1756
Ind. Eng. Chem. Res. 1994,33, 1756-1764
Knowledge-Based System for Mixing Type Selection Tuomas Koiranen, Andrzej Kraslawski,’ and Lars Nystriim Department of Chemical Technology, Lappeenranta University of Technology, P.O. Box 20, 53851 Lappeenranta, Finland
Mixing is an example of a process where the selection of the equipment is practically realized “by hand”. There are the following benefits from the introduction of knowledge-based system in mixing equipment selection: faster evaluation and generation of alternatives, systematic search of alternatives, and storage of experience. The subject of this paper is an expert system for the selection of mixing type. The main factors that are taken into account in the selection process are phases to be mixed, mixing requirements, residence time, mixing applications, maintenance, cost, type of application, and viscosity. The discussed mixing types are mixing in stirred vessel, dynamic in-line mixer, static in-line mixer, jet mixing in tanks, tee-pipe or injector, pneumatic mixing, roll mill, colloid mill, and kneader. In the presented system the knowledge base is separated in a n MSEXCEL spreadsheet, Nexpert Object acts as an inference machine, and MS-EXCEL spreadsheets work as a user interface. Two examples of mixing type selection are presented.
Introduction There is a growing recognition for the need of design automation. However, the nature of engineering design is not uniform. It is a mixture of heuristic rules and rigorous calculations. An engineering design where calculations are well determined is characteristic for the later phases of design. The first part of the design activities is based on the heurisitics and individual experience. Design steps where human judgement is a main factor are very difficult perform in a totally automatic way. As a consequence, the application of computers is a routine activity in the latter stages of design and is rather seldom at the beginning of the design process. The goal of total computer aided integration of the design process, from the process concept to the detailed process description, has not been achieved up to now. The introduction of the knowledge-based systems in the engineering practice started in the mid 1980s was not a success. The early systems had poor user interface, were slow, and were equipped in incomplete knowledge. Moreover they were difficult to maintain and support. After a few years of experience a new, modified generation of the knowledge-based system has been introduced into the practice. At the present, use of the knowledge-based systems (expert systems) is more and more popular in process engineering. There are the following main areas of use in process engineering: process design, planning and operations, modeling and simulation, fault diagnosis, process control, and product design. One of the decisions to be taken in an early phase of design process is equipment selection. There were several attempts to apply computers to the selection of equipment in process engineering (Lahdenpera, 1989; Blass and Goettert, 1990;Chuang et al., 1992;Hanratty and Joseph, 1992). Recently more and more attention has been paid to the mixing equipment selection (Yang et al., 1993;Knoch and Bottlinger, 1993; Koiranen et al., 1994a,b). Mixing is an example of the process where the selection of the equipment is practically realized “by hand”. There are two main reasons for a such situation: a great variety of processes and a lack of “applicable” theoretical tools. In such a situation experience and heuristic are the main engineer’s tools. There are the following benefits from the introduction of a knowledge-based system in mixing
equipment selection: faster evaluation and generation of alternatives, systematic search of alternatives, and storage of experience. Design of mixing process is divided into the selection of the mixing type and design of mixing system. As a consequence, a knowledge-based system should have a similar structure. The subject of this paper is an expert system for the selection of mixing type. The mixing types are mixing in stirred vessel, dynamic in-line mixer, static in-line mixer, jet mixing in tanks, tee-pipe or injector, pneumatic mixing, roll mill, colloid mill, and kneader. The later phase in design of a stirred vessel consists in determination of tank diameter, tank bottom, impeller type, impeller entering, baffles, and impeller supports and power consumption calculations. Their detailed description is given by Koiranen et al. (1994a,b) The article is divided into four main parh. The first one, Engineering Factors in the Selection of Mixing Types, gives basic information about some aspects of selection. Next, Selection Concept presents information about the suitability of the given alternative for the given situation. System Overview presents the program architecture and examples. In Summary basic system features are presented.
Engineering Factors in the Selection of Mixing Types The factors influencing the selection of mixingtype have been extensively presented in the main textbooks and papers in mixing (Liquid Agitation, 1975-1976; McDonough, 1992; Nagata, 1975; Oldshue, 1983;Uhl and Gray, 1986; Tatterson, 1991). As mentioned above, the main mixing types taken into account in the paper are mixing in stirred vessel, dynamic in-line mixer, static inline mixer, jet mixing in tanks, tee-pipe or injector, pneumatic mixing, roll mill, colloid mill, and kneader. The main factors influencing the selection are phases to be mixed, mixing requirements, residence time, mixing applications, maintenance, cost, type of application, and viscosity. The composition of the above mentioned factors is as follows. Phases to be mixed: liquid-liquid; liquid-solid; gasliquid; gas-liquid-solid.
oasa-5aa~19412633-i756$04.50/0 0 1994 American Chemical Society
Ind. Eng. Chem. Res., Vol. 33, No. 7, 1994 1757 Table 1. Fragment of the Knowledge Table
mixer stirred tank dynamic in line static in line jet mixing tee pipe or injector pneumatic roll mill colloid mill kneader ideal mixer
liquidliauid 1 1 1 1 1 1 1 1 1 1
mixing operation liquidgas; gas-liquidsolid liauid solid 1 1 1 0.2 0 1 1 1 1 0 0 1 1 0 0 0.3 1 0 1 0 0 1 0 0 1 1 1 1 1 1
Mixing requirements: shear; flow; fluid motion; heat transfer; gas dispersion; liquid dispersion; solid dispersion. Residence time: hours; minutes; seconds. Mixing applications: gas suspension; absorption; solid suspension; slurrying; crystallization; leaching; dissolving; liquid homogenization; liquid suspension; extraction; emulsification;storage homogenization;chemical reaction; fermentation. Maintenance: impossibility of blockage; cleaning. Cost: capital cost; energy consumption; working time. Viscosity: batch process; continuous process. The suitability of every type of mixing is estimated from the point of view of every criterion. The validation of the several criteria is presented in Table 1. It is based on the literature information (Liquid Agitation, 1975-1976; McDonough, 1992; Nagata, 1975; Oldshue, 1983; Uhl and Gray, 1986; Tatterson, 1991.
