A new method for online flow regime monitoring in bubble column reactors

A Method for Converting Thesauri toRDF/OWLMark van Assem1,Maarten R.Menken1,Guus Schreiber1,Jan Wielemaker2,and Bob Wielinga21Vrije Universiteit Amsterdam,Department of Computer Science{mark,mrmenken,schreiber}@cs.vu.nl2University of Amsterdam,Social Science Informatics(SWI){wielemaker,wielinga}@swi.psy.uva.nlAbstract.This paper describes a method for converting existing the-sauri and related resources from their native format to RDF(S)andOWL.The method identifies four steps in the conversion process.Ineach step,decisions have to be taken with respect to the syntax or se-mantics of the resulting representation.Each step is supported througha number of guidelines.The method is illustrated through conversions oftwo large thesauri:MeSH and WordNet.1IntroductionThesauri are controlled vocabularies of terms in a particular domain with hi-erarchical,associative and equivalence relations between terms.Thesauri such as NLM’s Medical Subject Headings(MeSH)are mainly used for indexing and retrieval of articles in large databases(in the case of MeSH the MED-LINE/PubMed database containing over14million citations3).Other resources, such as the lexical database WordNet,have been used as background knowl-edge in several analysis and semantic integration tasks[2].However,their native format,often a proprietary XML,ASCII or relational schema,is not compati-ble with the Semantic Web’s standard format,RDF(S).This paper describes a method for converting thesauri to RDF/OWL and illustrates it with conversions of MeSH and WordNet.The main objective of converting existing resources to the RDF data model is that these can then be used in Semantic Web applications for annotations. Thesauri provide a hierarchically structured set of terms about which a commu-nity has reached consensus.This is precisely the type of background knowledge required in Semantic Web applications.One insight from the submissions to the Semantic Web challenge at ISWC’034was that these applications typically used simple thesauri instead of complex ontologies.Although conversions of thesauri have been performed,currently no accepted methodology exists to support these efforts.This paper presents a method that 3/entrez/4http://www-agki.tzi.de/swc/swc2003submissions.html2Van Assem,Menken,Schreiber,Wielemaker and Wielingacan serve as the starting point for such a methodology.The method and guide-lines are based on the authors’experience in converting various thesauri.This paper is organized as follows.Section2provides introductory information on thesauri and their structure.In Sect.3we describe our method and the ratio-nale behind its steps and guidelines.Sections4and5each discuss a case study in which the conversion method is applied to MeSH and WordNet,respectively. Additional guidelines that were developed during the case studies,or are more conveniently explained with a specific example,are introduced in these sections. Related research can be found in Sect.6.Finally,Sect.7offers a discussion.2Structure of ThesauriMany thesauri are historically based on the ISO2788and ANSI/NISO Z39.19 standards[3,1].The main structuring concepts are terms and three relations between terms:Broader Term(BT),Narrower Term(NT)and Related Term (RT).Preferred terms should be used for indexing,while non-preferred terms are included for use in searching.Preferred terms(also known as descriptors) are related to non-preferred terms with Use For(UF);USE is the inverse of this relation.Only preferred terms are allowed to have BT,NT and RT relations.The Scope Note(SN)relation is used to provide a definition of a term(see Fig.1).Fig.1.The basic thesaurus relations.Scope note is not shown.Two other constructs are qualifiers and node labels.Homonymous terms should be supplemented with a qualifier to distinguish them,for example“BEAMS (radiation)”and“BEAMS(structures)”.A node label is a term that is not meant for indexing,but for structuring the hierarchy,for example“KNIVES By Form”.Node labels are also used for organizing the hierarchy in eitherfields or facets.The former divides terms into areas of interest such as“injuries”and “diseases”,the latter into more abstract categories such as“living”and“non-living”[3].The standards advocate a term-based approach,in which terms are related directly to one another.In the concept-based approach[7],concepts are inter-related,while a term is only related to the concept for which it stands;i.e.a lexicalization of a concept[12].The concept-based approach may have advan-tages such as improved clarity and easier maintenance[6].A Method for Converting Thesauri to RDF/OWL3 3Method DescriptionThe method is divided into four steps:(0)a preparatory step,(1)a syntactic conversion step,(2)a semantic conversion step,and(3)a standardization step. The division of the method into four steps is an extension of previous work[15].3.1Step0:PreparationTo perform this step(and therefore also the subsequent steps)correctly,it is essential to contact the original thesaurus authors when the documentation is unclear or ambiguous.An analysis of the thesaurus contains the following:–Conceptual model(the model behind the thesaurus is used as background knowledge in creating a sanctioned conversion);–Relation between conceptual and digital model;–Relation to standards(aids in understanding the conceptual and digital model);–Identification of multilinguality issues.Although we recognize that multilinguality is an important and complicating factor in thesaurus conversion(see also[8]),it is not treated in this paper.3.2Step1:syntactic conversionIn this step the emphasis lies on the syntactic aspects of the conversion pro-cess from the source representation to RDF(S).Typical source representations are(1)a proprietary text format,(2)a relational database and(3)an XML representation.This step can be further divided into two substeps.Step1a:structure-preserving translation.In Step1a,a structure-preserving translation between the source format and RDF format is performed, meaning that the translation should reflect the source structure as closely as pos-sible.The translation should be complete,meaning that all semantically relevant elements in the source are translated into RDF.Guideline1:Use a basic set of RDF(S)constructs for the structure-preserving translation.Only use constructs for defining classes,sub-classes,properties(with domains and ranges),human-readable rdfs:label s for class and property names,and XML datatypes.These are the basic build-ing blocks for defining an RDF representation of the conceptual model.The remaining RDF(S)and OWL constructs are used in Step2for a semantically oriented conversion.However,one might argue that the application of some constructs(e.g.domains and ranges)also belongs to semantic conversion.4Van Assem,Menken,Schreiber,Wielemaker and WielingaGuideline2:Use XML support for datatyping.Simple built-in XML Schema datatypes such as xsd:date and xsd:integer are useful to supply schemas with information on property ing user-defined XML Schema datatypes is still problematic5;hopefully this problem will be solved in the near future.Guideline3:Preserve original naming as much as possible.Preserving the original naming of entities results in more clear and traceable conversions.Prefix duplicate property names with the name of the source entity to make them unique.The meaning of a class or property can be explicated by adding an rdfs:comment,preferably containing a definition from the orig-inal documentation.If documentation is available online,rdfs:seeAlso or rdfs:isDefinedBy statements can be used to link to the original documen-tation and/or definition.Guideline4:Translate relations of arity three or more into structures with blank nodes.Relations of arity three or more cannot be translated directly into RDF properties.If the relation’s arguments are independent of each other,a structure can be used consisting of a property(with the same name as the original relation)linking the source entity to a blank node (representing the relation),and the relation’s arguments linked to the blank node with an additional property per argument(see examples in Sect.4).Guideline5:Do not translate semantically irrelevant ordering infor-mation.Source representations often contain sequential information,e.g.ordering of a list of terms.These may be irrelevant from a semantic point of view,in which case they can be left out of the conversion.Guideline6:Avoid redundant information.Redundant information creates representations which are less clear and harder to maintain.An example on how to avoid this:if the Unique Identifier(UI)of a resource is recorded in the rdf:ID,then do not include a property that also records the UI. Guideline7:Avoid interpretation.Interpretations of the meaning of infor-mation in the original source(i.e.,meaning that cannot be traced back to the original source or documentation)should be approached with caution,as wrong interpretations result in inconsistent and/or inaccurate conversions.The approach of this method is to postpone interpretation(see Step2b).Instead of developing a new schema(i.e.,thesaurus metamodel),one can also use an existing thesaurus schema,such as the SKOS(see Sect.3.4),which already defines“Concept”,“broader”,etc.This may be a simpler approach than to first develop a new schema and later map this onto the SKOS.However,this is only a valid approach if the metamodel of the source and of SKOS match. 5/2001/sw/WebOnt/webont-issues.html#I4.3-Structured-DatatypesA Method for Converting Thesauri to RDF/OWL5 A drawback is that the naming of the original metamodel is lost(e.g.“BT”instead of“broader”).For thesauri with a(slightly)different metamodel,it is recommended to develop a schema from scratch,so as not to lose the original semantics,and map this schema onto SKOS in Step3.Step1b:explication of syntax.Step1b concerns the explication of informa-tion that is implicit in the source format,but intended by the conceptual model. The same set of RDF(S)constructs is used as in Step1a.For example,the AAT thesaurus[11]uses node labels(called“Guide Terms”in AAT),but in the AAT source data these are only distinguished from normal terms by enclosing the term name in angle brackets(e.g.<KNIVES by Form>).This information can be made explicit by creating a class GuideTerm,which is an rdfs:subClassOf the class AATTerm,and assigning this class to all terms with angle brackets. Other examples are described in Sects.4and5.3.3Step2:Semantic ConversionIn this step the class and property definitions are augmented with additional RDFS and OWL constraints.Its two substeps are aimed at explication(Step 2a)and interpretation(Step2b).After completion of Step2a the thesaurus is ready for publication on the Web as an“as-is”RDF/OWL representation. Step2a:explication of semantics.This step is similar to Step1b,but now more expressive RDFS and OWL constructs may be used.For example,a broaderTerm property can be defined as an owl:TransitiveProperty and a relatedTerm property as an owl:SymmetricProperty.A technique that is used in this step is to define certain properties as special-izations of predefined RDFS properties,e.g.rdfs:label and rdfs:comment.For example,if a property nameOf is clearly intended to denote a human-readable label for a resource,it makes sense to define this property as a subproperty of rdfs:label.RDFS-aware tools will now be able to interpret nameOf in the intended way.Step2b:interpretations.In Step2b specific interpretations are introduced that are strictly speaking not sanctioned by the original model or documen-tation.A common motivation is some application-specific requirement,e.g.an application wants to treat a broaderTerm hierarchy as a class hierarchy.This can be stated as follows:broaderTerm rdfs:subPropertyOf rdfs:subClassOf. Semantic Web applications using thesauri will often want to do this,even if not all hierarchical links satisfy the subclass criteria.This introduces the notion of metamodeling.It is not surprising that the schema of a thesaurus is typically a metamodel:its instances are categories for describing some domain of interest. Guideline8:Consider treating the thesaurus schema as a metamodel.The instances of a thesaurus schema are often general terms or concepts,that oc-cur as classes in other places.RDFS allows one to treat instances as a classes:6Van Assem,Menken,Schreiber,Wielemaker and Wielingasimply add the statement that the class of those instances is a subclass of rdfs:Class.For example,an instance i is of class C;class C is declared to be an rdfs:subClassOf rdfs:Class.Because instance i is now also an instance of rdfs:Class,it can be treated as a class.The above example of treating broader term as a subclass relation is similar in nature.A schema which uses these constructions is outside the scope of OWL DL.Application developers will have to make their own expressivity vs.tractabil-ity trade-offhere.The output of this step should be used in applications as a specific interpretation of the thesaurus,not as a standard conversion.3.4Step3:StandardizationSeveral proposals exist for a standard schema for thesauri.6Such a schema may enable the development of infrastructure that can interpret and interchange thesaurus data.Therefore,it may be useful to map a thesaurus onto a stan-dard schema.This optional step can be made both after Step2a(the result may be published on the web)and Step2b(the result may only be used in an application-specific context).Unfortunately,a standard schema has not yet been agreed upon.As illustration,the schema of the W3C Semantic Web Advanced Development for Europe(SWAD-E)project is mapped to MeSH in Sect.4.7 The so-called“SKOS”schema of SWAD-E is concept-based,with class Concept and relations narrower,broader and related between Concepts.A Concept can have a prefLabel(preferred term)and altLabel s(non-preferred terms).Also provided is a TopConcept class,which can be used to arrange a hierarchy under special concepts(such asfields and facets,see Sect.2).TopCon-cept is a subclass of Concept.Note that because SKOS is concept-based,it may be problematic to map term-based thesauri.4Case one:MeSHThis section describes how the method has been applied to MeSH(version2004 8).The main source consists of two XMLfiles:one containing so-called descrip-tors(228MB),and one containing qualifiers(449Kb).Each has an associated DTD.Afile describing additional information on descriptors was not converted. The conversion program(written in XSLT)plus links to the original source and outputfiles of each step can be found at http://thesauri.cs.vu.nl/.The conversion took two persons approximately three weeks to complete.6/SWAD/thes links.html7/2001/sw/Europe/reports/thes/1.0/guide/8/mesh/filelist.htmlA Method for Converting Thesauri to RDF/OWL74.1Analysis of MeSHThe conceptual model of MeSH is centered around Descriptors,which contain Concepts[9].In turn,Concepts consist of a set of Terms.Exactly one Concept is the preferred Concept of a Descriptor,and exactly one Term is the preferred Term of a Concept.Each Descriptor can have Qualifiers,which are used to indicate aspects of a particular Descriptor,e.g.