The study on high-speed centrifugal-regenerative pumps with an inducer (HCRP) is carried
out. Thecombined structure of inducer, centrifugal impeller, and regenerative impeller is
presented, and a theoretical parallelcombinatorial hydraulic design method is investigated.
The comparative experimental results show that efficiency insmaller capacity region, head
coefficient and efficiency in larger capacity region of HCRPs is few lower, much higherand
lower than those of high-speed centrifugal pumps, respectively, anci that the suction
performance of HCRPs isdetermined only by inducer. HCRPs can be more suitably applied to
deliver small-capacity high-head liquids inchemical and petrochemical industries.

A full second-order moment (FSM) model and an algebraic stress (ASM) two-phase turbulence
modelare proposed and applied to predict turbulent bubble-liquid flows in a 2D rectangular
bubble column. Predictiongives the bubble and liquid velocities, bubble volume fraction,
bubble and liquid Reynolds stresses and bubble-liquidvelocity correlation. For predicted
two-phase velocities and bubble volume fraction there is only slight differencebetween
these two models, and the simulation results using both two models are in good agreement
with the particleimage velocimetry (PIV) measurements. Although the predicted two-phase
Reynolds stresses using the FSM are insomewhat better agreement with the PIV measurements
than those predicted using the ASM, the Reynolds stressespredicted using both two models
are in general agreement with the experiments. Therefore, it is suggested to usethe ASM
two-phase turbulence model in engineering application for saving the computation time.

The cell model developed since 1950s is a useful tool for exploring the behavior of
particle assemblages,but it demands further careful development of the outer cell boundary
conditions so that interaction in a particleswarm is better represented. In this paper, the
cell model and its development were reviewed, and the modificationsof outer cell boundary
conditions were suggested. At the cell outer boundary, the restriction of uniform liquid
flowwas removed in our simulation conducted in the reference frame fixed with the particle.
Zero shear stress conditionwas used to evaluate the outer boundary value of the stream
function. Boundary vorticity was allowed to evolve tovalues compatible to existing stream
function at the free shear outer boundary. The fore-aft symmetry of vorticitydistribution
at the outer boundary is thought critical to ensure the continuity of inflow and outflow
between touchingneighbor cells, and is also tested in the modified cell model. Numerical
simulation in terms of stream function andvorticity based on the modified cell models was
carried out to shed light on the interaction between liquid andparticles. Lower predicted
drag coefficient by the modified cell models was interpreted with the feature of
flowstructure. The drag coefficient from the simulation was also compared with correlations
of drag coefficient reportedin literature. It is found that the modified cell model with
the uniformity of external flow relaxed and the fore-aftsymmetry of boundary vorticity
enforced was the most satisfactory on the overall performance of prediction.

Modelling and simulations are conducted on velocity slip and interfacial momentum transfer
for super-sonic two-phase (gas-droplet) flow in the transient section inside and outside a
Laval jet(L J). The initial velocity slipbetween gas and droplets causes an interfacial
momentum transfer flux as high as (2.0-5.0) × 104 Pa. The relaxationtime corresponding to
this transient process is in the range of 0.015-0.090 ms for the two-phase flow formed
insidethe LJ and less than 0.5 ms outside the LJ. It demonstrates the unique performance of
this system for application tofast chemical reactions using electrically active media with
a lifetime in the order of 1 ms. Through the simulationsof the transient processes with
initial Mach number Mg from 2.783 to 4.194 at different axial positions inside theLJ. it is
found that Mg has the strongest effect on the process. The momentum flux increases as the
Mach numberdecreases. Due to compression by the shock wave at the end of the L J, the flow
pattern becomes two dimensionaland viscous outside the LJ. Laser Doppler velocimeter (LDV)
measurements of droplet velocities outside the LJ arein reasonably good agreement with the
results of the simulation.

The axial concentration distribution of both particles with better wetting (forming non-
attached system)and poorer wetting (forming attached system) was investigated in a vertical
gas-liquid-solid fluidized bed of 4.2 cm indiameter and 130 cm in height with the solids
holdup less than 0.05. The one-dimensional sedimentation-dispersionmodel could be used
satisfactorily to describe the axial distribution of solids holdup by modifying only a
modelparameter, i.e. by means of the terminal settling velocity minus a certain value,
which is a function of gas velocityand considers the effect of an additional drag force
resulted from attached rising bubbles. The axial profiles of solidconcentration predicted
are in good agreement with experimental results. This model also explains reasonably
thedifferent axial distributions of solid concentration, i.e. the solids holdup decreases
as the axial height increases innon-attached system, but increases with the axial height in
attached system at a given gas velocity.