Selection Concept The selection factors are divided into three groups, crucial, meaningful, and unimportant criteria, based on the information obtained from the main textbooks and papers in mixing (Liquid Agitation, 1975-1976; McDonough, 1992; Nagata, 1975; Oldshue, 1983; Uhl and Gray, 1966; Tatterson, 1991. There are the following reasons for such diversification. A group of crucial parameters ensures a potential suitability of a given type of mixing to a given type of process. The role of crucial criteria is elimination of the infeasible solutions. These parameters are checked first in order to make the selection candidate group smaller. The second group of criteria, meaningful parameters, is used for the ranking of the alternatives. The third type of criteria, unimportant parameters, is used to eliminate the criteria that are not important in the given type of applications. The suitability of the given mixing type i from the point of view of the given selection factor (criterion) is presented in Table 1. The only fuzzy criterion is viscosity of the mixture. A graphic illustration of fuzzy number is presented in Figure 1. The meaning of the symbols in Figure 1 is as follows: p ( x ) = membership function; x = viscosity; a = lower limit of possibility range; b = lower limit of certainty range; c = upper limit of certainty range; d = upper limit of possibility range. If the viscosity of a medium to be mixed is in a range from b to c, it is certain that the given equipment is good for this application. If the viscosity of a medium to be mixed is in the range from a to b or from c to d, then a given mixing equipment is possibly applicable. Its suitability equals a value of the membership function. The value of membership function is from 0 to 1. The greater the value of the membership function, the more suitable is given alternative. Introductory information
gas suspension 1 1 1 0 0 1 0 0 1 1
absorption
application solid suspension slurrying
1 0.3
0.3 0 0 0.8 0 0 0.1 1
a
1 0 1 0 0 0.3 0.3
1 0 1 0 0 0 1
0.5
0.2
1 1
1 1
crystallization
1 0.1 0.3 0 0 0 0 0 0 1
b
c
d
x
Figure 1. The membership function. Table 2. Control of the Inference Mechanism: 'Selection Table" parameter fluids mixed
importance crucial
process type residence time shear flow application fluid motion heat transfer solid dispersion gas dispersion liquid dispersion impossibilityof blockage cleaning easiness installation costa energy consumption working time
crucial crucial no no meaningful meaningful meaningful meaningful meaningful meaningful meaningful meaningful meaningful meaningful meaningful
a
importance factor 1
threshold values 0.67' 0.5b
1 1 1 1 0.8
0.7 0.7 0.7 0.7 0.7 1 0.6 0.4 0.2
0.5
Crucial variables. Viscosity variable.
about fuzzy numbers and operations on them is presented in Zimmermann (1992). The suitability Ci of the given mixing type i for the given mixing problem is determined based on the knowledge introduced in Table 1, importance factors presented in Table 2, and input data obtained from the user, Table 3. The input data are either numerical or string type. The suitability Ci of the given mixing type i for the given mixing problem is calculated according to the equation Ci = Mi/Mid4
(1) In the case of a numerical input value introduced by the user, Mi is defined as
In the case of a string input value given by the user, Mi
1758 Ind. Eng. Chem. Res., Vol. 33, No. 7, 1994 Table 3. Inmt Data for Examule 1: Variant 1 ~~
INPUT TABLE START APPLICATION, PARAMETER QUESTION mixing-fluids Fluid phase: liquid-liquid ,liquid-solid ,gas-liquid ,gas-liquid-solid proccss-type Operation: batch, continuous residence-time Which is the residence time nearest: hours, minutes, seconds shear Importance of shear for the process (0-1): 0 is no shear needed, 1 is very high amount of shear needed flow Importance of flow for the process (0-1): 0 is no flow needed, 1 is very high amount of flow needed application Application: gas suspension , solid suspension , liquid suspension , absorption , crystallization , leaching , dissolving , slurrying , emulsification , extraction , liquid homogenisation , storage homogetlisation , chemical reaction , fernahtion heat-trpnsfer Importance of beat transfer for the process (0-1): O=no heat transfer needed,l=heat transfer very essential impossibility-of-blockage Possibility of blockage (0-1) based on materials: O=no danger of blockage, 1 =blockage possibility affects on selection cleaning-easin.ss Cleaning easiness: O=pure materials or no consideration on cleaning,l=cleaning is very important viscosity Viscosity estimation (cP):
VALUE liquid-liquid batch hours 0.2 1 storage homogsaisation
0.2 0.2 0.5
0.8
Table 4. Outuut Data for Examule 1: Variant 1 CANDIDATE
CANDIDATE NAME
IDEAL-MIXER jet-mixing-in-unkr PneUnUtiC-miXing stirred-veurl CANDIDATE
ideal mixer jet mixing in unks pnCUI7dC mixing a t i d vu& BATCH
RANK 1
3 4 2 CONTINUOUS
TOTAL CERTAINTY CRUCIAL VARIABLES CERTAINTY I I 0.92 0.95
0.9 0.92
0.9s I RES. TIME (HOURS) RES. TIME w m m ) RES. TIME (SECONDS)
SHEAR CAPABILITY
FLOW CAPABILITY
I
I
0
0.1
0
0.6 I
I I 0.9 I
0
IDEAL-MIXER jet-mixing-in-unka pneurmlic-mixing stimd-vessel CANDIDATE IDEAL-MIXER jet-mixing-in-unkr pncumtic-mixing nimd-vcwl CANDIDATE IDEAL-MIXER ,cl-mixi"g-in-unkI pncunutic-mixing StiKd-"eSCl
CANDIDATE IDEAL-MIXER jet-mixing-in-unki pneurmtic-mixing atimd-vcawl
CANDIDATE
I I
I
0
I 0
1
0.3 I
I I
mum MOTION
HEAT TRANSFER
s o L m DISPERSION
I
1
I 0 I GAS DISPERSION
m m COSTS
ENERGY CONSUMF'TION
WORKING TIME
1 02 0.2 0
I 0.8
I I
0 0.2
0.5
BLOCKAGE
CLEANDIG EASINESS
I I
LIQUID DISPERSION
msslsam
I
1
I
I
1
I
I
0
0
0
0.2
I I GAS SUSPENSION I 0 I I LIQUID HOMffiENISATlON I 0.7 1 I MIN. VISCOSITY (CERTAIN RANGE,BATCW
0.25 1 ABSORPTION
0. I
0 0.7 I sLvRRymo
0 1 CRYSTALLIZATION
1 I LEACHING
I 0
I 0
0
0 I
IDEAL-MIXER
0
jet-mixing-in-unkr
0 0 0
P"C"IlMtif-mixing stirred-vcswl
1
0 0.8 1 LIQUID SUSPENSION I
I SOLID SUSPENSION I 0 03 I D(TRACl7ON
1
0 0 I
1
EMULSIFICATION
STORAGE HOMOGENISATION
CHEMICAL REACTION
1
I
I
0
I
I 0
0.5
0 I
MAX. VISCOSITY (CERTAIN RANGE,BATCW
I MIN. VISCOSITY (POSSIBLE RANGE.BATCH)
1 MAX. VISCOSITY (POSSIBLE RANGE.BATCW
I.OOE+IJ 50 50 500000
0 0 0 0
I.OOE+IJ
0 0 1
IWO loo0 1003000
is defined as (3)
where wj = weight of importance of t h e j selection criterion proposed by the system, column 3 in Table 2. Xi, = suitability of the given mixing type i according to the criterionj, elements in Table 1. Yj = weight of importance of the j selection criterion, numerical values in column 3 in Table 3. Midealis defined in a similar way as Mi. The respective values of Xij are taken for the ideal mixer from Table 1. The values of weights of importance wj and suitabilities Xij are based on information obtained from Liquid Agitation (197519761, McDonough (1992), Nagata (19751, Oldshue (19831,UhlandGray(1986),andTatterson (1991). The importance of any criterion and/or the values of the weights of importance W j could be tuned by the user from Table 2.