“ABDOMEN”has the Qualifiers “pathology”and“abnormalities”.Descriptors are related in a polyhierarchy, and are meant to represent broader/narrower document retrieval sets(i.e.,not a subclass relation).Each Descriptor belongs to one(or more)offifteen Categories, such as“Anatomy”and“Diseases”[10].The Concepts contained within one Descriptor are also hierarchically related to each other.This model is inconsistent with the ISO and ANSI standards,for several reasons.Firstly,the model is concept-based.Secondly,Descriptors contain a set of Concepts,while in the standards a Descriptor is simply a preferred term. Thirdly,Qualifiers are not used to disambiguate homonyms.4.2Converting MeSHStep1a:structure-preserving translation.In the XML version of MeSH, Descriptors,Concepts,Terms and Qualifiers each have a Unique Identifier(UI). Each Descriptor also has a TreeNumber(see[1]).This is used to indicate a position in a polyhierarchical structure(a Descriptor can have more than one TreeNumber),but this is implicit only.Relations between XML elements are made by referring to the UI of the relation’s target(e.g.<SeeRelatedDescriptor> contains the UI of another Descriptor).In Step1a,this is converted into in-stances of the property hasRelatedDescriptor.The explication of TreeNum-bers is postponed until Step1b.Most decisions in Step1a concern which XML elements should be translated into classes,and which into properties.The choice to create classes for Descrip-tor,Concept and Term are clear-cut:these are complex,interrelated structures.A so-called<EntryCombination>relates a Descriptor-Qualifier pair to another Descriptor-Qualifier pair.Following Guideline4,two blank nodes are created (each representing one pair)and related to an instance of the class EntryCombi-nation.As already mentioned,relations between elements in XML MeSH are made by referring to the UI of the target.However,each such relation also in-cludes the name of the target.As this is redundant information,the name can be safely disregarded.Guideline9:Give preference to the relation-as-arc approach over the relation-as-node approach.In the relation-as-arc approach,relations are modeled as arcs between entities(RDF uses“properties”to model arcs).In the relation-as-node approach,a node represents the relation,with one arc relating the source entity to the relation node,and one arc relating the relation node to the destination entity[7].The relation-as-arc approach is more natural to the RDF model,and also allows for definition of property semantics(symmetry,inverseness,etc.)in OWL.8Van Assem,Menken,Schreiber,Wielemaker and WielingaIt is not always possible to follow Guideline9, e.g.in the case of MeSH <ConceptRelation>.Although the straightforward choice is to create a prop-erty that relates two Concepts,this is not possible as each ConceptRelation has an additional attribute(see Guideline4).Again,a blank node is used(the relation-as-node approach).Guideline10:Create proxy classes for references to external resources if they are not available in RDF.Each Concept has an associated Seman-ticType,which originates in the UMLS Semantic Network9.This external resource is not available in RDF,but might be converted in the near fu-ture.In MeSH,only the UI and name of the SemanticType is recorded.One could either use a datatype property to relate the UI to a Concept(again, the redundant name is ignored),or create SemanticType instances(empty proxies for the actual types).We have opted for the latter,as this simplifies future integration with UMLS.In this scenario,either new properties can be added to the proxies,or the existing proxies can be declared owl:sameAs to SemanticType instances of a converted UMLS.Guideline11:Only create rdf:ID s based on identifiers in the original source.A practical problem in the syntactical translation is what value to assign the rdf:ID attribute.If the original source does not provide a unique identifier for an entity,one should translate it into blank nodes,as opposed to generating new identifiers.A related point is that if the UI is recorded using rdf:ID,additional properties to record an entity’s UI would introduce redundancy,and therefore shouldn’t be used.Guideline12:Use the simplest solution that preserves the intended se-mantics.In XML MeSH,only one Term linked to a Concept is the preferred term.Some terms are permutations of the term name(indicated with the attribute isPermutedTermYN),but unfortunately have the same UI as the Term from which they are generated.A separate instance cannot be created for this permuted term,as this would introduce a duplicate rdf:ID.Two obvious solutions remain:create a blank node or relate the permuted term with a datatype property permutedTerm to Term.In thefirst solution,one would also need to relate the node to its non-permuted parent,and copy all information present in the parent term to the permuted term node(thus introducing redundancy).The second solution is simpler and preserves the intended semantics.Step1b:explication of syntax.In Step1b,three explications are made. Firstly,the TreeNumbers are used to create a hierarchy of Descriptors with a[Descriptor]subTreeOf[Descriptor]property.Secondly,the TreeNum-ber starts with a capital letter which stands for one offifteen Categories.The class Category and property[Descriptor]inCategory[Category]are intro-duced to relate Descriptors to their Catogories.Thirdly,the ConceptRelations 9/pubs/factsheets/umlssemn.htmlA Method for Converting Thesauri to RDF/OWL9 are translated into three properties,brd,nrw and rel,thus converting from a relation-as-node to a relation-as-arc approach(see Guidelines12and9).This requires two observations:(a)the values NRW,BRD and REL of the attribute relationName correspond to narrower,broader and related Concepts;and(b) the relationAttribute is not used in the actual XML,and can be removed. Without the removal of the relationAttribute,the arity of the relation would have prevented us from using object properties.Some elements are not explicated,although they are clear candidates.These are XML elements which contain text,but also implicit information that can be used to link instances.For example,a Descriptor’s<RelatedRegistryNumber> contains the ID of another Descriptor,but also other textual information.Split-ting this information into two or more properties changes the original semantics, so we have chosen to create a datatype property for this element and copy the text as a literal value.Step2a:explication of semantics.In Step2a,the following statements are added(a selection):–The properties brd and nrw are each other’s inverse,and are both transitive, while rel is symmetric;–A Concept’s scopeNote is an rdfs:subPropertyOf the property rdfs:comment;–All properties describing a resource’s name(e.g.descriptorName)are de-clared an rdfs:subPropertyOf the property rdfs:label;–Each of these name properties is also an owl:InverseFunctionalProperty, as the names are unique in the XMLfile.Note that this may not hold for future versions of MeSH;–All properties recording a date are an owl:FunctionalProperty;–The XML DTD defines that some elements occur either zero or once in the data.The corresponding RDF properties can also be declared functional;–As a Term belongs to exactly one Concept,and a Concept to exactly one De-scriptor,[Concept]hasTerm[Term]as well as[Descriptor]hasConcept [Concept]is an owl:InverseFunctionalProperty;Unfortunately,the relation represented by class EntryCombination cannot be supplied with additional semantics,e.g.that it is an owl:SymmetricProperty (see Guideline9).Step2b:interpretations.In Step2b,the following interpretations are made, following Guideline8.Note that these are examples,as we have no specific application in mind.–brd is an rdfs:subPropertyOf rdfs:subClassOf;–Descriptor and Concept are declared rdfs:subClassOf rdfs:Class.10Van Assem,Menken,Schreiber,Wielemaker and WielingaStep3:standardization.In Step3,a mapping is created between the MeSH schema and the SKOS schema.The following constructs can be mapped(using rdfs:subPropertyOf and rdfs:subClassOf):–mesh:subTreeOf onto skos:broader;–mesh:Descriptor onto skos:Concept;–mesh:hasRelatedDescriptor onto skos:related;–mesh:descriptorName onto skos:prefLabel.There is considerable mismatch between the schemas.Descriptors are the cen-tral concepts between which hierarchical relations exist,but it is unclear how MeSH Concepts and Terms can be dealt with.SKOS defines datatype proper-ties with which terms can be recorded as labels of Concepts,but this cannot be mapped meaningfully onto MeSH’Concept and Term classes.For example, mesh:conceptName cannot be mapped onto skos:prefLabel,as the former’s do-main is mesh:Concept,while the latter’s domain is skos:Concept(skos:Concept is already mapped onto mesh:Descriptor).Furthermore,the mesh:Category can-not be mapped onto skos:TopCategory,because TopCategory is a subclass of skos:Concept,while mesh:Category is not a subclass of mesh:Descriptor.5Case two:WordNetThis section describes how the method has been applied to WordNet release 2.0.The original source consists of18Prologfiles(23MB in total).The con-version programs(written in Prolog)plus links to the original source as well as the outputfiles of each step can be found at http://thesauri.cs.vu.nl/.The conversion took two persons approximately three weeks to complete.Note that Step3for WordNet is not discussed here for reasons of space,but is available at the forementioned website.5.1Analysis of WordNetWordNet[2]is a concept-based thesaurus for the English language.The concepts are called“synsets”which have their own identifier.Each synset is associated with a set of lexical representations,i.e.its set of synonyms.The synset concept is divided into four categories,i.e.nouns,verbs,adverbs and adjectives.Most WordNet relations are defined between synsets.Example relations are hyponymy and meronymy.There have been a number of translations of WordNet to RDF and OWL formats.Dan Brickley10translated the noun/hyponym hierarchy directly into RDFS classes and subclasses.This is different from the method we propose, because it does not preserve the original source structure.Decker and Melnik11 have created a partial RDF representation,which does preserve the original 10/Archives/Public/www-rdf-interest/1999Dec/0002.html11/library/A Method for Converting Thesauri to RDF/OWL11 structure.The conversion of the KID Group at the University of Neuchatel12 constitutes an extension of representation both in scope and in description of semantics(by adding OWL axioms).We follow mainly this latter conversion and relate it to the steps in our conversion method.In the process we changed and extended the WordNet schema slightly(and thus also the resulting conversion).5.2Converting WordNetStep1a:structure-preserving translation.In this step the baseline classes and properties are created to map the source representation as precisely as pos-sible to an RDF representation:–Classes:SynSet,Noun,Verb,Adverb,Adjective(subclasses of SynSet), AdjectiveSatellite(subclass of Adjective);–Properties:wordForm,glossaryEntry,hyponymOf,entails,similarTo, memberMeronymOf,substanceMeronymOf,partMeronymOf,derivation, causedBy,verbGroup,attribute,antonymOf,seeAlso,participleOf, pertainsTo.Note that the original WordNet naming is not very informative(e.g.“s”repre-sents synset).For readability,here we use the rdfs:label s that have been added in the RDF version.All properties except for the last four have a synset as their domain.The range of these properties is also a synset,except for wordForm and glossaryEntry.Some properties have a subclass of SynSet as their domain and/or range,e.g.entails holds between Verbs.The main decision that needs to be taken in this step concerns the following two interrelated representational issues:1.Each synset is associated with a set of synonym“words”.For example,the synset100002560has two associated synonyms,namely nonentity and nothing.Decker and Melnik represent these labels by defining the(multi-valued)property wordForm with a literal value as its range(i.e.as an OWL datatype property).The Neuchatel approach is to define a word as a class in its own right(WordObject).The main disadvantage of this is that one needs to introduce an identifier for each WordObject as it does not exist in the source representation,and words are not unique(homonymy).2.The last four properties in the list above(antonymOf,etc.)do not representrelations between synsets but instead between particular words in a synset.This also provides the rationale for the introduction of the class WordObject in the Neuchatel representation:antonymOf can now simply defined as a property between WordObjects.We prefer to represents words as literal values,thus avoiding the identifier prob-lem(see Guideline11).For handling properties like antonymOf we defined a helper class SynSetWord with properties linking it to a synset and a word.For each subclass of SynSet,an equivalent subclass of SynSetWord is introduced(e.g. SynSetVerb).A sample representation of an antonym looks like this:12http://taurus.unine.ch/GroupHome/knowler/wordnet.html。

