruption within acceptable degradation parameters and to recover with a suitable time and reasonable costs and risks”[11].Jain et al.[10]has identified four characteristic aspects of resilient process which are“early detection(recognizing weak signals through monitoring and anticipation),error tolerant design(including inherently safer design),plasticity(also characterized as resistive flexibility),and recoverability”.

This concept of“resilience”includes ISD and further gives ISD a more optimized and result-oriented criterion to filter the more meaningful and effective protection measures.Hence,although the quenching,inhibition,emergency venting[12–14]etc.have been proven effective,they are not within the scope of this review.The aspects of“inherent safer reactor design”,“thermal risk assessment”and “early detecting technology”are focused in the discussion on current progress.

2.Inherently Safer Reactor

As we know,the heat generation,heat removal and heat accumulation are three core parts of heat management.In practice,the heat generation is decided by the thermodynamics of specific reaction which lacks space for optimization.Meanwhile,the heat removal ability of batch or semi-batch reactors is also inherently limited.Therefore,novel reactors with larger specific area,enhanced heat transfer coefficient or greater thermal inertia factor have been researched warmly,and transforming batch or semi-batch process to continuous process becomes the preferable way for reactors ISD.

2.1.Heat Exchanger(HEX)reactors

2.1.1.Structural advantages of HEX reactor

Multifunctional heat exchanger(MHE)can accomplish mixing and reaction as well as heat transfer by supplying or removing the heat almost as rapidly as it is absorbed or generated by the reaction.HEX reactors are developed from the MHE and own many advantages such as better reaction control,improved selectivity,by-product reduction and better safety[15].Fig.3 shows a distinct magnitude differences of specific area and heat transfer coefficient in four kinds of reactors[16].The novel structure designed HEX reactors like metallic foams or offset strip fins have greater performance on heat removal.

Fig.2.Evolution in process safety and risk management methods[8].

Fig.3.Heat exchange capability for different reactors[16].

Fig.4.(a)Reaction channel design;(b)utility channel design;(c)HEX assembled[21].

From this point of view,a number of new reactors research and design have been reported[17–20].Among them,the HEX reactor designed by Théron et al.[21]gave a typical and better performance.It was designed and assembled like Fig.4 and the design parameters were detailed in Table 1.Obviously,the small channel width and the high volume ratio of process and utility fluid guarantee the performance.

The experiments in this reactor demonstrated good performance by offering sufficient residence time to complete chemical conversion,and the flow behavior approached to plug- flow in any Reynolds number conditions.These characters guarantee the transformation from batch to continuous processes successfully.Nevertheless,the temperature profiles and heat transfer intensification factors,which we moreconcern about,had highlighted good thermal performances of the HEX reactor(5000–8000 kW·m?3·K?1when process flow rate ranges from 8.7 to 15.0 kg·h?1)as Fig.5 showed.

Table 1Geometrical properties of the HEX[21]

Fig.5.Intensification factor of HEX reactor with the process flow rate[21].

The test reaction of sodium thiosulfateoxidation by hydrogen peroxide had been carried out which owns great reaction heat(ΔHr= ?586.2 kJ·mol?1of Na2S2O3).According to the calculation,at the operating temperature of 60°C,the heat exchanged in the HEX reactor per unit of fluid volume(Q/V)equaled to 1.2×103k W·m?3.Compared with the classical batch reactors whose volume ranged from 1 L to 1000 L and Q/V ranged from 1200 to 100,the maximal heat removal ability of HEX was at least 10 times greater than batch reactors.This offered the reactor a strong capacity of heat removal.

2.1.2.Material advantages of HEX reactors

Besides the structural advantages of HEX reactors discussed above,the influence of material on the thermal characters has also been researched.As know n,the thermal characteristics like thermometric conductivity and thermal diffusivity are different from material to material.The thermometric conductivity has a positive correlation with thermal exchange capacity,and the level of thermal diffusivity decides the rate of heat spreading.Despènes et al.[22]demonstrated the difference between Plexiglas,stainlesssteel(SS),silicon carbide(SiC),SiC/stainless(SiCS)steel and SiC/aluminum(SiCA)in Fig.6,which showed two meaningful trends.

?With the introduction of SiC,whatever pure SiC or mixed with other material,the thermal exchange capacity increased obviously.

?Compared with conventional material like SS,the thermal exchange capacity of material with SiC increased obviously with the process flow rate.