1
1
I DlSSOLVING 1 0
0 I FERMENTATION 1
0 I
0.2 I 1 M m . VISCOSITY MAX. VISCOSITY urn.VISCOSITY MAX. vlscosrr~ (CERTAIN (CERTAIN (POSSIBLE (possmu RANGE.CONTNUOUS) RANGE,COWINUOUS) RANGE,CONTUWOUS) R A N G E , C O ~ O U S ) 0 I.WE+IS 0 I.WEC15 0 50 0 loo0 0 50 0 loo0 0 500000 0 Iwooo3
0.7 I
The suitability is calculated for the selection objects first for crucial parameters and then for the meaningful parameters from eq 1. The suitability of the crucial parameters is compared to the ideal mixer, which is the most suitable in any mixing duty. The form of eqs 1, 2, and 3 results from the studies of the different equations. It is used for the normalization of the suitability of the mixing type. Moreover it creates a possibility of calculating the compromise estimations based on the system knowledge and user opinion. The overall crucial parameter suitability of the approved selection candidates must exceed the crucial parameter threshold value. The threshold values result from the tuning of the system. The tuning was based on the comparison of the results obtained from the system with the proposal from the literature, opinion of experts, and existing solutions. The suitability of the meaningful parameter is calculated only for the approved candidates. The selected candidates
Ind. Eng. Chem. Res., Vol. 33, No. 7, 1994 1759 Table 5. Input Data for Example 1: Variant 2 INPUT TABLE START APPLICATION: QUESTION PARAMETER VALUE liquid-liquid Fluid phpse: liquid-liquid ,liquid-solid ,gas-liquid .gas-liquid-solid mixing-fluids tmtch Operation: batch, contiouous P--W Which is the residence time nearest: hours, minutes. 8econds residence-time hours 0 Importance of shear for the process (0-1): 0 is no shear needed. 1 is very high amount of shear needed abapr flow 0.8 Importance of flow for the process (0-1): 0 is no flow needed, 1 is very high pmount of flow needed Application: gas suspension , solid suspension , liquid suspension , absorption , crystnllizntion , leaching , dissolving , slurrying , application storage homogeisltion emulsification , extraction , liquid homogeaisation , storage homogeaisation , chemiul reaction , fenneatation 0 Importance of heat transfer for the process (0-1): O=no heat transfer necded,l=hcat transfer very essential heat-transfer 0 Possibility of blockage (0-1) based on materials: O=no danger of blockage, 1 =blockage possibility affects on selection impossibility-of-blffihge 0.3 Cleaning easiness: O=purc materials or no consideration on cleaning.1 =cleaning is very important cluning-casin~ 0.8 Viscosity estimation (cP): viscosity
Table 6. Output Data for Example 1: Variant 2 CANDIDATE
CANDIDATE NAME
IDEAL-MIXER jet-mixing-in-u~ pnrurmtic-mixing stirred-vnul CANDIDATE
idul mixer jet mixing in unki PMVMtiC mixing
IDEAL-MIXER jet-mixing-in-Unks pmunutic-mhing dmd-vessel CANDIDATE
1 1
I I
I I 1 I
LIQUDLIQUID
LIQUDSOLID
IDEAL-MIXER
I
jet-mixing-in-unkr
PMUnUtiS-midng
I I
slimed-veucl
I
CANDIDATE
FLUID MOTION
I 0 0.3 I HEAT TRANSFER
IDEAL-MIXER jet-mixing-in-unkr QMumlis~mixing
I
1
I I
0 0.15
1 GAS SUSPENSION
ABSORP~ON
SOLID SUSPENSION
I 0 0.8 I
I 0 0.3
I
1
0 0
0 0
1 EXTRACnON
1 EMULSIFICATION
1 STORAGE HOMOGENISATION
stimd-verul
CANDIDATE IDEAL-MIXER jet-mixing-in-uck Q"c"lMtiC-mi**
nimd vessel BATCH
1
0 I
RANK
TOTAL CERTAINTY
CRUCIAL VARIABLES CERTAINTY
1
I
I
2 4
0.97 0.91
1
3
0.91
commuous
RES. TIME (HOURS) I I
I
1 I GAS-LIQUID 1
0 1
I
0.97
I RES.TU&
mww
RES. T M E (SECONDS)
S H E M CAPABUITY
FLOW CAPABILITY
1
I
I
I
0 0 0.8
0 0
0.I
OW-LIQUIDsoLm I
I 0 I
0 m m L COSTS
I
I 0.9 I
ENERGY CONSUMPTION
WORKING TIME
I
I I I
0.6
I 0.2
0.8 0
0.1 0
soLm DISPERSION I
OAS DISPERSION
I
I
0 0.1 I
0 0.7 i
0 0 I
SLURRYING
CRYSTALLIZATION
LIQUID DISPERSION
0.1
B-GE
mssmnm I 0.2 I
I I I
1 LWcHrno
1 DISSOLVING
I 0 0 I
I LIQUID HOMWENISATIO N
LIQUID SUSPENSION
IDEAL-MIXER jet-mixinp-in-unkr PMUrIUlk-dnp
I 0.7 I
I
1
I
I
I
0
0
0
i
0
0
os
1 MIN. VISCOSITY (CERTAIN RANGE,BATCH)
I
I
I I
0.1
Uid-"WCl
0.2 I
nirrrd-"C.Ul CANDIDATE
CANDIDATE
IDEAL-MIXER jet-mixing-in-@& QnaUlNtiS-mhiap
0 0 0
nimd-VCS~i
0
I
(CERTAIN RANGE.BATCH) I.OOE+lS 50
50 100000
CHEMICAL REACTION
0
IWO
0 0
IWO
MAX.VISCOSITY (CERTAIN RANGE.CONTINUOUS) I.OOE+IS 10 50
IoOawM
5oowo
MAX. VISCOSITY MIN. VISCOSITY MAX.VISCOSITY MIN. VISCOSITY (CERTAIN (POSSIBLE &GE,BATCH)
0
(POSSIBLE RA~~GE.BATCH) I.OOE+IJ
are ranked on the basis of their suitability to the mixing task.