Quantum Hall effectFrom Wikipedia, the free encyclopediaThe quantum Hall effect (or integer quantum Hall effect) is a quantum-mechanical version of the Hall effect, observed in two-dimensional electron systems subjected to low temperatures and strong magnetic fields, in which the Hall conductance G undergoes certain quantum Hall transitions to take on the quantized valueswhere is the channel current, is the Hall voltage, e is the elementary charge and h is Planck's constant. The prefactor ν is known as the "filling factor", and can take on either integer (ν = 1, 2, 3, .. ) or fractional (ν = 1/3, 2/5, 3/7, 2/3, 3/5, 1/5, 2/9, 3/13, 5/2, 12/5 ...) values. The quantum Hall effect is referred to as the integer or fractional quantum Hall effect depending on whether ν is an integer or fraction respectively. The integer quantum Hall effect is very well understood, and can be simply explained in terms of single-particle orbitals of an electron in a magnetic field (see Landau quantization). The fractional quantum Hall effect is more complicated, as its existence relies fundamentally on electron–electron interactions. It is also very well understood as an integer quantum Hall effect, not of electrons but of charge-flux composites known as composite fermions. In 1988, it was proposed that there was quantum Hall effect without Landau levels. This quantum Hall effect is referred to as the quantum anomalous Hall (QAH) effect. There is also a new concept of the quantum spin Hall effect which is an analogue of the quantum Hall effect, where spin currents flow instead of charge currents.[1]Contents◾1 Applications◾2 History◾3 Integer quantum Hall effect – Landau levels◾4 Mathematics◾5 See also◾6 References◾7 Further readingApplicationsThe quantization of the Hall conductance has the important property of being incredibly precise. Actual measurements of the Hall conductance have been found to be integer or fractional multiples of e2/h to nearly one part in a billion. This phenomenon, referred to as "exact quantization", has been shown to be a subtle manifestation of the principle of gauge invariance.[2] It has allowed for the definition of a new practical standard for electrical resistance, based on the resistance quantum givenby the von Klitzing constant R K = h/e2 = 25812.807557(18) Ω.[3] This is named after Klaus von Klitzing, the discoverer of exact quantization. Since 1990, a fixed conventional value R K-90 is used in resistance calibrations worldwide.[4] The quantum Hall effect also provides an extremely precise independent determination of the fine structure constant, a quantity of fundamental importance in quantum electrodynamics.HistoryThe integer quantization of the Hall conductance was originally predicted by Ando, Matsumoto, and Uemura in 1975, on the basis of an approximate calculation which they themselves did not believe to be true. Several workers subsequently observed the effect in experiments carried out on the inversion layer of MOSFETs. It was only in 1980 that Klaus von Klitzing, working at the high magnetic field laboratory in Grenoble with silicon-based samples developed by Michael Pepper and Gerhard Dorda, made the unexpected discovery that the Hall conductivity was exactly quantized. For this finding, von Klitzing was awarded the 1985 Nobel Prize in Physics. The link between exact quantization and gauge invariance was subsequently found by Robert Laughlin. Most integer quantum Hall experiments are now performed on gallium arsenide heterostructures, although many other semiconductor materials can be used. In 2007, the integer quantum Hall effect was reported in graphene at temperatures as high as room temperature,[5] and in the oxide ZnO-Mg x Zn1-x O.[6] Integer quantum Hall effect – Landau levelsIn two dimensions, when classical electrons are subjected to a magnetic field they follow circular cyclotron orbits. When the system is treated quantum mechanically, these orbits are quantized. The energy levels of these quantized orbitals take on discrete values: , where ωc = eB/m is the cyclotron frequency. These orbitals are known as Landau levels, and at weak magnetic fields, their existence gives rise to many interesting "quantum oscillations" such as the Shubnikov–de Haas oscillations and the de Haas–van Alphen effect (which is often used to map the Fermi surface of metals). For strong magnetic fields, each Landau level is highly degenerate (i.e. there are many single particle states which have the same energy E n). Specifically, for a sample of area A, inmagnetic field B, the degeneracy of each Landau level is (where g s represents a factor of 2 for spin degeneracy, and is the magnetic flux quantum). ForHofstadter's butterfly sufficiently strong B-fields, each Landau level may have so many states that all of the free electrons in the system sit in only a few Landau levels; it is in this regime where one observes the quantum Hall effect.MathematicsThe integers that appear in the Hall effect are examples oftopological quantum numbers. They are known inmathematics as the first Chern numbers and are closelyrelated to Berry's phase. A striking model of much interest inthis context is the Azbel-Harper-Hofstadter model whosequantum phase diagram is the Hofstadter butterfly shown inthe figure. The vertical axis is the strength of the magneticfield and the horizontal axis is the chemical potential, whichfixes the electron density. The colors represent the integer Hall conductances. Warm colors represent positive integersand cold colors negative integers. The phase diagram isfractal and has structure on all scales. In the figure there is an obvious self-similarity.Concerning physical mechanisms, impurities and/or particular states (e.g., edge currents) are important for both the 'integer' and 'fractional' effects. In addition, Coulomb interaction is also essential in the fractional quantum Hall effect. The observed strong similarity between integer and fractional quantum Hall effects is explained by the tendency of electrons to form bound states with an even number of magnetic flux quanta, called composite fermions .See also◾quantum Hall transitions◾Fractional quantum Hall effect◾Composite fermions◾Hall effect◾Hall probe◾Graphene◾Quantum spin Hall effect◾Coulomb potential between two current loops embedded in a magnetic fieldReferences1.^ Ezawa, Zyun F. (2013). Quantum Hall Effects: Recent Theoretical and Experimental Developments (3rd ed.). World Scientific. ISBN 978-981-4360-75-3.2.^ Laughlin, R. (1981). "Quantized Hall conductivity in twodimensions" (/abstract/PRB/v23/i10/p5632_1). Physical Review B 23 (10): 5632–5633. Bibcode:1981PhRvB..23.5632L (/abs/1981PhRvB..23.5632L).doi:10.1103/PhysRevB.23.5632 (/10.1103%2FPhysRevB.23.5632). Retrieved 8 May 2012.3.^ Tzalenchuk, Alexander; Lara-Avila, Samuel; Kalaboukhov, Alexei; Paolillo, Sara; Syväjärvi, Mikael;Yakimova, Rositza; Kazakova, Olga; Janssen, T. J. B. M.; Fal'ko, Vladimir; Kubatkin, Sergey (2010)."Towards a quantum resistance standard based on epitaxialgraphene" (/nnano/journal/v5/n3/abs/nnano.2009.474.html). NatureNanotechnology5 (3): 186–189. arXiv:0909.1220 (https:///abs/0909.1220).Bibcode:2010NatNa...5..186T (/abs/2010NatNa...5..186T).doi:10.1038/nnano.2009.474 (/10.1038%2Fnnano.2009.474). PMID 20081845(https:///pubmed/20081845).4.^ "conventional value of von Klitzing constant" (/cgi-bin/cuu/Value?rk90). NIST.5.^ Novoselov, K. S.; Jiang, Z.; Zhang, Y.; Morozov, S. V.; Stormer, H. L.; Zeitler, U.; Maan, J. C.;Boebinger, G. S.; Kim, P.; Geim, A. K. (2007). "Room-Temperature Quantum Hall Effect in Graphene".Science315 (5817): 1379. arXiv:cond-mat/0702408 (https:///abs/cond-mat/0702408).Bibcode:2007Sci...315.1379N (/abs/2007Sci...315.1379N).doi:10.1126/science.1137201 (/10.1126%2Fscience.1137201). PMID 17303717(https:///pubmed/17303717).6.^ Tsukazaki, A.; Ohtomo, A.; Kita, T.; Ohno, Y.; Ohno, H.; Kawasaki, M. (2007). "Quantum Hall Effectin Polar Oxide Heterostructures". Science315 (5817): 1388–91. Bibcode:2007Sci...315.1388T(/abs/2007Sci...315.1388T). doi:10.1126/science.1137430(/10.1126%2Fscience.1137430). PMID 17255474(https:///pubmed/17255474).Further reading◾Ando, Tsuneya; Matsumoto, Yukio; Uemura, Yasutada (1975). "Theory of Hall Effect in a Two-Dimensional Electron System". J. Phys. Soc. Jpn.39 (2): 279–288.Bibcode:1975JPSJ...39..279A (/abs/1975JPSJ...39..279A).doi:10.1143/JPSJ.39.279 (/10.1143%2FJPSJ.39.279).◾Klitzing, K.; Dorda, G.; Pepper, M. (1980). "New Method for High-Accuracy Determination of the Fine-Structure Constant Based on Quantized Hall Resistance". Phys. Rev. Lett.45 (6): 494–497. Bibcode:1980PhRvL..45..494K(/abs/1980PhRvL..45..494K). doi:10.1103/PhysRevLett.45.494(/10.1103%2FPhysRevLett.45.494).◾Laughlin, R. B. (1981). "Quantized Hall conductivity in two dimensions". Phys. Rev. B.23(10): 5632–5633. Bibcode:1981PhRvB..23.5632L(/abs/1981PhRvB..23.5632L). doi:10.1103/PhysRevB.23.5632(/10.1103%2FPhysRevB.23.5632).◾Yennie, D. R. (1987). "Integral quantum Hall effect for nonspecialists". Rev. Mod. Phys.59(3): 781–824. Bibcode:1987RvMP...59..781Y(/abs/1987RvMP...59..781Y). doi:10.1103/RevModPhys.59.781(/10.1103%2FRevModPhys.59.781).◾Hsieh, D.; Qian, D.; Wray, L.; Xia, Y.; Hor, Y. S.; Cava, R. J.; Hasan, M. Z. (2008). "A topological Dirac insulator in a quantum spin Hall phase". Nature452 (7190): 970–974.arXiv:0902.1356 (https:///abs/0902.1356). Bibcode:2008Natur.452..970H(/abs/2008Natur.452..970H). doi:10.1038/nature06843(/10.1038%2Fnature06843). PMID 18432240(https:///pubmed/18432240).◾25 years of Quantum Hall Effect, K. von Klitzing, Poincaré Seminar (Paris-2004). Postscript (http://parthe.lpthe.jussieu.fr/poincare/textes/novembre2004.html). Pdf(/PhysicsHorizon/25yearsQHE-lecture.pdf).◾Magnet Lab Press Release Quantum Hall Effect Observed at Room Temperature (/mediacenter/news/pressreleases/2007february15.html)◾Avron, Joseph E.; Osadchy, Daniel, Seiler, Ruedi (2003). "A Topological Look at the Quantum Hall Effect" (/resource/1/phtoad/v56/i8/p38_s1).Physics Today56 (8): 38. Bibcode:2003PhT....56h..38A(/abs/2003PhT....56h..38A). doi:10.1063/1.1611351(/10.1063%2F1.1611351). Retrieved 8 May 2012.◾Zyun F. Ezawa: Quantum Hall Effects - Field Theoretical Approach and Related Topics.World Scientific, Singapore 2008, ISBN 978-981-270-032-2◾Sankar D. Sarma, Aron Pinczuk: Perspectives in Quantum Hall Effects. Wiley-VCH, Weinheim 2004, ISBN 978-0-471-11216-7◾Baumgartner, A.; Ihn, T.; Ensslin, K.; Maranowski, K.; Gossard, A. (2007). "Quantum Hall effect transition in scanning gate experiments". Physical Review B76 (8).Bibcode:2007PhRvB..76h5316B (/abs/2007PhRvB..76h5316B).doi:10.1103/PhysRevB.76.085316 (/10.1103%2FPhysRevB.76.085316).◾E. I. Rashba and V. B. Timofeev, Quantum Hall Effect, Sov. Phys. - Semiconductors v. 20, pp.617-647 (1986).Retrieved from "/w/index.php?title=Quantum_Hall_effect&oldid=605304404"Categories: Hall effect Condensed matter physics Quantum electronics SpintronicsQuantum phases Mesoscopic physics◾This page was last modified on 22 April 2014 at 14:51.◾Text is available under the Creative Commons Attribution-ShareAlike License; additional terms may apply. By using this site, you agree to the Terms of Use and Privacy Policy.Wikipedia® is a registered trademark of the Wikimedia Foundation, Inc., a non-profitorganization.。