From view of safety,the materials of SiC offered HEX reactor greater thermal removal capacity which has already been in high level and also owned potential extending space as process flow rate raising in emergency situation.The positive effect on heat removal with increasing process flow rate would be less limited by thermal properties of material like stainless steel.This is a preferable way to enhance resilience of process which means a high level of fault-tolerant and inherently safety.

Meanwhile,there is another way to promote the resilience of a system that is elevating the thermal capacitance of HEX reactors.Thermal inertia factor is a parameter to characterize thermal capacitance which is defined by

which indicates the absorption ratio of total heat of reaction by reaction mixture.In a batch reactor,the typical thermal inertia factor does not generally exceed 1.2,meaning that the main heat of reaction contributes to the maximum temperature of synthesis reaction(MTSR).On the contrary,HEX reactors own greater mass or specific heat.Hence the greater thermal capacitance results in lower MTSR,which means less igniting feasibility of secondary decomposition reaction in mixture.

Bena?ssa et al.[23]reported the changing trend of maximum temperature in different thermal inertia scenarios of the HEX reactor(Fig.7).The utility flow(UF)was supported by two stainless steel plates:transition plates(TPs)and sandwich plates(SPs),and was separated with reactive flow plates(RPs)by SP.A typical exothermic reaction was carried out in this reactor and a scene of flow stoppage was observed.As Fig.8 show s,the “normal operation”temperature profile was recorded and the hypothetic situations 1)–5)were also calculated.These situations contained different functional plates,which leaded to different thermal inertia factors of 1.00,1.28,1.41,1.69 and 11.96 respectively.As can be seen,the temperature of situation 5),which was closed to the normal operation line,was under control all the time.On the contrary,other 4 situations all showed various degrees of temperature runaway.In conclusion,the part of energy dissipated by reactors played a key role in reaction temperature control,and the HEX reactors owned higher level of inherent safety.

Fig.7.Structure of the HEX reactor by the sequence of functional plates[23].

Fig.6.Thermal exchange capacity of different materials[22].

Fig.8.Final temperature profiles reached by the reaction mixture after flow s stoppage according to five hypotheses[23].

Based on the same purpose,Despènes et al.[22]investigated the influence of material properties on the thermal capacitance of HEX reactors.As Fig.9 shows,in the period of 0–35.5 min,the temperature and flow rate of process fluid and utility fluid were stable.As the utility fluid stopped at 35.5 min,the outlet temperature of process fluid showed various degree of increase,that can be explained semi-quantitatively by parameters in Table 2.The heat absorbed by reactor and dissipated to external environment is depended on mCpand UextAextrespectively.The two parameters of SiC are both weaker,which result in a higher outlet temperature.On the contrary,the value of UextAextof SS and mCpof SiCA is greater,which contributes to more heat dissipation.

From the overall consideration of thermal exchange capacity and thermal capacitance,the SiC mixed material is a preferable route of material optimization from the view of heat management and passive safety.

2.2.Microreactors

Microreactor is a kind of continuous reactor whose reaction channel dimension is smaller than 1 mm and has been accepted as inherently safer reactors.Based on microreactors,a number of inherently high risk processes like potential reaction runaway or gas-phase explosion,have been achieved and de-risked[24–30].Due to the bigger specific area and higher ratio of reaction volume and inertia volume,microreactors show greater advantages than HEX reactors on specific area,heat transfer coefficient and thermal inertia factor,which meansa greater capacity of heat management.In addition to this,the other two obvious characteristics of mass transfer enhancement and explosion prevention are the guarantees for microreactors to handle risks inherently.

Table 2Thermal capacitance of each reactor[22]

2.2.1.Mass transfer enhancement

In most situations,the flow in microchannel is laminar,that means the mass transfer is mainly based on diffusion in radial direction and the velocity is distributed as parabola in the axial direction.How ever,due to the microscale of radius,the time for sufficient radial diffusion can be much shorter than reaction residence time and the axial backmixing can be neglected.Hence,in macroscopic view,the flow behavior in microchannel is close to plug- flow,which also has been confirmed in publications[31–33].Due to enhanced mass transfer,the efficiency and performance of reaction process are also enhanced,which leads to the significantly decrease of residence time and material inventory in reactors.This characteristic is in good agreement with principles of ISD.As assessment by Zhang et al.[34],for the reason of less residence time and smaller reactor volume,the Beckmann rearrangement reaction carried out in the microreactor showed a less damage index(DI),indicating a greater improvement of inherent safety level.