The selective oxidation of n-butane to maleic anhydride (MA) on a vanadium-phosphorus oxide
(VPO)catalyst was studied using on-line gas-chromatography combined with mass spectrometry
(GC-MS) and transientresponse technique. The reaction intermediates, butene and furan, were
found in the reaction effluent under nearindustrial feed condition (3% butane+15%O2), while
dihydrofuran was detected at high butane concentration (12%butane, 5%O2). Some
intermediates of MA decomposition were also identified. Detection of these
intermediatesshows that the vanadium phosphorus oxides are able to dehydrogenate butane to
butene, and butene further to formMA. Based on these observations, a modified scheme of
reaction network is proposed. The transient experimentsshow that butane in the gas phase
may directly react with oxygen both on the surface and from the metal oxidelattice, without
a proceeding adsorption step. Gas phase oxygen can be adsorbed and transformed to surface
latticeoxygen but it can not participate in selective oxidation. Adsorbed oxygen leads to
deep oxidation, while latticeoxygen leads to selective oxidation.

Based on Bloor & Ingham’s approach for determining the fluid field and on the analyses of loci offluid particles inside hydrocyclones, analytical models are developed for calculating the migration probability ofsingle-cone and two-cone hydrocyclones separating light dispersions. The calculated results are in good agreementwith Thew’s correlation at different flow rate, split ratio or fluid properties if the structural parameters keep thesame as those of Thew’s 35 mm hydrocyclone. The difference between predictions according to two-cone model andsingle-cone model is nearly negligible, which is very close to Thew’s original idea that major separation happensin the small cone-angle zone. Calculated results indicate that split ratio has little effects on reduced migrationprobability at least for F ≤ 20%, which is consistent to the conclusion drawn from Thew’s correlation that reducedmigration probability will remain constant when split ratio changes. Because no simplifying hypothesis is imposed onstructural parameters, the presented models can also be reliably applied in predicting the effects of those parameterson migration probability.

Formation of mineral scales on heat exchangers is a persistent and expensive problem. In
the presentpaper, the calcium carbonate scale inhibition by two inhibitors, polyacrylic
acid (PAA) and 2-phosphonobutane-1,2,4-tricarboxylic acid (PBTCA) has been studied in a
pool boiling system. It is found that PBTCA has abetter inhibition effect than PAA under
the identical conditions. X-ray diffraction(XRD) and Fourier transform-infrared(FTIR)
analyses demonstrate that the content of vaterite increases as inhibition effects increase.
Themetastable crystal forms of vaterite and aragonite are stabilized kinetically in the
presence of inhibitors. Therelationship between the inhibition effect and the fractal
dimension has also been investigated. The result showsthat the fractal dimension is higher
in the presence of inhibitors. The better the inhibition effect, the higher thefractal
dimension. The step morphology was observed by atomic force microscopy (AFM) images. It is
shown thatthe step space on the calcium carbonate surface increases in the presence of
inhibitors. Moreover, with the increaseof inhibition effect, both the step space and the
fractal dimension increase. The step bunching is found on thecalcium carbonate by AFM. The
better the inhibition effect of the inhibitor, the slower the step velocity in theunbunched
location. As a result, the step space becomes wider in the presence of PBTCA than that in
the presenceof PAA.

A mathematical model is developed for an industrial acrylonitrile fluidized-bed reactor
based on arti-ficial neural networks. A new algorithm, which combines the characteristics
of both genetic algorithm (GA) andgeneralized delta-rule (GDR) is used to train artificial
neural network (ANN) in order to avoid search terminatedat a local optimal solution. For
searching the global optimum, a new algorithm called SM-GA, incorporating ad-vantages of
both simplex method (SM)and GA, is proposed and applied to optimize the operating
conditions of anacrylonitrile fluidized-bed reactor in industry.

The possibility of multiplicity in an isothermal continuous mixed suspension-mixed product
removalcrystallizer is explored using the bifurcation theory. A process involving
agglomeration controlled precipitationis considered in which secondary nucleation occurs
simultaneously with primary nucleation. The determinantequations for the existence of
multiple steady states are developed and the multiplicity boundaries dependent on
thephysical and kinetic properties and operational parameters of the process are obtained
by resolving these determinantequations. The number of steady states in the precipitator
for various multiplicity regions is determined and thelinear stability of these steady
states is analyzed by using the Routh criterion.

In order to overcome the elementary heterogeneous nucleation while octahydro trisodium salt
of fructose-1,6-diphosphate(FDPNaa.8 H2O) is crystallized with ethanol precipitation at low
temperature, a new crystallizationmethod with alcohol precipitation combined with salt
precipitation has been presented. The ethanol-sodium ac-etate system for crystallization of
salt of fructose-1,6-diphosphate is based on the mechanism of crystallization ofFDPNa3.8
H2O in the ethanol-low temperature system. It is found that crystal size may be controlled
by regulatingtemperature or pH value of solution in the crystallization process, and the
crystal yield increases to 95% from 78%which obtained in the ethanol-low temperature
system.