System Overview The knowledge-basedsystems traditionally have three main parts: the knowledge base, inference machine, and user interface. In this system the knowledge base is separated in MS-EXCEL spreadsheet, Nexpert Object acts as an inference machine, and MS-EXCEL spreadsheets work as a user interface. Knowledge Base. The knowledge about selection of mixing types is stored into an MS-EXCEL spreadsheet in database form. The knowledge is divided into two main P a . In the first part, there is the knowledge for the selection of different mixing types. The knowledge has the form of the table containing information on how good is a given mixing type from the point of view of the given criterion.
RANGE.CONI?NUOUS)
0.1 CLEANUW EASINESS
1
0 0 I FERMEhTATION
MLN. V I S C O S m (POSSIBLE RANGLCONTNUOUS)
MAX. VlpcOSITY (POSSIBLE
The suitabilityestimationsare crisp or fuzzy. The example of the knowledge storage is described in Table 1. The selection objects, mixing types, are in the rows, and selection parameters, criteria, are given in the columns of the spreadsheet. There are many benefits in this kind of arrangement. First, the knowledge can be tuned by changing the selection parameter values of selection objects from the knowledge spreadsheet. Second,new selectionobjectscan be updated easily into the system by opening a new row from the spreadsheet and by writing there the parameter values of a selection object. Also, deleting of selection objects can be simply done by deleting those rows from the spreadsheet where the parameter values of the selection object appear. In the second part of knowledgeabout selectionof mixing types there is knowledge about the control of the inference machine stored into an MS-EXCEL spreadsheet. An example of the inference control table is shown in Table
1760 Ind. Eng. Chem. Res., Vol. 33, No. 7, 1994 Table 7. Input Data for Example 1: Variant 3 INPUT TABLE START APPLICATION: QUESTION Fluid phase: liquid-liquid ,liquid-solid ,gas-liquid ,gas-liquid-solid Operation: batch, continuous Which is the residence time nearest hours, minutes, seconds Importance of shear for the process (0-1): 0 is no shear needed, 1 is very high amount of sheor needed Importance of flow for the process (0-1): 0 is no flow needed, 1 is very high amount of flow needed Application: gas suspension , solid suspension , liquid suapsnsion , absorption , crystallihou , leaching , dissolving emulsification , extraction , liquid homogcnisntion , storage homogslrisstion , chemical reaction , f e m h t i o n Importance of heat transfer for the process (0-1): O=no heat transfer needed,l=heat tmkr very essential Possibility of blockage (0-1) based on mnterids: O=no danger of blockage, 1 =blockage possibility affects on selection Cleaning easiness: O=pure materials or no consideration on cl&g,l=cleaning is very important Viscosity cstimntion (cP):
PARAMETER mixing-fluids pr-s-type resideace-time shear flow application
, slurrying ,
heat-transfer impossibility-of-blockage cleaning-easiness viscosity
VALUE liquid-liquid batch houn 0 0.6
storage homogenisation
0 0 0.1 0.8
Table 8. Output Data for Example 1: Variant 3 CANDIDATE
CANDIDATE NAME
IDEAL-MIXER jct-mixlng-in-unb pwumatic-mixing atid-veucl CANDIDATE
idul mixer jet mixing in umks pneumatic mixing nined v e ~ s e l BATCH
IDEAL-MIXER jet-mixing-in-unlo pneumatic-mixing
I I
I I
1
I I
I I Lipurn-sorm
1
Sid-WWl
CANDIDATE IDEAL-MIXER jet-mixing-in-unka pneumatic-mixing atid-vruol CANDIDATE IDEAL-MIXER jet-mixing-in-umks pncumtic-mixing atirred-vessel CANDIDATE IDEAL-MIXER jet-mixing-in-unkr p"r"rmtiC-mixing
stirred-vessel CANDIDATE IDEAL-MIXER jet-mixing-in-unka pneumatic-mixing stimd_vcarcl CANDIDATE
IDEAL-MIXER jet-mixing-in-unlo pnaumatic-mixing stimd-vriui
LIQUID-LIQUID 1
I I I
RAM(
TOTAL CERTAwry
1
I
2 3
0.95
4
co"uous
I 0
0.9
0.9 RES. TIME (HOURS)
I 1
GAS-LIQUID I 0 I
CRUCW VARlABLES CERTAwry I 1 0.98 1 RES. " M E
(MINUTES) I 0
0 0.8 GAS-LIQUIDSOLID
RES. T M E (SECONDS) I 0 0
0 INITIAL COSTS I
I
I I 0 I
SOLID DISPERSION
GAS DISPERSION
1
1
0 LIQUID DISPERSION I
02 02
SHEAR CAPABILITY
FLOW c A p A B n m
I
I
01
I 0.9
0.6 i ENERGY CONSUMPTION I 08
1
0.3 I HEAT TRANSFER I
1
0
0
0
0
I I GAS SUSPENSION 1 0 1 I LIQUID HOMffiENISATlON
0.25 i ABSORPTION I
0.1 i SOLID SUSPENSION 1
0.7 I
0
0 0.8 1 LIQUID SUSPENSION I
EXTRACTION
1
0
I 0
I
0
I
I
0
0.5
I
07
I
1
I
I
MIN. VISCOSITY (CERTAIN RANGE,BATCH)
MAX. VISCOSITY (CERTAIN RANGE,BATCH) I.OOECI5
MIN. VISCOSITY (POSSIBLE RANGE,BATCH)
M A X VISCOSITY
I MIN VISCOSITY (CERTAIN RANGE,COKIWUOUS)
0 02 I MAX VISCOSITY (CERTAIN RANGE,CONTINUOUS) IOOE+IS 50
FLUID MOTION
I 0.7
0 0 0 0
50 50 500000
I WORKING TIME 1 1
0
I
0.5
SLVRRYING
I CRYSTALLlZATlON
02 BLOCKAGE WSSlBlLITY I 02 I I LEACHING
1
I
I
I
0
0
0
0
0.3
0 I EMULSIFICATION
0
0
I STORAGE HOMffiENISATlON
I CHEMICAL REACTION
0 0 I
I
I
0 0 0 0
(POSSIBLE RANGE.BATCH) I.WE+lJ IOM)
loo0 1000003
2. The selection parameters are in the first column on the