Scientia Iranica B(2011)18(4),923–929Sharif University of TechnologyScientia IranicaTransactions B:Mechanical EngineeringExperimental investigation of air–water,two-phase flow regimes in vertical mini pipeP.Hanafizadeh,M.H.Saidi∗,A.Nouri Gheimasi,S.GhanbarzadehSchool of Mechanical Engineering,Sharif University of Technology,Tehran,P.O.Box11155-9567,IranReceived8November2010;revised28April2011;accepted12June2011KEYWORDSMini pipes;Two-phase flow; Flow pattern; Visualization; Flow pattern map.Abstract In this study,the flow patterns of air–water,two-phase flows have been investigated experimentally in a vertical mini pipe.The flow regimes were observed by a high speed video recorder in pipes with diameters of2,3and4mm and length27,31and25cm,respectively.The comprehensive visualization of air–water,two-phase flow in a vertical mini pipe has been performed to realize the physics of such a two-phase flow.Different flow patterns of air–water flow were observed simultaneously in the mini pipe at different values of air and water flow rates.Consequently,the flow pattern map was proposed for flow in the mini-pipe,in terms of superficial velocities of liquid and gas phases.The flow pattern maps are compared with those of other researchers in the existing literature,showing reasonable agreement.©2011Sharif University of Technology.Production and hosting by Elsevier B.V.All rights reserved.1.IntroductionGas–liquid,two-phase flow in micro structures has played an important role in several industrial and medical applications, such as micro heat exchangers,lab-on-chips,bio-MEMS and micro cooling electronics.Physical perception of micro flows is critical in order to optimize and develop the design of such devices.Two-phase flows at mini and micro scale have recently attracted the attention of scientists as a result of its wide usage in advanced science and technology,namely Micro-Electro-Mechanical Systems(MEMS),chemical engineering, bioengineering,medical devises,micro cooling systems,micro structures in computers,etc.The literature survey on this issue has been categorized into adiabatic and phase change work, which has been summarized in this paper.∗Corresponding author.E-mail address:saman@(M.H.Saidi).1.1.Adiabatic worksThe works of Suo and Griffith[1]were among the first studies concentrating on flow patterns in microchannels.They detected three different flow patterns,namely,bubbly/slug, slug and annular flow,in their studies,using channels with widths in the range of0.514–0.795mm.Sadatomi et al.[2] proposed flow regime maps in vertical rectangular channels and indicated that channel geometries have little influence in noncircular channels with large hydraulic diameters greater than10mm.Xu et al.[3]investigated concurrent vertical two-phase flow in a vertical rectangular channel with a narrow gap, experimentally.They reported that with a decrease in channel gap,the transition from one flow regime to another occurs at smaller gas flow rates.They developed a new criterion to predict transition from annular flow,as well.Hestroni et al.[4] performed experiments for air–water and steam–water flow in parallel triangular micro-channels,developed a practical modeling approach for two-phase micro-channel heat sinks and considered the discrepancy between flow patterns of air–water and steam–water flow in parallel micro-channels. Fukagata et al.[5]simulated an air–water two-phase flow in a20µm ID tube,numerically,with focus upon flow and heat transfer characteristics in the bubble train flows.He and Kasagi[6]simulated numerically adiabatic air water slug flow in a micro tube.They focused on pressure drop characteristics and their modeling.They found that the total pressure drop of a slug flow can be decomposed into a frictional pressure drop and a pressure drop over the bubble itself.Carlson et al.[7]1026-3098©2011Sharif University of Technology.Production and hosting by Elsevier B.V.All rights reserved.Peer review under responsibility of Sharif University of Technology.doi:10.1016/j.scient.2011.07.003924P.Hanafizadeh et al./Scientia Iranica,Transactions B:Mechanical Engineering18(2011)923–929investigated characteristics of multiphase dynamics,especially two-phase gas–liquid flow,by means of advanced numerical simulations.They compared two Computational Multi-Fluid Dynamic(CMFD)codes,Fluent and TransAT,and reported a prediction of recirculating flow in the bubbly flow case using TransAT,while significant recirculation was not observed in the solution using Fluent.Saison and Wongwises[8]performed a series of experiments in a horizontal circular micro channel with an inner diameter of0.15mm.They presented a flow pattern map in terms of the phase superficial velocities, and proposed a new pressure drop correlation for practical application.1.2.Phase change worksThe pool boiling heat transfer,in a vertical narrow annular with closed bottoms,was observed through a transparent quartz shroud by Yao and Chang[9],and stages of evolving boiling phenomena with an increase in heat flux were reported. Several researchers observed three basic flow patterns,namely, bubbly,slug and annular flow,in the mini pipe and channel. Damianides and Westwater[10]performed experiments with a1mm tube,and Mertz et al.[11]and Kasza et al.[12] studied the flow visualization of water nucleation in a single rectangular channel of2.5mm by6mm.Lin et al.[13]used a single round tube with2.1mm inside diameter for their experiments,and compared the flow transitions with those predicted by Bernea et al.[14].Sheng and Palm[15]performed their experiments with1–4mm diameter tubes.Cornwell and Kew[16]found three different flow patterns for R-113, namely,isolated bubbles confined bubbles and slug/annular flow,in rectangular channels with cross sectional areas of 1.2–0.9mm and3.5–1.1mm.Ory et al.[17]considered the effects of capillary,inertia,friction and gravity forces on the velocity distribution and temperature field along a single capillary two-phase flow in a heated micro-channel.Research dealing with gas–liquid,two-phase flow in micro-channels,in situations where fluid inertia was significant in comparison with surface tension,was reviewed by Ghiaasiaan and Abdel-Khalik[18].Jiang et al.[19]studied the boiling of water in triangular micro-channels,having widths of50and100µm. They observed individual bubbles at low heat fluxes,and an abrupt change in flow pattern to an unstable slug flow with increasing heat flux.Chedester and Ghiaasiaan[20]addressed the hydro-dynamically controlled Onset of a Significant Void (OSV)in heated micro tubes.They derived a simple semi-empirical correlation for the radius of departing bubbles at the OSV point to show the accuracy of their hypothesis.Some experimental studies have been reported on gas liquid two-phase flow in mini and micro conduits by Kandlikar[21],Lee and Mudawar[22]and Serizawa et al.[23].The three zone boiling heat transfer model was developed by Thome et al.[24]. Revellin and Thome[25]used an optical measurement method for two-phase characteristics of R-134a and R-245fa,in0.5mm and0.8mm diameter channels,to determine the frequency of bubbles existing in the microevaporator.They detected four flow patterns,namely,bubbly,slug,semi-annular and annular flow,whose transitions were not well compatible with neither the macroscale map of refrigerants nor the microscale map of air–water flow.Sobierska et al.[26] experimentally investigated the water boiling phenomena in a vertical rectangular microchannel,with a hydraulic diameter of0.48mm.They observed three main flow patterns,namely, bubbly,slug and annularflow.Figure1:Schematic of test apparatus.Due to the effects of surface tension,two-phase flows at mini and micro scale have different behavior in comparison with the macro scale.The aim of the present work is to visualize flow regimes in air–water two-phase flows and propose a flow regime map for such flows in vertical mini pipes.The neural network technique is implemented to recognize and predict a gas–liquid,two-phase flow pattern in mini tubes,having diameters of2,3and4mm.2.Experimental setupThis study is carried out by experimental apparatus schematically shown in Figure1.Air and water are used as gas and liquid phases in the experiments.The water flow rates are regulated by the needle valves and are measured by the cali-brated rotameter.Air and water are mixed together in a mixer made of acrylic glass and placed at the bottom of the riser pipe. The compressed air is fed by the compressor via an air injec-tor,which is schematically depicted in Figure2.The water flows from the center hole of a mixer,with a diameter of2mm,while air is injected into the holes around the center hole,each having 1mm diameter.The air flow rates are set by the regulator valve and are continuously measured by the calibrated gas rotameter. The overall height and inside diameter of the riser pipe are sum-marized in Table1.In order to have the opportunity to visually observe the two-phase flow patterns,the riser pipe was made of transparent glass.The air water mixture was directed upward through the riser,separated in the separation tank at the top of the riser and the air was discharged into the atmosphere.Differ-ent flow regime images were captured by a digital high speed camera,with a frame rate of1200fps,from the test section of the upriser.The test section is placed after the entrance section to diminish the effect of the entrance region.The length of the entrance section is about500mm.The superficial air and water velocities are0.5–10m/s and0.05–1m/s,respectively.P.Hanafizadeh et al./Scientia Iranica,Transactions B:Mechanical Engineering 18(2011)923–929925Figure 2:Schematic of air and watermixer.Figure 3:(a)RGB picture;(b)gray picture;and (c)subtracting and median process of flow in the pipe (3mm diameter).3.Experimental results 3.1.Image processingImage processing techniques must be performed in order to extract features from the images of the two-phase flow.Each picture has 8bit RGB (red,green and blue)color format,being converted from RGB to a grey scale mode.The output image has 256grey levels from 0(black)to 255(white).It is difficult to extract the bubbles directly from an original digital image and therefore preprocessing procedures must be undertaken to reduce noise and improve the quality of the images.An image-subtracted algorithm was used to reduce background noise by subtracting the background image from each dynamic image.In order to smooth the image border,a median filter was also used.A sliding window (3×3)was used in this process,and the median gray level of the pixels in the window was ter,the gray level of the pixels located at the center of the window was replaced by the median.The result of these processes is shown in Figure 3.3.1.1.Inverting binary imageThe images were converted from grayscale to binary mode by threshold segmentation,and an iterative procedure was used to calculate the optimizing threshold as follows [27]:Figure 4:Binary image of two-phase flow in the mini pipe.(a)The minimum and maximum of the gray level,namely Z land Z k ,are found in the image,and the initial value of the threshold is derived from their arithmetic average as:T 0=(Z k +Z l )/2.(1)(b)According to the initial value of threshold T K ,the imageis divided into two parts,namely,object and background,and the average value of the gray level in each part is calculated as:Z O =−Z (i ,j )<T kZ (i ,j )N O,(2)Z B =−Z (i ,j )>T kZ (i ,j )N B,(3)where Z (i ,j )is the gray level of the pixel (i ,j )in the image,N O is the number of the pixels in which Z (i ,j )is less than T K ,and N B is the number of pixels in which Z (i ,j )is more than T K .(c)The new threshold is calculated based on the arithmeticaverage of the object and background segments of the image as:T k +1=(Z O +Z B )/2.