Meanwhile,it is well accepted that the reaction kinetic investigation always being an intrinsic way to comprehend the reaction process for the purpose of risk control.How ever,in the conventional way,the kinetic investigation is time consuming and complex,and for most time,inaccurate.On the contrary,the kinetic parameters obtained by using microreactors are close to the intrinsic reaction kinetics due to the absence of mass and heat transfer effect as well as smaller concentration and temperature gradient.Salmi et al.[35]and Nijhuis et al.[36]conducted kinetic investigations of oxidation of ethylene to ethylene oxide,oxidation of propene to propene oxide in microreactors respectively,which all acquire positive results as expectation.

2.2.2.Explosion prevention

Fig.9.Comparison of temperature variation for different reactor materials.[22].

Technically,due to the extremely fast reaction and heat generation rate,explosion is indeed a kind of reaction runaway.The advantage of explosion prevention in microreactorsis based on the strong capacity of heat removal which leads to explosion quenching.As previously mentioned,numerous explosive reaction processes have been carried out in controlled and steady conditions.Their gas-phase mixture composition and concentration are undoubtedly lying in the range of explosion,for instance the oxidation of ethylene to ethylene oxide,direct synthesis of hydrogen peroxide from hydrogen and oxygen,etc.How ever,along with the extensive research and utilizing of microreactor advantages,it was demonstrated that limitation and specificity should be taken into consideration.

Heinrich et al.studied the explosion inside the microreactor with the ignition of catalyst[37].The stoichiometric mixture of ethene and oxygen was tested in temperaturerange of 548–593 K and the mixture flow rate ranged from 240 to 669 ml·min?1.It turned out that only if the condition reached 593 K and 669 ml·min?1,the mixture ignited.This negative result is reported for the first time that in specific high temperature and high flow rate condition,the ethene–oxygen mixture explosion could not be quenched.Meanw hile,Libber et al.carried out the propagation experiments by the same apparatus showing moredetailed results as in Fig.10[38].The stoichiometric mixture of ethene and oxygen was tested in a microchannel of 0.25 mm and P1–P8 were dynamic pressure sensors arranged along with the explosion propagating path from ignition to exit.It showed the initial pressure having a huge influence on the propagating behavior.Only in the lower initial pressure of 32 kPa,the explosion was prevented successfully.

Fig.10.Detonation propagation velocities of ethene–oxygen mixture at various initial pressures between neighboring sensors[38].

Furthermore,aimingat the corerole of micro-channel dimension,Fischer et al.[39]and Liebner et al.[38]studied the safe maximum diameter of micro-channel in ethane–oxygen mixture.The results showed that the so-called λ/3-rule[40](λ being the detonation cell width),established in macroscopic tubes,is also applicable to microreactors in that mixture system.Detonations usually propagate if λ/3 is lower than the tube diameter.

From the above results of research,a more common conclusion is suggested that microreactor is indeed a kind of inherently safer reactor but has limitation and precondition.The heat removal capacity of microreactor is huge but limited,the raise of initial pressure,temperature or flow rate etc.of flammable mixture will enhance the instantaneous power of reaction heat releasing.When it is above the heat removal capacity,the runaway will occur.Hence,for each specific reaction process,the mixture composition,operation temperature and pressure,geometry structure of micro-channel,way of explosion ignition,etc.should be taken into account,and more beforehand investigation will lower the risk.

Based on the advantages discussed above,there would be three directions to promote the further research and application of inherently safer reactors.Firstly,developing the new reaction mechanism or new catalyst to raise the reaction rate,conversion or selectivity,which could benefit the batch to continuous transposition.Secondly,optimizing the structure or material of inherently safer reactors to enhance the mass and heat transfer.For instance,utilizing the structure of annular channel to further weaken the axial backmixing in microreactors[32].Thirdly,instead of scaling up the size of a single reactor,the industrialization of microreactor would be based on reactor stacking.Hence,the research on mixing and flow behavior in microscale or mesoscale should be paid more attention.

3.Thermal Risk Assessment

Just like reactor design,the thermal risk assessment also belongs to ISD and is an effective and indispensable way to achieve two applications:recognizing and ranking the thermal risk of reaction processes,and determining the boundary condition to pre-design the safety operating region.