left of the table, the importance of the selection parameters is stored in the second column, and in the third column there are the parameter weights. The meanings of selection parameters and parameter weights are described in the previous section. The threshold values for fluid viscosity ranges and crucial parameters can also be changed in this table. The threshold values are used to put a suitability value limit for selection objects which approved selection objects should exceed. The engineer can change the importance of the parameters and the threshold values straight from the spreadsheet. In this way it can affect inferencing without going into the "program code" level. Inference Machine. An object-oriented programming style has been utilized because the selection objects can be treated as objects in a class-and-object hierarchy. The selection objects that belong to the same class have the same attributes, and the same methods which are used to
0 0 0 0
CLEANING EASINESS
I I
I I DISSOLVING
FERMEKTATlON
I 0 I I MIN. VISCOSITY (POSSIBLE RANGE,CO"UOUS) 0
50
0 0
MAX. VISCOSITY (WSSIBLE RANGE,COKTINUOUS) I CQE+IS loo0 loo0
500000
0
IOOO
evaluate them. The selection objects are retrieved dynamically from the database to Nexpert Object as dynamic objects of one class. A class has methods to calculate suitability which dynamically linked objects inherit from aclass. The calculation methods are activated for an entire class of objects when parameter suitability is needed. The selection objects which are approved based on selection criteria are transferred to the class of feasible solutions with "pattern matching" techniques. "Pattern matching" is checking at once the attribute value, i.e., suitability of the objects of a class, to identify feasible solutions. User Interface. The user interface consists of the input and results tables. The engineer formulates the problem of selecting suitable mixing types in the input table. Such an approach enables a use of case studies instead of having interactive questions generated by the system. Thus, when input information is changed, the engineer only has to change some answers which are to be changed. The results of the selection are written in the results table by the system. The results consist of possible mixing
Ind. Eng. Chem. Res., Vol. 33, No. 7,1994 1761 Table 9. Input Data for Example 2: Variant 1 INPUT TABLE START APPLICATION: QUESTION Fluid phase: liquid-liquid ,liquid-solid ,gas-liquid ,gas-liquid-solid Operation: batch, continuous Which is the residencc time nsprsst: hours, minutas, scconds Importance of shea for the p(0-1): 0 is no shear needed, 1 is very high amount of shear needed Importance of flow for the prouss (0-1): 0 is no flow needed, 1 is very high PmOuDt of flow needed Application: gas suspension , solid suspension , liquid suspension , absorption , crystallization , leaching , dissolving emulsification , extraction , liquid homogcuisation , storage homogcuisation , chemical reaction , fermentation Importance of heat transfer for the process (0-1): O=no heat transfer needed,] =heat transfer very essential Possibility of blockage (0-1) b e d on materials: O=no danger of blockage, l=bl&ge possibility nffwts on selection Cleaning easiness: O=pun materials or no considerationon cleaning.1 =clcauing is very important Viscosity estimation (cP):
PARAMETER mixing-fluids process-typc residence-time shear flow
, slurrying ,
application heat-transfer impossibility-of-blockage cleaning-easiness viscosity
VALUE gas-liquid CONTINUOUS seconds 0.9 0.7 absorption 0 1 0.7
so
Table 10. Output Data for Example 2: Variant 1 CANDIDATE
rnw-MIxm mtic-in-line-mixer tccgipr-or-injcctor dyrumicjn-liae-mixcr
CANDIDATE
CANDIDATE NAME
-tie
i d u l mixor in line mixer
tccpip. or injector dylvmis in linc mixer
BATCH
1 3 2 4 CONTINUOUS
TOTAL CERTAINTY
I 0.83 0.83 0.81
RES. T M E (HOURS)
CRUCW VARUBLES CERTAINTY I
I I 1 RES. TIME
RES. TIME (SECONDS)
SHEAR CAPABILIly
FLOW CAPABILITY
I I I I
1
ENERGY CONSUMPnON
WORKING TIME
rnvns)
I
I
I
0
0.4
i
0 0
0
I
0
I I I I
0.3
1
CANDIDATE
LlQUrPUQUlD
LIQIJID-SOUO
GAS-LIQUID
GAS-LIQUID-SOLID
I N I T W COSTS
I
I
I
0.65
0.6
I
I
I 0.5 CLEANING EASINESS
rnw_t.mER mtic-in-lim-mixer teeqipc-or-injector dyrumic-in-liw-mixer
I 0 0.2
RANK
1
1
I
IDEAL-MIXER
I 1
I I
I
mtis-in-line-mixer
I
1 1
tecgipr-or-injector dynrmis-in-line-mixer CANDIDATE
1
0.1
I
0
1 FLUID MOTlON
0.2
I SOLID DISPERSION
0
I 0.6
GAS DISPERSION
LlOUID DISPERSION
1
1
I
1
0 0.1
0
0.I
LEACHING
DISSOLVING I
HEAT TRANSFER
OS BLOCKAGE
wssmnm IDEAL-MIXER rutis-in-line-mixer tcegipc-or-injector dyrumis-in-line-mixer CANDIDATS IDEAL-MIXER atis-in-line-mixer teegipe-or-injcotor dylumis-in-line-mixer CANDIDATE
1 1 1
I GAS SUSPENSION I 1 0
I UQUID HOMOGENISATION
IDEAL-MIXER
I
sutic-in-line-mixer kegipc-or-imjcstar
1 1
dynamic-in-line-mixer
I
CANDIDATE
MIN. VISCOSITY (CERTAIN RANGE,BATCH)
IDEAL-MIXER rutisin-line-nuxer tregipc-ar_injcstor dynamic-in-line-mixer
0 0 0 0
I 0.7
I 0.8
0 0
0
I 0
0.1
I
ABSORPTION
s o L m SUSPENSION
SLURRYING
1 1 CRYSTALLIZATION
I 0.3 0 0.3
1
1
I I
I 0.3
O
0
0
0