(4)If T K =T K +1,then the algorithm is finished,else K ≪=K +1,and turn to step (b).The binary image of the bubbles in the vertical pipe,which is the result of the above procedure,is shown in Figure 4.3.1.2.Image morphology processingSome morphological functions,such as dilation,erosion,opening and closing operations,were applied to modify the shapes of bubbles.Dilation adds pixels to the boundaries of the objects in an image,while erosion removes pixels on the object boundaries.The definition of a morphological opening of an image is erosion followed by dilation,using the same structuring element for both operations.The related operation,morphological closing of an image,is the reverse.It consists of dilation followed by erosion,with the same structuring element.Both of them do not significantly alter the area or shape of objects.The opening operation removes small objects and smoothes boundaries.Borders removed by erosion are restored by dilation,but small objects that were absorbed during erosion do not reappear after dilation.The closing operation was used to fill tiny holes and smooth boundaries.Objects were expanded by dilation and then reduced by erosion,so borders were smoothed and holes were filled [28,29].After926P.Hanafizadeh et al./Scientia Iranica,Transactions B:Mechanical Engineering 18(2011)923–929Figure 5:Final image of two-phase flow in the mini pipe.these operations,the result of image processing is shown in Figure 5.Bubble images of two-phase flow were clear using the above image processing,and it prepared bubbles for quantitative analysis,such as measuring area,perimeter and diameter.3.2.Flow pattern mapIn the experimental procedure while varying gas or liquid mass flow rate,a 10s film was recorded from the flow regime at a speed of 1200fps.The recorded film was replayed in slow motion for recognition of flow regimes.Each film converted to separate frames in a picture format using Adobe Premiere software.The achieved pictures were used as inputs of image processing techniques.The final binary pictures were used for the mentioned post processing procedure,such as flow regimedetection,void fraction and bubble velocity calculation,etc.Figure 6shows those typical flow regimes observed in the vertical,co-current,air–water,two-phase flows,in the 3mm mini pipe.Four basic flow patterns,namely,bubbly,slug,churn and annular,accompanied by their transitions,are illustrated in these figures.The visualization shows that air–water two-phase flows in mini pipes do not have three dimensional behaviors,especially in bubbly and slug flows.The final processed images of different flow regimes in air–water,two-phase flow in mini pipes have been presented in Figure 7.Figures 8–10show the flow pattern map for a vertical round tube with inner diameters of 2,3and 4mm,respectively.The proposed maps are in terms of superficial velocities of phases,and the four main flow patterns are depicted in these maps.In Figure 11,the achieved flow pattern for the pipe with 2mm ID was compared to the work of Ide et al.[30],shown by a solid line.They divided the flow pattern map into the four main regions,namely,dispersed bubbly flow,intermittent flow,churn flow and annular flow.The comparison shows that the bubbly and annular flows in the present work are not well in accordance with those of Ide et al.In the present work,the dispersed bubbles were not seen,because the air bubble injector did not have very thin holes.As a result,the created bubbles mostly have diameters in the range of the pipe diameter.Even the existence of air injectors with thin holes cannot guarantee the creation of bubbly flow.In the case of small bubbles occurring,as a result of thin holes in the air injector and the developed two-phase flow,they would collapse,resulting in large bubbles know as intermittent flow.Bubbly flows are mainly promoted by bubble breaking mechanisms,due to turbulence effects.It seems that in small diameter pipes,the formation of a specific flowpatterns(a)Bubbly.(b)Bubbly-slug.(c)Slug.(d)Messy-slug.(e)Churn.(f)Wispy-annular.(g)Ring.(h)Wavy-annular.(i)Annular.Figure 6:Different flow patterns in a vertical pipe with 2mm diameter.P.Hanafizadeh et al./Scientia Iranica,Transactions B:Mechanical Engineering18(2011)923–929927(a)Bubbly.(b)Slug.(c)Messy-slug.(d)Churn.(e)Ring.(f)Wavy-annular.Figure7:Final processed image of different two-phase flow regimes in the minipipe.Figure8:Flow patterns for2mm innerdiameter.Figure9:Flow patterns for3mm inner diameter.mainly depends on mixer configuration.The radial air supplierused in this study makes intermittent flow patterns,such asslug and churn flows,while the air supply in the tube centerfavors annular flow.This can be the reason for an absence ofannular flow in the proposed flow patterns.The comparison offlow patterns also reveals that the slug,messy slug andsemi-Figure10:Flow patterns for4mm inner diameter.annular flows in the proposed map are in accordance with theintermittent flow of Ide et al.[30].In the present study,a noticeable difference between flowpattern maps for vertical pipes with various diameters of2,3and4mm is not seen.This can be justified in regard tothe fact that the dominant forces acting on the air–watermixture in the small diameter pipes,namely,gravitation,inertia,surface tension and buoyancy forces,are in the sameorder of magnitude.This concept clearly indicates that thesethree flow patterns can be combined to form a new flow patternfor the gas–liquid,two-phase flow in small diameter pipes.A combination of these three flow patterns results in a newflow pattern map,which is illustrated in Figure12.A FuzzyC-Means clustering technique(FCM)was used to classify theflow patterns.The solid lines in the figure show the transitionregion of the flow patterns.This figure shows the achieved flowmap for mini pipes with diameters in the range of2–4mm.4.ConclusionIn this paper,air–water,two-phase flow patterns wereinvestigated experimentally for mini pipes with diameters of2,3and4mm.An image processing technique was used fordetection of flow patterns from pictures derived from filmsrecorded with a high speed camcorder.The obtained flowpatterns reveal that there is no noticeable difference between928P.Hanafizadeh et al./Scientia Iranica,Transactions B:Mechanical Engineering 18(2011)923–929Figure 11:Comparison between the achieved flow patterns with the work of Ide et al.[30]for a pipe with diameter of 2mm.Figure 12:Proposed two-phase vertical upward flow pattern map.two-phase,upward flow patterns in this range of diameters.A new flow pattern map was achieved for vertical mini pipes,due to a comparison of the flow patterns of these three diameters of pipe.The proposed map was compared with existing research.A comparison of the present work and previous research shows that the flow patterns of slug,messy slug and semi-annular in the present work are compatible with the intermittent flow pattern of Ide et al.[30].However,in the present study,the annular flow is seen at a lower superficial air velocity than that in the work of Ide et al.[30].AcknowledgmentsThis research was funded by Iran Supplying Petrochemical Industries,Parts,Equipment and Chemical Design Corporation (SPEC),as a joint research project with Sharif University of Technology (project No.KPR-8628077).References[1]Suo,M.and Griffith,P.‘‘Two-phase flow in capillary tubes’’,Int.J.Basic Eng.,86,pp.576–582(1964).[2]Sadatomi,Y.,Sato,Y.and Saruwatari,S.‘‘Two-phase flow in verticalnoncircular channels’’,Int.J.Multiphase Flow ,8,pp.641–655(1982).[3]Xu,J.L.,Cheng,P.and Zhao,T.S.‘‘Gas–liquid two-phase flow regimes inrectangular channels with mini/micro gaps’’,Int.J.Multiphase Flow ,25,pp.411–432(1999).[4]Hetsroni,G.,Mosyak, A.,Segal,Z.and Pogrebnyak, E.‘‘Two-phaseflow patterns in parallel micro-channels’’,Int.J.Multiphase Flow ,29,pp.341–360(2003).[5]Fugakata,K.,Kasagi,N.,Ua-arayaporn,P.and Himeno,T.‘‘Numericalsimulation of gas liquid two-phase flow and convective heat transfer in a micro tube’’,Int.J.Heat and Fluid Flow ,28,pp.72–82(2007).[6]He,Q.and Kasagi,N.‘‘Numerical investigation on flow pattern andpressure drop characteristics of slug flow in a micro tube’’,6th Int.ASME Conf.on Nanochannels,Microchannels and Minichannels ,Darmstadt,Germany,pp.24–35(2008).[7]Carlson,A.,Kudinov,P.and Narayanan,C.‘‘Prediction of two-phase flowin small tubes:a systematic comparison of state-of-the-art CMFD codes’’,5th Europe Thermal-Sci.Conf.,The Netherlands,pp.138–150(2008).[8]Saisorn,S.and Wongwises,S.‘‘An experimental investigation of two-phaseair–water flow through a horizontal circular micro-channel’’,Exp.Thermal Fluid Sci.,33,pp.306–315(2009).[9]Yao,S.C.and Chang,Y.‘‘Pool boiling heat transfer in a confined space’’,Int.J.Heat Mass Transf.,26,pp.841–848(1983).[10]Damianides,D.A.and Westwater,J.W.‘‘Two-phase flow patterns in acompact heat exchanger and in small tubes’’,2nd UK National Conf.on Heat Transf.,11,United Kingdom,London,pp.1257–1268(1988).[11]Mertz,R.,Wein,A.and Groll,C.‘‘Experimental investigation of flow boilingheat transfer in narrow channels’’,Calore e Technologia ,14(2),pp.47–54(1996).[12]Kasza,K.E.,Didascalou,T.and Wambsganss,M.W.‘‘Microscale flow visu-alization of nucleate boiling in small channels:mechanisms influencing heat transfer’’,Int.Conf.on Compact Heat Exchanges for the Process Indus-tries ,New York,USA,pp.343–352(1997).[13]Lin,S.,Kew,P.A.and Cornwell,K.‘‘Two-phase flow regimes and heattransfer in small tubes and channels’’,11th Int.Heat Transf.Conf.,Kyongju,Korea,2,pp.45–50(1998).[14]Barnea,D.,Luninsky,Y.and Taitel,Y.‘‘Flow pattern in horizontal andvertical two-phase flow in small diameter pipes’’,Canadian J.Chem.Eng.,61,pp.617–620(1983).[15]Sheng, C.H.and Palm, B.‘‘The visualization of boiling in small-diameter tubes’’,Int.Conf.on Heat Transport and Transport Phenomena in Microsystems ,Banff,Canada,pp.44–53(2001).[16]Cornwell,K.and Kew,P.A.‘‘Boiling in small parallel channels’’,CEC Conf.on Energy Eff.in Process Tech.,Athens,Greece,pp.624–638(1992).[17]Ory, E.,Yuan,H.,Prosperetti, A.,Popinet,S.and Zaleski,S.‘‘Growthand collapse of a vapor bubble in a narrow tube’’,Phys.Fluids ,12,pp.1268–1277(2000).[18]Ghiaasiaan,S.M.and Abdel-Khalik,S.I.‘‘Two-phase flow in micro-channels’’,Adv.Heat Transf.,34,pp.145–253(2001).[19]Jiang,L.,Wong,M.and Zohar,Y.‘‘Forced convection boiling in a micro-channel heat sink’’,Int.J.Micro-Electro-Mech.Sys.,10,pp.80–87(2000).[20]Chedester,R.C.and Ghiaasiaan,S.M.‘‘A proposed mechanism for hydrodynamically-controlled onset of significant void in microtubes’’,Int.J.Heat Fluid Flow ,23,pp.769–775(2002).[21]Kandlikar,S.G.‘‘Fundamental issues related to flow boiling in minichan-nels and microchannels’’,Exp.Therm.Fluid Sci.,26,pp.389–407(2002).[22]Lee,J.and Mudawar,I.‘‘Two phase flow in high heat flux micro channelheat sink for refrigeration cooling applications’’,Int.J.Heat Mass Transf.,48,pp.928–955(2005).[23]Serizawa,A.‘‘Gas liquid two-phase flow in microchannels’’,In MultiphaseFlow Handbook ,C.T.Crowe,Ed.,2nd ed.,pp.830–887,CRC Press (2006).[24]Thome,J.R.,Dupont,V.and Jacobi, A.M.‘‘Heat transfer model forevaporation in micro channels’’,Int.J.Heat Mass Transf.,47,pp.3375–3385(2004).P.Hanafizadeh et al./Scientia Iranica,Transactions B:Mechanical Engineering18(2011)923–929929[25]Revellin,R.and Thome,J.R.‘‘Experimental investigation of R-134a andR-245fa two-phase flow in microchannels for different flow conditions’’, Int.J.Heat Fluid Flow,28,pp.63–71(2007).[26]Sobierska, E.,Kulenovic,R.and Mertz,R.‘‘Heat transfer mechanismand flow pattern during flow boiling of water in a vertical narrow channel experimental results’’,Int.J.Thermal Sci.,46,pp.1172–1181 (2007).[27]Shi,L.‘‘Fuzzy recognition for gas–liquid two-phase flow pattern based onimage processing’’,Proc.of13rd IEEE Int.Conf.on Control and Automation, pp.1424–1427(2007).[28]Heijmans,H.J.A.M.,Morphological Image Operators,Academic Press,NewYork(1994).[29]/help/toolbox/images/index.html.[30]Ide,H.,Kariyasaki,A.and Fukano,T.‘‘Fundamental data on the gas–liquidtwo-phase flow in minichannels’’,Int.J.Thermal Sci.,46,pp.519–530 (2007).Pedram Hanafizadeh received his M.S.and Ph.D.Degrees in Mechanical Engineering from the Centre of Excellence in Energy Conversion at Sharif University of Technology,Tehran,Iran,in2005and2010,respectively.His work is mainly concentrated on the field of Multiphase Flow,Experimentally, Numerically and Analytically.His research interests include Characteristics of Multiphase Flow,Heat Transfer,Boiling and Condensation,Instrumentation in Fluid Flow,Image Processing for Flow Field Analysis,and Industrial and Applicable Usage of Multiphase Flow.Mohammad Hassan Saidi is Professor and Chairman of the School of Mechanical Engineering at Sharif University of Technology,Tehran,Iran. His current research interests include Multiphase Flows,Heat Transfer Enhancement in Boiling and Condensation,Modelling of Pulse Refrigeration, Vortex Tube Refrigerator,Indoor Air Quality and Clean Room Technology, Energy Efficiency in Home Appliances and Desiccant Cooling Systems.Arash Nouri Gheimasi obtained his B.S.Degree in Mechanical Engineering in2010,and is currently an M.S.student at the Centre of Excellence in Energy Conversion at the School of Mechanical Engineering,Sharif University of Technology,Tehran,Iran,under the supervision of Professor Saidi.His B.S. thesis involved work on the Characteristics of Gas-Liquid Two-Phase Flow in Mini Pipes and he is now working on Application of Visual Techniques in Two-Phase Flow.His research interests include the area of Two Phase Flow and Its Industrial Applications.Soheil Ghanbarzade received his B.S.and M.S.Degrees in Mechanical Engineering from the Centre of Excellence in Energy Conversion at Sharif University of Technology in2008and2010,respectively.Since then he has worked under the supervision of Professor M.H.Saidi as research staff in the Multiphase Group.His research interests include:Analytical,Numerical and Experimental Methods to Study Characteristics of Large Scale and Mini Scale Air-Water,Two-Phase Flows.He holds a Gold medal from the13th National Olympiad of Mechanical Engineering in Iran,and is currently a Ph.D.student of Petroleum Engineering at the University of Texas,Austin,USA.。