The lumped time distribution functions were proposed, which can be used for describing the
dynamicsystems with two or more than two states of the end of growing polymer chain during
chain addition polymerization.Numerical analysis of the lumped time distribution functions
was carried out. The method for calculating molecularweight distribution of polymer in the
stable free radical polymerization and more general cases was developed basedon the lumped
time distribution functions.

The copolymerization kinetics of ethylene oxide and propylene oxide in an atomizing-
circulation reactorunder semi-continuous operation is studied which is of great importance
for molecular designation. The kineticparameters are obtained by numerical optimization of
the kinetic model.

The performance of supercritical fluid (SCF) as a solvent can be greatly affected by
addition of anentrainer to the system. In this study, a constant volume visual method is
used to measure the critical point ofCO2+n-butyraldehyde, CO2+ i-butyraldehyde and
CO2+alcohol binary systems and CO2+entrainer+trisodiumsalt of tri-(m-sulfonphenyl)phosphine
(TPPTS) ternary systems, which provides us good theoretical basis for super-critical
extraction and chemical reaction. The relationship between critical point and concentration
of the entrainerare discussed. The phase behavior of binary system and that of ternary
system are compared. The relationshipbetween the concentration of TPPTS and critical point
of binary systems are also discussed.

The pressure swing adsorption (PSA) models discussed here are divided into three
categories: partialdifferential equation model, electrical analogue model and neural
network model. The partial differential equationmodel, including equilibrium and kinetic
models, has provided an elementary viewpoint for PSA processes. Usingthe simplest
equilibrium models, some influential factors, such as pressurization with product,
incomplete purge,beds with dead volume and heat effects, are discussed respectively. With
several approximate assumptions i.e.,concentration profile in adsorbent, "frozen" column,
symmetry and heat effects of bed wall, the more complexkinetic models can be simplified to
a certain degree at the expense of a limited application. It has also been foundthat the
electrical analogue model has great flexibility to handle more realistic PSA processes
without any additionalhypothesis.

With the xanthan synthesis in Xanthomonas campestris as an example, two methods for
metabolicflux analysis of overdetermined system, the experimental data error minimization
method and the equation errorminimization method, are compared from their mathematical
basis, rationality of the results and the easiness ofcomputation. The results show that the
experimental data error minimization method is appropriate in metabolicflux analysis of
overdetermined system.

A series of ethylene oxide (EO)-propylene oxide (PO) random co-polymers (EOPO) were used to
formaqueous two-phase systems (ATPS) with ammonium sulfate. Effects of EOPO’s properties on
the phase separationbehaviors and on the partition of cephalexin and 7-
aminodesacetoxicephalosporanic acid (7-ADCA) in ATPS wereinvestigated. Both the molar mass
and molar ratio of EO to PO of EOPO could greatly influence partition behaviorsof
cephalexin and 7-ADCA as well as the binodal curve of ATPS. With the increase of molar mass
of co-polymeror the decrease of molar ratio of EO to PO, the critical point of ATPS
decreased, the binodal curve became moreasymmetry, and both cephalexin and 7-ADCA followed
the same tendency to partition into the polymer-poorbottom phase. The experimental results
shows that it is feasible to partition cephalexin and 7-ADCA in either thepolymer-rich top
phase or the polymer-poor bottom phase by choosing a specific phase-forming EOPO.

A response surface modeling approach for simultaneous optimization of optical purity and
yield wasapplied to the resolution of tartaric acid to study the effects of both the amount
of the solvent and the amountof the resolving agent α-methylbenzyl amine on the resolution
results. The direction of changing the experimen-tal conditions was determined from the
initial response study, and expected high yield of the pure L(+)-tartaricacid-L(-)-α-
methylbenzyl amine salt was obtained.

This paper discusses potential application of fuzzy set theory, more specifically, pattern
matching, in assessing risk in chemical plants. Risk factors have been evaluated using
linguistic representations of the quantity of the hazardous substance involved, its
frequency of interaction with the environment, severity of its impact and the uncertainty
involved in its detection in advance. For each linguistic value there is a corresponding
membership function ranging over a universe of discourse. The risk scenario created by a
hazard/hazardous situation having highest degree of featural value is taken as the known
pattern. Each sample pattern of hazard/hazardous situation with their known featural values
is then matched with the known pattern. The concept of multifeature pattern matching based
on fuzzy logic is used to derive the rank ordering of process hazards. In multifeature
pattern recognition/matching, a sample pattern is compared to a number of known data
patterns or a known pattern is compared to a number of sample data patterns. The process
assesses which known pattern resembles most closely data sample using Wang’s approaching
degree method. A methodology has been developed and the same has been exemplified by
presenting a case example with a limited number of hazards.

The distribution of vertical stress for both active and passive state in the silo with a
central innerdowncomer is reported in this paper. Experimental measurement of the axial
distribution of vertical stress for bothactive and passive state in the silo are in good
agreement with that predicted by theoretical analysis. The meanaxial stress is reduced due
to the presence of the inner downcomer in the silo.