3.1.Ranking of thermal risk

As for this application,whatever theoretical framework,experimental measures and ranking criterion are mature and have been applied for years in chemical industry.Still,due to the importance of this function,this section gives a brief scan of typical related works in Table 3.

The common features of the above researches can be summarized as followings.The assessment work can be divided into two parts:aiming to the material(reactants,products,intermediate products)decomposition reaction and the target synthesis reaction.With the reasonable assumption of adiabatic conditions,the parameters of ΔTad,TMRad,MTSR,etc.are acquired and utilized without consideration of external heat removal.Therefore,the thermal risk assessment result is intrinsic and conservative.Utilizing the parameter of ΔTadand TMRadto rank the possibility and severity of reaction runaway,ΔH is used to characterize the potential energy releasing.And parameter comparison of TP,MTSR,TD24,etc.gives a detailed ranking of thermal risks and decides the igniting feasibility of secondary decomposition reaction[47].

3.2.Determination of boundary condition

Based on the first aspect discussed above,a further application of boundary condition determination could be researched if the reactionkinetic parameters and heat transfer parameters of the synthesis reaction are available.As know n,there are numerous parameters helping in characterizing the reaction state like temperature,pressure,p H,composition,etc.How ever,for the most time,due to the lack of comparison criterion,the absolute value of parameters seems meaningless.Taking temperature for example,in normal operation,the temperature or rate of temperature rising etc.should not over the preset critical limitation(LIM),such as TLIM,(d T/d t)LIM,etc.Hence the criteria of T＞TLIM,d T/d t＞(d T/d t)LIMor d2T/d t2＞(d2T/d t2)LIMare the most common boundary condition of processes.How ever,the determination of critical limitation is uncertain.According to the actual reaction condition,many factors should be taken into consideration to determine the boundary conditions,such as the mass transfer kinetics,heat balance calculation(heat of stirring,reactant dosing,and heat removal by coolant etc.),reaction kinetics,and others.For different reaction processes,the boundary conditions were widely divergent[48–50],but the research method was common.In a typical work,Woezik and Westerterp[51]chose the oxidation of 2-octanol by nitric acid which occurred in a two-phase system where the reaction proceeds via two steps:2-octanol to 2-octanone and 2-octanone to carboxylic acid.Based on the mass and heat balance calculation and kinetic parameter collection[52],the mathematical model was constructed.Meanwhile,the criterion for dangerous state recognition proposed in[53,54]was adopted.The experiment revealed the varying trend of MTSR along with the coolant temperature in different dosing times or cooling capacity.Also,the boundary condition of cooling capacity UDa/?＞ 45 was confirmed,which guarantees the invariable safe operation of first step oxidation.Furthermore,the boundary condition of Tcoolto keep reaction limited in the“quick onset,fair conversion and smooth temperature profile”(QFS)region[53]for both steps,and the boundary condition of Exto prevent reaction runaway in any coolant temperature was also predicted and confirmed.

Table 3Typical works of thermal risk ranking

As can be seen,thermal risk assessment offers a pre- filtering of thermal risk and a pre-design of safe operating boundary.Although the assessment processes are time and resource consuming,the result is applicable and meaningful.As for the further research and application of thermal risk assessment,research would be better focused on the ranking method of continuous process and heterogeneous process.Also,the acquiring method and approach of reaction kinetic parameters would be the main obstacle and the focus in the near future.

4.Early Detection of Reaction Runaway

During the operation of reaction process,the detection timing of reaction runaway is critical.The earlier the detection is,the moretime and space to launch emergency response.

4.1.Criteria of early detection

The early detection methods are divided into two kinds:model based and model-free.They are both based on the real-time parameters measurement like temperature or pressure.How ever,the difference was that the model-free route carried out the measurement treatment by a novel mathematical theory like chaos.On the other hand,the model-based route constructed models by conventional mass and heat balance calculation and reaction kinetics acquired off-line.Whatever the route is,the criterion selection of reaction runaway was critical.Due to the complexity of detailed mathematic model and the lengthy derivation path of each criterion,only the results are shown here.Barkelew[55]set the E(Tmax? Tθ)/RTθ2≥ 1 as the runaway region.Adler and Enig[56]set the criterion of dθ/d z＞ 0 and d2θ/d z2＞ 0 as runaway region.Hub and Jones[57]set the criterion of d(T?Tθ)/d t＞ 0 and d2T/d t2＞0 as the runaway region.And seemingly as the most promising one,the divergence criterion proposed by Strozzi et al.[58]set the?F(θ,z)＞ 0 as the reaction runaway region.