0
0.1 STORAGE HOMOGENlSATION I
LIQUID SUSPENSION 1
I 1 1 MAX.VISCOSrrY (CERTAIN RANGE,BATCH) l.WE+15 0 0 0
0
0
MTRACTION
EMULSIFlCATlON
I 0.7 0.3
I 0.I 0
0.5 MIN.VISCOSrrY
0
(POSSIBLE
MAX.VISCOSrrY
0
(POSSIBLE RANGE,BATCH) I.OOE+I5
0
0
0
0
0
0
RANGE,BATCH)
types in ranked order with their suitability values and the parameters affected by the selection which is considered adequate for the explanation of the reasoning. EXCEL macros can be used an update of input table based on the importance of parameters in the table of inference control knowledge. An example of an input table is described in Table 3.The results table is presented in Table 4. The total number of rules that control the information flow in mixing type selection equals 117. There are five classes, which stand for the different states of inferencing. The objects are divided into static and dynamic objects. Static objects are input parameters (e.g., viscosity) and hypotheses of the rules. Dynamic objects are selection candidates. There are 31 static and 10 dynamic objects in the system. It has taken 3 months to organize and to implement the knowledge, collected in advance. Examples. Two industrial examples are presented to illustrate the applicability of the system. The first example
I
I 0.1 0.I
0.5 0
0 CHEMICAL REACTION
0.1 FERMENTATION
1
I
0 0 0
0.8
0.5
0 0
0.3
0
MIN. VISCOSITY (CERTAIN RANGE,CONTINUOUS)
MAX.VISCOSITY
0 0 0 0
MIN. VISCOSITY MAX,VISCOSITY (CERTAIN (POSSIBLE (POSSIBLE RANGE,COKIWUOUS) RANGE,CONTlNUOUS) RANGE.CONTINUOUS) IOOEC15 0 I OOE+IS
1OOOOOO 50
5
m
0 0 0
lMxo00
IWO 2soooo
is based on the study of petroleum oil storage homogenization presented by Uhl and Gray (1986). A description of the problem is as follows: tank volumes are 80 OOO and 122 000 barrels; mixing time is about 24 h; flow of fluid is essential, but shear forces have almost no influence in storage homogenization; viscosity of crude oil is little less than that of water; homogenization of the content of the small diameter solid particles is low in crude oil. All the input information is given in Tables 3,5,and 7. The system has selected jet mixing, stirred vessel, and pneumatic mixing. The values of the suitability from the point of view of the particular criteria as well as the total applicability, “total certainty”, are presented in Tables 4, 6, and 8. A side-entering propeller in a stirred tank was used in an industrial plant. According to Nagata (1975) jet mixing is also often used in blending of liquids. According to Kurronen (1985) pneumatic mixing is also possible in such cases. The different input conditions have been simulated
1762 Ind. Eng. Chem. Res., Vol. 33, No. 7, 1994 Table 11. InDut Data for Example 2 Variant 2 INPUT TABLE START APPLICATION: QUESTION PARAMETER Fluid phrpe. liquid-liquid Jiquid-solid ,gas-liquid ,gas-liquid-solid mixing-fluib Operation: h h , continwus pm=-tYpe nsidcnce-tine which is the m i d c a w rims neuest: hwrs. minutea, seconds Importance of shear for chs pmcsg (0-1): 0 is no shear needed, 1 is very high amount of shapr needed shear Importance of flow for the pmcsg (0-1): 0 is no flow needed, 1 is very high amount of flow needed flow Application: gas surpcDsion , solid PuSpsDIion , liquid suspmsion , h r p t i o n , crystrllization , leaching , dissolving , slunying , application emulsification, extnaiocl , liquid homogonidon , storage homogcaidon , c b s m i d rsrction, fenneatation Importance of heat transfer for the pmcsg (0-1): 0-no heat transfer nceded,l=heat transfer very ssscatid heat-trsnsfcr impossibility-of-blockage Possibility of blockage (0-1) buDd on rmterids: O=no danger of blockage. 1 =blockage possibility affects on sslection Cleaning easinsac: 0-pure nutarids or no Ecasidention on clerning,l=clepning is very impo-t clerning-easiness viscosity Viscwity estimation (cP):
VALUE gas-liquid CONTINUOUS Seconds
0.5 0.3 absorption 0
0.6 0.3 50
Table 12. Output Data for Example 2 Variant 2 CANDIDATE
CANDIDATENAME
RANK
TOTALCERTAINTY
CRUCW VARIABLE3
CERTAwry IDEAL-MlXEn
idealmixer
I 3 2 4
1
1
0.86
I I I
rutis-in-line-mixer
rutif in line mixer
taoge-or-injector dylumis-in-line-mixer
ur-pipe or i+cW dynunis b I*r mixu BATCH
comous
m.TlME (Horn)
I 0
1 1
I
I 0.4
I I
I
0
0.2
I I
0 0
0
1
0
0.3
I
UQurPLIQurO
LIQUID-SOUD
GAS-LlQWUI
GAS-UQvIDsoLm
m n m corn
I I m G Y CONSUMPTION
I I I
I
1
1
I 0.1
I
I I
CANDIDATE
rnwyrxmt &tic-in-line-mixer taogipc-or-injutor dyrumic-in-line-mixer CANDIDATE
rns.u-himm rutis -in-line-mixer tcrqipe-or-injector dyMmic-in-line-mixcr CANDIDATE
IDEAL-MIXER rutif-in-line-mixer tccqipt-or-injcctor dynamic-in-line-mixer CANDIDATE
IDEAL-MIXER salic-in-line-mixer tccgipc-or-injcctor dyMmic-in-line-mixcr CANDIDATE
I
FLUD MOTION I
I I I GAS SUSPWSION I I 0 I LIQUID HOMOOENISATlON
rnuL-MIXm rutis-in-lime-mixrr trcgipc-or-injector dynamic-in-line-mixer CANDIDATE
I I
I I
0.86
0.83
RES. 7ha
mnmrras)
RES. fh(E (SECONDS) SHEAR CAPABarrY
0.65 1 0.6
I 0 0.2 I 0 HEAT TRANSFER soLm DISPERSION GAS DISPERSION LIQUID DBPERSION I 0.7 0
I 0.8 0
0
0