岩土工程专业词汇——A辑•absorbed water 吸着水•aging 老化abutment 桥台•allowable bearing capacity 容许承载力•acoustic emission of rock 岩石声发射•allowable settlement 容许沉降•acoustic prospecting 声学探测•alluvial fan 冲积扇•active earth pressure 主动土压力•alluvial soil 冲积土•alternating load 交变荷载•active fault 活断层•anchoring，bolting 锚固bolting（螺栓连接, 拧紧）•activity index 活动性指数•anchor slab wall 锚定板墙•additional stress,superimposed stress 附加应力•anticline 背斜•aquifer 含水层•adverse geologic phenomena 不良地质现象•artesian(自流井的) pressure head 承压水头•aeolian deposit 风积土•artificial fill 人工填土•aeration(通风) zone water 包气带水•attitude 产状•average thickness method 平均厚度法•aggregate for concrete 混凝土骨料岩土工程专业词汇——B辑•back analysis 反演分析•bearing capacity factor 承载力因数•base course 道路基层•bearing stratum 持力层•bed rock 基岩•berm 反压平台( 反压马道) •boulder(stone block) 漂石( 块石) •Biot consolidation theory 比奥固结理论•Boussinesq theory 布辛涅斯克理论•Bishop's somplified method of slices 毕肖普简化条分法•bridge pier 桥墩brittle failure 脆性破坏•bulk modulus 体积模量blasting 爆破•buoyancy 浮托力block theory 块体理论•buttress retaining wall 支墩式挡墙•body force 体积力borrow area 取土场•by-pass seepage 绕渗岩土工程专业词汇——C辑•Cahtornia Bearing Ratio(CBR) 加州承载比•cobble 卵石( 碎石) •coefficient of collapsibility 湿陷系数•cantilever retaining wall 悬臂式挡墙•coefficient of compressibility 压缩系数•capillary water 毛细管水•categortzatlon of geotechnical projects 岩土工程分级•coefficient of consolidation 固结系数•cemment stabilization 水泥加固•coefficient of curvature 曲率系数•central load 中心荷载( 轴心荷载) •coefficient of deformation due to leaching 溶滤变形系数•Cerruti‘ssolution 色卢铁解答•coefficient of permeability 渗透系数•check dam 谷坊chemical grouting 化学灌浆•coeffieient of secondary consolidation 次固结系数•coarse aggregate 粗骨料•coefficient of self-weight collapsibility 自重湿陷系数•coarse-grained soil 粗粒土•coefficient of uniformity 均匀系数•compression index 压缩指数•coefficient of viscosity 粘滞系数•compressive strength 抗压强度•coefficient of volume compressibility 体积压缩系数•concentrated load 集中荷载( 点荷载) •coefficient of weathering 风化系数•concrete faced rockfill dam 混凝土面板堆石坝•cohesion 粘聚力•cone penetration test(CPT) 静力触探试验•cohesionless soil 无粘性土•cohesive soil 粘性土•confined warer 承压水•collapsibility 湿陷性•consistency limit 稠度界限•collapsibility test of loess 黄土湿陷试验•consolidated-drained triaxial test 固结排水三轴试验•collapsible soil 湿陷性土•consolidated quick shear test 固结快剪试验•combined bolting and shotcrete 喷锚支护•consolidated - undrainbd triaxial test 固结不排水三轴试验•compaetibility 压实性•compaction by rolling 碾压法•compaction grouting method 挤密喷浆法•consolidation 固结•consolidation cure 固结曲线岩土工程专业词汇——C辑•compacting mechinery 碾压机械•consolidation grouting 固结灌浆•compaction test 击实试验•composite ground 复合地基•consolidation settlement 固结沉降•composite slip surface 复合滑动面•consolidation test 固结试验•comprehensive engineering geological map 综合工程地质图岩土工程专业词汇——C辑•constitutive relation of soil土的本构关系( 本构模型) •comoressibility 压缩性•constrained grain size 限制粒径•Coulomb's earth pressure theory库仑土压力理论•counterfort retaining wall 扶壁式挡墙•constrained modulus 压缩模量•crack growth 裂纹扩展creep 蠕变•contact pressure 基底压力( 接触压力) •critical edge pressure 临塑荷载•contour map of groundwater 地下水等水位线图•critical height of slope( 土坡) 临界高度•control of underground water 地下水控制•critical hydraulic gradient 临界水力梯度•core of rock 岩芯•core recovery 岩芯采取率•critical void ratio 临界孔隙比•core wall 心墙•corner-points method 角点法•cross - hole method 跨孔法•curtain grouting 帷幕灌浆•Coulomb - Navier strength theory 库仑—纳维强度理论•cushion 垫层cutting 路堑•cutoff wall 截水墙cyclic load 周期荷载岩土工程专业词汇——D辑•Darcy's law 达西定律•Degree of consolidation 固结度•debris flow 泥石流•degree of saturation 饱和度•deep mixing method 深层搅拌法•delayed elasticity 弹性后效•deep soil stabilization 深层土加固•densification by explosion 爆炸加密法•deep well method 深井法•degree of compaction 压实度•densification by sand pile 挤密砂桩•density 密度•depletion coefficient 疏干系数•dispersion coefficient 弥散系数•developing chart of exploratory drift坑洞展示图•dispersive clay 分散性粘土•disturbed soil sample 扰动土样•dewatering method 降水法•differential settlement 沉降差•diversion tunnel 导流洞•down-hole method 下孔法•dike ，levee 堤drainage method 排水法•dilatancy 剪胀性dynamic consolidation 强夯法•dilatancy of rock 岩石扩容•dynamic load 动荷载diluvial fan 洪积扇•dynamic penetration test 动力触探试验•diluvial soil 洪积土[ dai’lju:vjel ]•direct shear test 直剪试验•dynamic simple shear test 动单剪试验•disaster geology 灾害地质学•dynamic triaxial test 动三错轴试验•dislocation 位错岩土工程专业词汇——E辑•earth dam 土坝•effective grain size 有效粒径•earth material 土料•earth pressure at rest 静止土压力•effective porosity 有效孔隙率•effective stress 有效应力•earthquake engineering 地震工程学•effective stress analysis 有效应力分析•earthquake subsidence 震陷•elastic strain 弹性应变•electrical prospecting 电法勘探•earth — rock dam 土石坝•earthwork 土石方工程electro-osmosis 电渗法•eccentric load 偏心荷载embankment 路堤•engineering geologic columnar profile •工程地质柱状图•environmental geotechnics 环境岩士工程•engineering geologic exploration 工程地质勘探•equipotential line 等势线•engineering geologic drilling 工程地质钻探•equivalent opening size(EOS) 等效孔径•engmeering geologic evaluation 工程地质评价•excavating machinery 挖掘机械•engineering geologic map 工程地质图•excavation 挖方•excess pore water pressure 超静水压力•engineering geologic mapping 工程地质测绘•expanded polystyrene(EPS) [ ’poli ’staieri:n ]•聚苯乙烯发泡材料•engineering geologic profile 工程地质剖面图•expansive soil 膨胀土岩土工程专业词汇——F辑•fabric sheet reinforced earth 铺网法•fine-grained soil 细粒土fissuerd clay 裂隙粘土•factor of safety 安全系数fissure water 裂隙水•failure strength 破坏强度•flat jack technique 扁千斤顶法•fall of ground 冒顶fatigue strength 疲劳强度•flow line 流线fault 断层•flow net 流网•field identification of soil 土的现场鉴别•fold 褶皱fraction 粒组•field observation 现场观测•fracture mechanics 断裂力学•fill 填方fracture zone 断裂破碎带•filter 反滤层free face 临空面•final settlement 最终沉降•free swelling ratio 自由膨胀率•fine aggregate 细骨料free water 自由水•frost heave 冻胀freezing method 冻结法•frost — heave capacity 冻胀量•french drain 盲沟fresh rock 新鲜岩石•frost heaving pressure 冻胀力•friction — resistance ratio 摩阻比•frozen soil 冻土岩土工程专业词汇——G辑•gap — graded soil 不连续级配土•geosynthetics 土工合成材料•geotechnical centrifugal model土工离心模型试验•general shear failure 整体剪切破坏•geotechnical engineering 岩土工程•geocomposite 土工复合材料•geotechnical engineering investigation 岩土工程勘察•Geodrain/prefabricated（预制）strip drain 塑性排水带法•Geofabriform 土工模袋geogrid 土工格栅•geotextiles 土工织物•geologic environment 地质环境•gradation 级配grain size 粒径•grain size distribution curve 粒径分布曲线•geologic structure 地质构造•gravelly soil 砾类土gravitational water重力水•geomechanical model test 地质力学模型试验•gravity retaining wall 重力式挡墙•geomembrane 土工膜geomorphology 地貌•Griffith 's strength criterion 格里菲斯强度准则•geophysical exploration 地球物理勘探•ground fracturing 地裂ground water 地下水•ground treatment 地基处理•geostatic stress 自重应力岩土工程专业词汇——G辑•groundwater dynamics 地下水动力学•groundwater regime 地下水动态•groundwater storage 地下水贮藏量•groundwater hardness 地下水硬度•groundwater monitoring 地下水监测•groin，spur dike丁坝grouting 灌浆•groundwater pollution 地下水污染•grouting test 灌浆试验guide adit 导洞•groundwater recharge 地下水补给量岩土工程专业词汇——H辑•hardness degree of rock 岩石坚硬程度•hydrogeological drilling 水文地质钻探•hydrogeological investigation 水文地质勘查•hydraulic fracturing technique 水力劈裂法•hydrogeology 水文地质学•hydrostatic pressure 静水压力•hydraulic fill 水力冲填•hydraulic gradient 水力梯度•heaving of the bottom 基坑底隆胀•hydraulic tunnel 水工隧道岩土工程专业词汇——I辑•influence chart 感应图•influence chart 感应图•infra - red detection 红外探测•internal friction angle 内摩擦角•investigation stage 勘察阶段•isochrone 等时孔压线•isoline method 等值线法•interstrated water 层间水岩土工程专业词汇——J、K辑•jet grouting method 高压喷射注浆法•joint 节理•Kaiser effect 凯塞效应•karst land feature 喀斯特地貌•karst 喀斯特( 岩溶)•karst collapse 喀斯特塌陷•Ko — consolidation Ko 固结岩土工程专业词汇——L辑•Lamilar flow 层流line load线荷载•landform unit 地貌单元•liquefaction of sand 砂土液化•landslide 滑坡landslide mass 滑坡体•liquefaction