Fig.11.Runaway boundaries using different criteria for a first-order reaction[58].

Fig.12.Runaway boundary using different criteria for an autocatalytic reaction[58].

The comparison of each criterion was also carried out as Figs.11 and 12[58].In Fig.11,the abscissa B was dimensionless adiabatic temperature rise which represented the heat release capacity of reaction mixture and the ordinate α/B represented the heat transfer capacity of the reaction system.The comparison aimed at a first order reaction.Five curves showed the same trend which demonstrated the consistency of the runaway boundary conditions.However,compared to the model-based limited criterion of Barkelew,the divergence criterion showed an earlier detection of runaway than the other two model-free limited criteria.Also,the autocatalytic reaction was compared in Fig.12.In this situation,the Adler criterion gave an extremely conservative result.On the other hand,due to the characteristic of autocatalytic reaction,the two items of Hub criterion were always positive which gave warning all the time.Apparently,the divergence criterion was earlier and more reliable than the other two on-line ways.More than that,the divergence criterion was applicable for both model-based and model-free route.

4.2.Utilizing of divergence criterion

4.2.1.Model-based route

Utilizing the advantages of divergence criterion,Guo et al.[59]gave an early detecting simulation of reaction runaway in adiabatic condition.Evolving from?F＞0,the criterion of

was proposed.The experiment was carried out in a semi-batch way and one reactant was dosed continuously.In Fig.13a and b,the ordinates are the reactant conversion and the value of criadrespectively.The normal operation would lead to reaction runaway as black line.Only if criad＞0 was detected and dosing was stopped at τ=0.239,the violent conversion rate and runaway were avoided.How ever,as the author concluded,to utilize this criterion on-line,the thermal and kinetic parameters of reaction,real-time temperature and conversion were needed.These would obstacle its wide application.

In addition,Casson et al.[50]gave a thorough research of the epoxidation process of soybean oil with the detailed model construction.As in Fig.14,time period(τ)of 0–1 describes the start-up state and steady state which were characterized by both divergence criterion and Hub criterion.The positive signal was explained due to the fast epoxidation rate in the fresh mixture.As the stoppage of cool reactant at τ=1,the reaction runaway was detected by both criteria.Apparently,the signals acquired by the divergence criterion were earlier and more obvious.

Fig.13.Simulated conversions profile of dosing continues scenario and dosing stop scenario if criad＞0[59].

4.2.2.Model-free route

Technically,the abbreviation of“EWDS”,which is based on chaos theory,belongs to the early detection method of Strozzi et al.[58]exclusively.The system consists of two parts:the phase-space reconstruction,and the real-time state parameter measurement like temperature or pressure.As for the phase-space reconstruction,the three elements are critical:reconstruction method,time delay and embedding dimension.Due to the derivatives being sensitive to noise,the delay coordinate method of reconstruction was chosen in EWDS.Omitting the complicated mathematic treatment and deduction,the results are listed below.As showed in Fig.15,in the contrast to the model-based divergence method,the model-free method of phase-space reconstruction acquiring the equal results and the same trends of divergence profiles.Due to the high heat-transfer coefficient,the divergence of column(a)was negative which means the non-runaway situations.On the other hand,the column(b)was opposite.

As commented in[58],the research explored“a new route,between the simple use of measured variables and the complex model-based state reconstruction techniques”to achieve the early detection of reaction runaway.Based on this theory,several related works have been reported.Bosch et al.[60]gave a comparison by utilizing the real-time measurement of temperature and pressure to calculate the divergence of the system.Results are showed in Fig.16.The criterion ΔVPS＞ 0 was evolved from?F＞0.Based on that criterion,the temperature or pressure way was both effective in early detection.However,in the situation of vapor system(acetic anhydride hydrolysis with surfactant),the temperature based detection was earlier than pressure based one.On the other hand,in the gas generating system(tert-butylperoxy-2-ethylhexanoate decomposition),the result was opposite.This showed that the real-time parameter is not limited to temperature.The sensors of parameters,whose performance are more sensitive,reliable,and also suitable for the specific system,need careful discretion.

Fig.15.Model-based calculated(.-.-)and reconstructed divergence profiles with different α[58].

Fig.16.Reconstructed ΔV PS using temperature and pressure for a vapor(Exp4)and a gassy(Exp8)runaway experiments[60].