0.1 SOLID SUSPENSION I
1 SLURRYR4G I
I
ABSORPTION I
I
I
1
I 1 1 CRYSTALLIZATION
I 0.3
MAX. VISCOSW
MIN. V B C O S M
MAX. VISCOSITY
MIN. VISCOSM
(CERTAIN RANQE,BATCH) I .JOE+ 15 0 0 0
(eossmu
w m L E RANGE,BATCW) I .JOE+ I5
(CERTAIN RANGE,CON"UOUS) 0
0
0.1
0. I
RANGE,BATCW 0 0 0 0
0
0
0
0 0
0
(Tables 3,5, and 7) in order to study the sensitivity of the results. The parameters changed were importance of shear, importance of flow, importance of heat transfer, possiblity of blockage, and cleaning easiness. They were changed in the followingranges: 0-0.2;0.61;0-0.2; 0-0.2; 0.1-0.5,respectively. It could be seen that the obtained results are stable and there is only a small influence of the parameter variability on the "total certainty". The results of simulation with the different variants are seen in Tables 4, 6, and 8. The second example concerns the pulp and paper industry. It is described by a Finnish static mixer manufacturer (Hiirkanen, 1985). The problem consists in absorption of oxygen into an alkaline washing liquid. The input information is given in Tables 9,11,and 13. A description of a case is as follows: a continuous process is studied; the residence time is measured in seconds; in absorption shear forces are more important than fluid flow; the viscosity of washing liquid is maximum 50 cP; absorption fibers are present in the liquid. The system
1
0
0.I
LEACHING I
DISSOLVING I 0.5 0 0.1
0.5 0.3
(CERTAIN
0
0.I
0.3
MIN.VL9C€lSW
tccqtpc-or-injector dyMnUC-ln-llrm-mXeI
I 0
I 0.8
0 0
0 0
Possmmn
1 0.1
I I
I
0.5 CLEANMG EASINESS
1 0.7
STORAGE HOMOOWLFATION
WORKPIG TlME
0.5 BLOCKAGE
I
0
I I
I I
MlRACnON
0 0.I
1 i
I
LIQUID SUSPENSION
0
0.5
IDEAL-MIXER ruIIc-in-lm-mxer
I 0.6
i 0 0 EMULSmCATlON
0.3 0 0.3
I
RANGKBATCW
1
FLOW CAPABILrn
0 0 0 CHDllCAL REACTION
MAX. VISCOSW (CERTAIN RANQE.CO"UOUS) I.OOE+IS
lMxlo00 5
50 m
FgCMEKTATlON
MIN. YlSCOSM (possmu RANGE,CO"UOUS) 0 0 0 0
MAX. VISCOSITY
m.mE
RANGE,CONTINUOUS) 1.JOE+ I5
has selected a tee-pipe or injector, a static mixer, and a dynamic in-line mixer. A static miser installation has been used successfully in a pulp and paper factory in Varkausaccording to Hiirkanen (1985). The same source gives the information that sometimesan injector has been designed so well that there is no need for a static mixer. According to Oldshue (1983), the performances of static and dynamic in-line mixers are similar. Studies of the system sensitivity have been performed (Tables 10, 12,and 14). The parameters changed were importance of shear, importance of flow, possiblity of blockage, and cleaning easiness. They were changed in the following ranges: 0.5-0.9; 0.3-0.7; 0.61; 0.3-0.7, respectively. The results obtained show that the system is not too sensitive to changes of the input parameters.
Summary The presented system is intended to be used in design and marketing. It is part of a greater project aiming a t
Ind. Eng. Chem. Res.,Vol. 33, No. 7,1994
1763
Table 13. Input Data for Example 2 Variant 3 INPUT TABLE START APPLICATION: PARAMETER QUESTION mixing-fluids Fluid phnsc. liquid-liquid ,liquid-solid ,gas-liquid ,gas-liquid-solid Operation: batch, continuous procass-type residence-time Which is the residence time nearest: hours, minutes, seconds Importance of shear for the pmess (0-1): 0 is no shear needed, 1 is very high amount of shear needed shear flow Importance of flow for the procsss (0-1): 0 is no flow nwded, I is very high amount of flow needed application Application: gas suspeusion , solid suspcnsion , liquid nrspsasion , absorption , crystallization , leaching , dissolving , slurrying , emulsification , extraction , liquid homogmisation , storage homogeaisntion , chemicpl reaction , fermentation heat-transfer Importance of heat transfer for the process (0-1): O=no heat transfer needed,l =heat transfer very essential impossibility-of-blockage Possibility of blockage (0-1) based on materials: O=no danger of blockage, 1 =blockage possibility affects on sslection clssning-wines3 Cleaning easiness: O=pun materials or no considerationon cl..ning,l =cleaning is very import.nt viscosity Viscosity estimation (cP):
VALUE gM-liquid
co"uous Seconds
0.7 0.5 absorption 0
0.8 0.5 50
Table 14. Output Data for Example 2: Variant 3 CANDIDATE
IDEAL-MIXER Ntis-in-line-mixer ucgipe-or-injector dynamic-in-lim-mixer CANDIDATE
CANDIDATE NAME
RANK
TOTAL CERTAINTY
idul mixer -tic in line mixer a-pipr or injector dynamic in line mixer
i 3 2 4
0.84 0.85 0.82
BATCH
CONTINUOUS
RES.T M E (HOURS)
1 0 0.2 0
1 I 1 1
I 0
LIQUID-LIQUID
LIQUID-soLm
1
I I 0.1 0.2 HEAT TRANSFER
I
CRUCIAL VARIABLES CERTAINTY I 1 1
I RES.TIME
RES.TIME (SECONDS)
SHEAR CAPABILITY
m r r w COSTS
ENERGY CONSUMPTION
WORKING TIME
I 0.65 1
I 0.6 1
I 1
0.6
0.5
0.5 c L w N m o EAsmEss
M
W CAPABWY
(MINUTES) IDEAL-MIXER Ntiojn-line-mixer uegip-or-iajcotor dynunis-in-line-mixer CANDIDATE IDEAL-MIXER Ntio-in-line-mixer tugipr-or-injector dynamic-in-line-mixer CANDIDATE IDEAL-MIXER Ntic-in-line-mixer teegipe-or-injector dyMmisjn-line-mixcr CANDIDATE rnw-MmER Nticjn-line-mixer tugipe-or-injector dynamio-in-lim-mixer CANDIDATE IDEAL-MIXER *tic-in-line-mixer ucgipe-or-injector dynrmic-in-line-mixer CANDIDATE