potential 液化势•land subsidence 地面下沉laterite 红土•liquidity index 液性指数liquid limit液限•layerwise summation method 分层总和法•local shear failure 局部剪切破坏•loess 黄土/’louis/ lime pile method 石灰桩法•loess - like soil 黄土状土lime soil pile 灰土桩lime treated soil 灰土longitudinal dike 顺坝•limit equilibrium method 极限平衡法•long-term modulus 长期模量•linear shrinkage ratio 线缩率•long-term stability 长期稳定性•Lugeon unit 吕荣单位岩土工程专业词汇——M辑•magmatic rock(igneous rock) 岩浆岩( 火成岩)•modulus of elasticity 弹性模量( 杨氏模量)•marine soil 海积土•masonry retaining wall 砌体挡墙•modulus of pressuremeter 旁压仪模量•maximum dry density 最大干密度•Mohr-Coulomb law 摩尔库仑定律•mechanical properties of rock 岩石力学性质岩土工程专业词汇——M辑•monitoring of landslide 滑坡监测•mechanics of granular media 散体力学•monitoring of pore-water pressure孔隙水压力监测•metamorphic rock 变质岩•monitoring of settlement and deformation 沉降变形监测•method of slice 条分法micro-crack 微裂纹岩土工程专业词汇——M辑•Mindlin’s solution 明德林解答muck 淤泥•monitoring of surrounding rock deformation of tunnel 洞室围岩变形监测•modified Griffith 's criterion•修正的格里菲斯准则•mini pile 微型桩mud pumping 翻浆冒泥•modulus of deformation 变形模量岩土工程专业词汇——N辑•natural angle of repose 天然休止角•New Austrian Tunnelling Method(NA TM) 新奥法•natural building materials 天然建筑材料•non-uniform settlement 不均匀沉降•needle — punched geotextile 针刺土工织物•nonwoven geotextile 无纺土工织物•normalization 归一化•normally consolidation soil 正常固结土岩土工程专业词汇——O辑•optinum moisture content 最优含水率•overburden pressure 覆盖压力•organic soil 有机质土•overconsolidation soil 超固结土•osmosic pressure method 渗透压法•over consolidation ratio(OCR) 超固结比•out and cover method 明挖法•over coarse-grained soil 巨粒土•overbreak 超挖overburden layer 覆盖层岩土工程专业词汇——P辑•packed drain ，fabric-enclosed drain 袋装砂井•permeability 渗透性•parallel section method 平行断面法•phreatic line 浸润线phreatic water 潜水•parapet wall 防浪墙•peremeability test 渗透试验•particle size analysis 颗粒分析试验•physical properties of rock 岩石的物理性质•passive earth pressure 被动土压力•piezocone test(CPTU) 孔压静力触探试验•peak strength 峰值强度•piezometric head 测压管水头•peat 泥炭penetration resistace 贯入阻力•pipe jacking method 顶管法•perched water 上层滞水piping 管涌•plane strain test 平面应变试验•plastic failure 塑性破坏•plastic flow 塑流•porosity 孔隙率•plastic limit 塑限•preconsolidation pressure 先期固结压力•plastic strain 塑性应变•plastic zone 塑性区•preloading method 预压法•plasticity chart 塑性图•pressure bulb 压力泡•plasticity index 塑性指数•pressuremeter test(PMT) 旁压试验•plate loading test 平板荷载试验•primary stress 初始应力( 地应力) •point loading test 点荷载试验•primary consolidation 主固结•Poisson's ratio 泊松比•poorly-graded soil 不良级配土•principle of effective stress 有效应力原理•pore air pressure 孔隙气压力•progressive failure 渐近破坏•pore pressure 孔隙压力•prototype monitoring 原型监测•pore pressure parameter 孔隙水压力系数•punching shear failure 冲剪破坏•pore pressure ratio 孔隙压力比•pump — in test 压水试验•pore water 孔隙水•pumping test 抽水试验•pore water pressure 孔隙水压力岩土工程专业词汇——Q辑•quick sand 流砂•quick shear test 快剪试验岩土工程专业词汇——R辑•radial flat jack technique 径向扁千斤顶法•radius of influence 影响半径•radial wells 辐射井•Rankine's earth pressure theory•兰金土压力理论•rate of settlement 沉降速率•retaining wall 挡土墙( 挡墙) •retardation 滞后•rate of stripping 剥离比•rheology 流变学•rebound modulus 回弹模量•road pavement 道路路面•rebound of foundation 地基回弹•rockburst 岩爆•recharge area 补给区rechargerate 补给率•rock classification 岩石分类•rock engineering 岩石工程•recharge method 回灌法rock fall 岩崩•reclamation 围垦工程rockfill dam 堆石坝•reinforced earth 加筋土rock mass 岩体•reinforced soil wall 加筋土挡墙•rock mass basic quality 岩体基本质量•relative density 相对密度•rock mechanics 岩石力学( 岩体力学) •relaxation time 松弛时间relief well 减压井•rock quality designation(RQD) 岩石质量指标•remolded strength 重塑强度•rolled fill earth dam 碾压土坝•remote sensmg prospecting 遥感勘测•rolling compaction test 碾压试验•reserve of building material 建材储量•root pile 树根桩residual soil 残积土•rose diagram of joints 节理玫瑰图•residual strength 残余强度runoff area 径流区•resonant column test 共振桩试验•rupture，fracture 断裂岩土工程专业词汇——S辑•saline soil 盐渍土sandy soil 砂井土•sand drain 排水砂井saturation curve 饱和曲线•scale effect 尺度效应•shield driving method 盾构法•seasonally frozen soil 季节冻土•shortcrete 喷射混凝土•secondary consolidation 次固结•shrinkage index 缩性指数•shrinkage limit 缩限side ditch 侧沟•secondary consolidation settle ment 次固结沉降•simple shear test 单剪试验•sedimentary rock 沉积岩•seepage 渗流slaking 湿化•air slaking潮解, 风化dry slaking干式消化[熟化]•lime slaking石灰熟化; 石灰消化•seepage deformation 渗透变•slide — resistant pile 抗滑桩•slip surface 滑动面•seepage failure 渗透破坏slip zone 滑动带•seepage force 渗流力sloping wall 斜墙•seepage path 渗径•slow shear test 慢剪试验•soft clay 软粘土•self —boring pressuremeter 自钻式旁压仪•soil dynamics 土动力学•oil fabric 土的组构•semi 一infinite elastic body 半无限弹性体•soil flow 流土soil mass 土体•sensitivity 灵敏度soil mechanics 土力学•settlement 沉降soil nailing 土钉•settlement calculation depth 沉降计算深度•soil sampler 取土器soil skeleton 土骨架•settlement curve 沉降曲线special soil 特殊土•shear modulus 剪切模量soldier pile 排桩•shear strain 剪应变soil structure 土的结构•shear strength 抗剪强度sheet pile wall板桩墙•specific gravity of soil particle 土粒比重•specific grout absorption•单位吸浆量(比吸浆量)•stress concentration 应力集中•specific penetration resistance 比贯入阻力•stress controlled test 应力控制试验•specific surface 比表面积•stress distribution 应力分布•specific water absorption 单位吸水量•stress history 应力历史•specific yield 给水度stress level 应力水平•split test 劈裂试验( 巴西试验) •stress path 应力路径•stress recovery method 应力恢复法•stability analysis 稳定分析•stability number 稳定数•stress relaxation 应力松弛•stable crack growth 稳定裂纹扩展•stress relief method 压力解除法•standard penetration test 标准贯入试验•stress space 应力空间•stip load 条形荷载•state of plastic equilibrium 塑性平衡状态•structural block 结构体•steady seepage 稳定渗流•structural plane 结构面stone material 石料•structural types of rock mass 岩体结构类型•stereographlc projection 赤平投影•subsurface runoff 地下径流•storage coefficient 储水系数•sureharge 超载surface force 表面力•strain controlled test 应变控制试验•surface soil stabilization 浅层土加固•strain hardening 应变硬化•surface water 地表水strain space 应变空间•surface wave vecosity method 表面波法•strain softening 应变软化•strength envelope 强度包线•surrounding rock 围岩•surrounding rock stress 围岩应力( 二次应力)•swelling force 膨胀力•swelling index 回弹指数•Swedish circle method 瑞典圆弧法•swelling ratio 膨胀率•syncline 向斜岩土工程专业词汇——T辑•tailings dam 尾矿坝•total stress analysis 总应力分析•tensile strength 抗拉强度•Terzaghi's consolidation theory太沙基固结理论•transducer 传感器transient load 瞬时荷载•texsol ，fibre soil 纤维土•transmissivity 导水系数•thaw collapsibility 融陷性trench 探槽•thick wall sampler 厚壁取土器•trench out method 挖沟法•triangular method 三角形法•thin wall sampler 薄壁取土器•triaxial compression test 三轴压缩试验( 三轴剪切试验)•thixotropy 触变性three phase diagram 三相图•tieback wall ，anchored wall 锚杆挡墙•triaxial extension test 三轴伸长试验•time factor 时间因数•torsional shear test 扭剪试验•total mineralization of groundwater •lining 隧道衬砌turbulent flow 紊流•total stress 总应力•true triaxial test 真三轴试验•triaxial test 三轴试验岩土工程专业词汇——U辑•ultimate bearing capacity 极限承载力•unconfined compressive streng th test •无侧限抗压强度试验•underlying stratum 下卧层•unconsolidated undrained 不固结不排水•underpinning 托换技术•under-consolidated soil 欠固结土岩土工程专业词汇——U辑•undisturbed soil sample不扰动土样( 原状土样) •uniformly distributed load 均布荷载•underground diaphragm wall 地下连续墙•unit weight 容量•up — hole method 上孔法•underground opening 地下洞室•uplift pressure 扬压力岩土工程专业词汇——V辑•vacuum preloading 真空预压法•vertical shaft 竖井vibroflotation 振冲法•valley terrace 河谷阶地•void ratio 孔隙比•vane shear test 十字板剪切试验•volumetric strain 体应变•volume shrinkage ratio 体缩率•varved clay 带状粘土岩土工程专业词汇——W辑•water bearing capacity 容水量•weathered crust 风化壳•water content 含水率•weathered rock 风化岩石•water injeeting test 注水试验•weathered zone 风化带•weathering 风化作用•water retaining capacity 持水度•weathering degree of rock 岩石风化程度•weak intercalated layer 软弱夹层•well point 井点排水•well — graded soil 良好级配土•weak structural plane 软弱结构面。