Ampelli et al.[61]gave another application of the divergence criterion in model-free approach.In the batch-w ise polymerization of methylmethacrylate,the divergence was calculated on-line to contrast with the jacked and reactant mixture temperatures in Fig.17.The ΔVPSgave a synchronousand accurate response along with the reaction process.At 0–4 min,the fluctuating of ΔVPSsignal which was under the horizontal line of 0.01,was caused by heating and cooling characteristics of calorimeter.At about 10 min,the obvious signal was caused by sharp increasing of TR.The EWDS warning is 3.58 min earlier than the time instant of the maximum temperature.In addition,with the monitoring of EWDS,the timing of inhibitor injection was determined by the ΔVPSsignal as Fig.18.Comparing the left profile with the right one of Fig.18,the inhibitor was effective to prevent reaction runaway under the guidance of EWDS.On the other hand,adding at the start of reaction,hydroquinone(Hq)or 1,4-benzoquinone(Bq)just played a role of delaying reagents with the introduction of induction period and had little moderation of temperature arising.

Fig.17.Profiles of reactant mixture(T R)and jacket(T J)temperature(upper part)and phase space volume change ΔV PS(lower part)vs.time.[61].

As can be seen,by combining the divergence criteria with the phases pace reconstruction,the EWDS achieved the early detection of reaction runaway in laboratory.How ever,as for the future application in industry,problems still remain.Taking the selection of time delay for instance,if the time delay is too small,each coordinate would be too close to separate;if the time delay is too large,the dynamic system at one time is irrelevant with itself at the later time.Hence the selection of the time delay,the embedding dimension and the reconstruction method is critical for phase-space reconstruction.In addition,noise is always problem for detection systems.The noise reduction scheme would be helpful to stabilize the feedback of EWDS.

5.Conclusions

The heat management is the core strategy of ISD,which is consisted of three parts:heat generation,heat removal and heat accumulation.The design and optimization of reactors is mainly based on the latter two aspects.Based on that,the four research and development strategies toward reactors ISD have been formed.First of all,transforming batch and semi-batch reactors to continuous reactors will elevate the specific area and thermal capacitance.Secondly,combining the reactors and heat exchangers with special structures to form the HEX reactors.In this process,the heat transfer coefficient is enhanced and the extra thermal inertia is introduced.The enhancement of these three factors of specific area,thermal capacitance and heat transfer coefficient belong to the strategy of“passive”,which enhance the resilience of process.Despite the reaction runaway happening,the influence and consequence could be absorbed and overcame.Thirdly,microminiaturizing the reaction channels.Due to the extremely reduced dimension,the flow behavior and explosion propagation behavior is changed obviously.The enhanced mass and heat transfer are beneficial to decrease the residence time and material inventory and also to acquire instinct kinetic parameters.Meanwhile,although the character of explosion prevention should be utilized with limitations and preconditions,microreactor is indeed a kind of inherently safer reactor.Fourthly,optimization of reactor materials.Capacities of heat removal and heat capacitance should be taken into overall consideration,and blending modification like Sic/metal is a recommended way.

Fig.18.Reactant mixture(T R)and jacket(T J)temperature profiles with situations of standard,hydroquinone(Hq)and 1,4-benzoquinone(Bq)injection.Left,injection at EWDS warning point;right,injection at the start of reaction[61].

Thermal risk assessment plays a role of pre- filtering of thermal risk and pre-design of safety boundary condition.With the ranking of△Hd,TP,MTSR,TD24,etc.,the potential energy releasing,possibility and severity of reaction runaway,igniting feasibility of secondary decomposition reaction are quantified.On the other hand,with the model construction of the reaction process including the mass and heat balance calculation and kinetic parameters researching,the boundaries like QFS region,runaway region are sketched.Apparently,the acquiring of reaction kinetic parameters is the main obstacle and will be the research focus in the future.

As for the reaction process operation,construction of EWDS is a valuable inherently safer procedure,which also enhance the resilience of process.The criterion of divergence has been compared with other criteria and been proven earlier and more reliable to carry EWDS.On the contrary of thermal risk assessment,model-free route is preferred by researchers for the reasons of pervasive and robust.Phase-space reconstruction based EWDS has been applied in several reaction runaway situations and has been cooperative worked with the other technical and emergency measures effectively.Although the accurate determination of the time delay and the embedding dimension is uncertain,the EWDS is still the promising route to detect reaction runaway early.