IDEAL-MIXER
1 1 I FLUID MOTION 1 I I I GAS SUSPENSION I I
I 0.7 0 0 ABSORPTION
0
0
I
0.3
1
0.3
LIQUID LIQUID SUSPENSION HOMOOENISATION I I I I
0
0 ms-LiQurn
I I I
I 0.4 0 0.3 oAs-LiQum.soLm
I
I s o L m DISPERSION
I 0 0 GAS DISPERSION
1 0.8 0
I I 0
0.1
I
s o L m SUSPENSION I I 0 0
sLumYmo
EXIRACTION I 0.7
I
I
0.3
I MIN. VISCOSITY (CERTAIN RANGE.BATCH) 0
1 MAX. VISCOSITY (CERTAIN RANGE,BATCH) l.OOE+lJ 0 0 0
0.5 MIN. VISCOSITY (POSSIBLE RANGE,BATCH) 0 0 0 0
LiQum DISPERSION
I I 0 0 EMULSIFICATION I 0.1 0 0 MAX,VISCOSITY (POSSIBLE RANGE,BATCH) I.COE+IS 0 0 0
full automation of the mixing equipment design. The most important features of the system are as follows: Object-oriented programming style allows addition of new or deletion of old selection candidates. There is a great flexibility in the form of the stored knowledge. It could be easily modified by changing the values of the selection parameters. There is a possibility of inference mechanism modifications, outside the programming code, by changing parameter weights, importance of parameters, and threshold values for the viscosity and crucial parameters. There is a possibility of case studies by changing the input values in the input table. Programming results can be browsed in the results table as well as all the parameter values of the approved candidates (why function). The architecture of this program enables easy tuning and modifications of the system. Such aprogram structure is suggested for future developments of the knowledgebased systems for the equipment type selection.
BLOCKAGE
FQssIBm
I
1 I I 1 0 0.1 1 0.1 0.1 1 0 0.1 CRYSTALLIZATION LEACHING DISSOLVING 1 I I 0.3 0 0.5 0 0 0 0 0.1 0.1 STORAGE CHEMICAL REACnON FERMENTATION HOMOOENISATION 1 I I 0 0.8 0 0 0,5 0 0.3 0 0 MIN. VISCOSITY MAX. VISCOSITY MIN. VISCOSITY MAX.VISCOSITY (CERTAIN (CERTAIN (POSSIBLE (POSSIBLE RANGE.CONTINUOUS) RANGE,CONIWUOUS) RANGE,CONTINUOUS) RANGE,CONTINUOUS) 0 1.DIE+ I5 0 I.CQE+IJ 0 lMxxKl0 0 50 0 5oooo
Acknowledgment This research is financially supported by the Finnish Technology Development Center (TEKES)and Neste Oy.
Literature Cited Blass, E.; Goettert, W. Computer-assisted selection of equipment for liquid-liquid extraction. In Computer Applications in Chemical Engineering; Buesemaker, H. Th., Iedema, P. D., Eds.; Elsevier: Amsterdam, 1990;pp 111-116. Chuang, K. T.; Chen, G. X.;Rao,M. An Expert System for Selecting a Vapour-Liquid Contactor. Can. J. Chem. Eng. 1992, 70,794799. Hanratty, J. P.; Joseph, B. Decision-making in ChemicalEngineering and Expert Systems: Application of The Analytic Hierarchy Process to Reactor Selection. Comput. Chem. Eng. 1992,16,849860. Hardee, R. Third-party pump selection and evaluation software. World Pumps, 1993, Jan, 16-20. HPukijnen, I. In ZNSKO Publications 101-85; INSKO Publications: Helsinki, 1985;Chapter VIII.
1764 Ind. Eng. Chem. Res., Vol. 33,No. 7, 1994 Knoch, A.;Bottlinger,M. Expertensysteme inder VerfahrenetechnikKonfiguration von Rfihrapparakn. (Expert System in Chemical Engineering-Exampleain the Configurationof Agitators.) Chem.Zng.-Tech. 1993,65,802-809. Koiranen, T.; Kraslaweki,A.; Yang, J.; Nystrbm, L. An Expert System for Impellers Selection. Presented at the Second World Congresa on Expert System,Lisbon, Portugal, 1994a. Koiranen, T.; Kraslawski,A.; Yang, J.; Nystrtjm, L. An expert system for the preliminary design of mixing equipment. Presented at the Fifth International Symmium on Proeem Systems Engineering, Kyongju, Korea, 1994b.Kurronen. A. In ZNSKO Publications 101-85:INSKO Publications: Helaink, 1985; Chapter I. Lahdenpera, E.; Korhonen, E.; Nystrbm, L. An Expert System for the Selection of Solid-Liquid Separation Equipment. Comput. Chem. Eng. 1989,13,467-474. Lesage, G.; Chacar, R. An expert system for the design, operation and diagnosisof solid/liquidseparators. In Computer Applications in Chemical Engineering; Bussemaker, H. Th., Iedema, P. D., Eda.; Elsevier: Amsterdam, 1990;pp 93-96. Liquid Agitation by Chemineer. Parts 1-12. In Chem. Eng. from 1975,Dee 8,110-114,to 1976,Dec 6,165-170.
McDonough,J. R. Mixing for ProceeeIndustries;van Nostrand Reinhold New York, 1992. Nagata, S. Mixing: Principles and Applications; Wdey New York, 1975. Oldshue, J. Y. Fluid Mixing Technology; McGraw-Hill: New York, 1983. Tattereon, G. B. Fluid Mixingand GasDispersion in Agitated Tanks. McGraw-Hilk New York, 1991. Uhl, V. W.; Gray, J. Mixing Theory and Practice; Academic Press Inc.: New York, 1966;Vol. 1. Yang,J.; Koiranen, T.; Kraelaweki, A.; Nystram, L. Object-Oriented Knowledge Based System for Process Equipment Selection. Comput. Chem. Eng. 1993,17,1181-1190. Zimmermann, H.d. Fuzzy Set Theory and Its Applications; Kluwer Aced. Publ.: Norwell, 1992.
Received for review October 6, 1993 Revised manuscript received February 24, 1994 Accepted April 4, 1994. 0
Abstractpublishedin Advance ACSAbstracts, May 15,1994.