高三英语计算机语言单选题40题1. When you are programming, you often need to use a(n) ______ to store data.A. algorithmB. variableC. functionD. loop答案：B。

本题考查计算机语言中的常见词汇。

选项A“algorithm”意为“算法”；选项B“variable”指“变量”，在编程中用于存储数据，符合题意；选项C“function”是“函数”；选项D“loop”是“循环”。

2. In computer programming, a(n) ______ is a set of instructions that tells the computer what to do.A. codeB. scriptC. commandD. syntax答案：A。

选项A“code”指“代码”，是一组指令；选项B“script”通常指“脚本”；选项C“command”意为“命令”；选项D“syntax”表示“语法”。

本题强调的是一组指令，所以选A。

3. Which of the following is NOT a type of programming language?A. PythonB. ExcelC. JavaD. C++答案：B。

选项A“Python”、选项C“Java”和选项D“C++”都是常见的编程语言；选项B“Excel”是电子表格软件，不是编程语言。

4. The process of finding and fixing errors in a program is called ______.A. debuggingB. compilingC. optimizingD. documenting答案：A。

“debugging”意为“调试”，即查找和修复程序中的错误；“compiling”是“编译”；“optimizing”指“优化”；“documenting”表示“文档化”。

文章编号：1000 − 7393(2023)06 − 0704 − 08 DOI: 10.13639/j.odpt.202310001钻井压耗工程公式估算漏层位置罗黎敏1 谭睿2 耿立军1 李小波3 徐正贤2 闫伟21. 中海石油(中国)有限公司天津分公司；2. 中国石油大学(北京)非常规油气科学技术研究院；3. 中海油安全技术服务有限公司引用格式：罗黎敏，谭睿，耿立军，李小波，徐正贤，闫伟. 钻井压耗工程公式估算漏层位置［J ］. 石油钻采工艺，2023，45（6）：704-711.摘要：钻井作业过程中井漏现象会损害油气储层，引发井塌、井喷、卡钻以及部分井段报废等恶性事件，是制约油气田安全高效开发的技术难题，准确判定漏层位置是现场解决钻井井漏问题的关键。

常规漏层位置确定方法计算复杂、误差大、经济性差，因此基于钻井压耗公式和立压、套压数据提出了一种快速估算漏层位置和喷漏转换时间的方法。

根据入口注入流量、排量、循环压耗和流态数据计算流态摩阻，结合漏失前后立压、套压数据迭代计算漏失层位、确定漏点位置，并根据转喷立压、套压确定漏喷转换时间。

对渤海区域2口漏失井的漏层位置计算发现，估算的漏层位置与现场探明的漏层位置重叠性较高，计算误差低于5.06%，估算漏层范围100%覆盖现场漏失层位；并且立压越大，漏喷转换时间越长。

该方法计算过程方便简洁，降低过多输入参数精度不足带来的误差，可作为漏层位置确定的前置计算方法为现场堵漏作业提供技术支撑。

关键词：钻井；压耗；井漏；漏层位置；立管压力；套管压力； 工程技术中图分类号：TE28 文献标识码： AEstimating the location of leakage layer by drilling pressure loss engineering formulaLUO Limin 1, TAN Rui 2, GENG Lijun 1, LI Xiaobo 3, XU Zhengxian 2, YAN Wei 21. CNOOC (China ) Tianjin Branch Company , Tianjin 300452, China ;2. Unconventional Oil and Gas Science and Technology Research Institute , China University of Petroleum (Beijing ), Beijing 102249, China ;3. CNOOC Safety Technology Service Co., Ltd., Tianjin 300452, ChinaCitation: LUO Limin, TAN Rui, GENG Lijun, LI Xiaobo, XU Zhengxian, YAN Wei. Estimating the location of leakage layer by drilling pressure loss engineering formula ［J ］. Oil Drilling & Production Technology, 2023, 45(6): 704-711.Abstract: The phenomenon of well leakage during drilling operations can damage oil and gas reservoirs, leading to malignant events such as well collapse, blowout, stuck drilling, and scrapping of some sections of the well, which poses a technical challenge in the safe and efficient development of oil and gas fields. Accurately determining the location of formation leakage is crucial for on-site resolution of drilling leakage issues. Conventional methods for determining leakage locations are complex, prone to errors, and economically inefficient. Therefore, a rapid method for estimating the leakage location and the time of leakage transition based on drilling pressure loss formula, standpipe pressure, and casing pressure was proposed. The method involves calculating flow regime friction based on the injected flow rate, displacement, circulating pressure loss, and flow regime, calculating leakage formation and determining leakage location based on the iteration of data from standpipe pressure and casing pressure before and after leakage, and determining the time for leakage transition according to standpipe pressure and casing pressure. The calculated leakage location for two leaking wells in the Bohai Sea region show a high overlap with the confirmed leakage location on site, with a calculation error below 5.06%, and the estimated leakage range 100% covers the on-site leakage locations. Furthermore, as the standpipe pressure第一作者: 罗黎敏(1979-)，2003年毕业于江汉石油学院石油工程专业，现主要从事海上油气田钻完井作业管理和技术研究工作，高级工程师。

九年级英语科研技术单选题50题1. In the field of artificial intelligence research, the term "algorithm" refers to _.A. a set of computational steps and rules for performing a specific taskB. a physical device used for data storageC. a type of software for graphic designD. a chemical substance for experiments答案：A。

解析：“algorithm”（ 算法）的定义就是执行特定任务的一组计算步骤和规则。

选项B，用于数据存储的物理设备是硬盘、U 盘等，不是算法的含义。

选项C，用于图形设计的软件有Photoshop 等，和算法概念不同。

选项D，算法与化学实验物质毫无关系。

2. When scientists conduct genetic research, they often study "DNA replication". "Replication" here means _.A. the process of making an exact copy of DNAB. the destruction of DNA structureC. the separation of different DNA segmentsD. the combination of DNA with other substances答案：A。

解析：“replication”在“DNA replication”DNA复制）中的意思就是制造DNA精确副本的过程。

选项B，破坏DNA结构不是“replication”的含义。

Flexible Flow Aggregation for Adaptive Network MonitoringFalko DresslerAutonomic Networking GroupDept.of Computer Science7 University of Erlangen,Germany dressler@informatik.uni-erlangen.deGerhard M¨u nzComputer Networks and Internet Wilhelm Schickard Institute for Computer Science University of T¨u bingen,Germanymuenz@informatik.uni-tuebingen.deAbstractNetwork monitoring is a major building block for many domains in communication networks.Besides typical ac-counting mechanisms and the emerging area of charging in next generation networks,especially network security solu-tions rely on efficient and optimized work monitoring in high-speed networks is usually based onflow accounting and aggregation techniques represent a neces-sary enhancement in order to cope with increasing amounts of monitoring data that accrue with the ever-growing net-work capacities.In this paper,we propose aflexibleflow aggregation mechanism that can be directly employed on a monitoring probe to reduce the memory and processing demands.Alternatively,it can work as a concentrator that collectsflow data from multiple monitoring probes,com-bines and aggregates them and forwards the results to an analyzer.We verified and evaluated the aggregation mech-anism by integrating it into our monitoring probe Vermont. Our approach opens new prospects for high-speed network monitoring and allows coping with special situations that cannot be treated satisfyingly by traditionalflow account-ing,such as distributed denial-of-service attacks causing very high numbers offlows.Aggregatedflow data are an easy-to-handle form of packet information especially for anomaly detection and accounting issues.1.IntroductionNetwork monitoring has become an important research area since the development of thefirst computer networks. At the beginning,monitoring data were used for network maintenance.In this context,network operators employed network monitoring techniques[1]in order to verify the de-ployed routing strategies or to locate potential bottlenecks for sufficient re-dimensioning of the networks[6].Soon, additional application domains appeared such as account-ing and the detection of suspicious activities such as at-tacks,propagating worms etc[10].While the aforemen-tioned applications are still relevant,other domains also re-quire appropriate network monitoring such as intrusion de-tection and traceback mechanisms[13]or charging applica-tions for next generation networks[7].Manufacturers have responded by developing specific monitoring devices or by adding monitoring functionality into existing products like routers orfirewalls.A common monitoring technique isflow accounting. The key idea offlow accounting is to store information about packetflows and the corresponding statistics instead of individual packet information.In this context,aflow is defined as a unidirectional stream of IP packets identified by a common IP-five-tuple(protocol type,source IP address, destination IP address,source port,destination port).To this end,a single measurement record can contain informa-tion on up to several thousand packets.For transmission of monitoring data to a remote analyzer,a series of protocols named Netflow were developed by Cisco.The latest version Netflow.v9has been documented as informational RFC[2]. Its successor is the IPFIX protocol[3]and the correspond-ing information model[15],both developed by the IPFIX working group at the IETF.First open-source implementa-tions that support Netflow.v9as well as IPFIX are available, e.g.nprobe[4],NetMate[19],and Vermont[11].Althoughflow accounting works well under normal con-ditions(typical connections consist of about7.7packets per flow[12]),there is a major problem during DDoS(Dis-tributed Denial of Service)attacks[14].Many attacks ran-domly forge the source addresses of the attack packets, which results in an enormous number of different IP-five-tuples,each constituting an individualflow.As a conse-quence,manyflows get lost because of limited resources at the monitoring probe,and the processedflows cause a huge amount of mostly useless data that is sent to the analyzer. Moreover,the analyzer will face a serious performance de-ficiency in processing all the received monitoring data.To cope with this problem,flow sampling orflow aggre-gation schemes need to be employed.Flow sampling uti-N e t f lo w .v 9I P F I XI P F I XMonitoring ProbesMonitoring DataI P F I XN et f l o w .v 9ConcentratorConfiguration Figure 1.Adaptive network monitoringlizes similar sampling and filtering algorithms as proposed in PSAMP but applies them to flows instead of packets [8].Adaptive flow aggregation has been addressed by Hu et al.[9],proposing to adapt the aggregation level dynami-cally according to the available resources of the monitoring probe.Although this approach avoids flow loss in DDoS attack situations as described above,it does not take into account that arbitrarily defined flow aggregates usually do not meet the requirements of the analyzer.If,on the other hand,the analyzer was satisfied with aggregated flow data,the corresponding aggregation could be constantly applied.In this paper,we introduce an aggregation mechanism that performs flow aggregation in a configurable manner.This aggregation mechanism allows adapting the amount of exported monitoring data to the current needs and available resources of the analyzer.Flexibility is obtained by rule-based control of the aggregation process,thus providing the basic functionality to build a control-loop between monitor-ing probe and analyzer as sketched in figure 1.With exam-ple configurations,we demonstrate the applicability and the advantages of our approach.Furthermore,we evaluated the capability to reduce the amount of flow data by monitoring two typical network links:the network connection of our group server and the Internet connection of the University of Erlangen.The remainder of the paper is organized as follows.The proposed aggregation mechanism is discussed in section 2.The implementation and evaluation of the method is pre-sented in section 3.A discussion in section 4concludes the paper.2.Flow Aggregation2.1.Reasons for Introducing Flow Aggre-gation TechniquesUsually,a flow is defined as a set of packets that share the same IP-five-tuple information and that are observed within a given time interval at a specific observation point in the network.Distinguishing flows on the basis of the IP-five-tuple information results in a large amount of monitoring data that has to be exported by the monitoring probe,and later received and processed by the analyzer.In order to limit the amount of exported monitoring data,commercial monitoring probes often export only flow records with high traffic volume.Another approach is to apply sampling al-gorithms on the generated flow records that –similar to packet sampling techniques –select a subset of all gener-ated flow records for export [8].On the other hand,the fine IP-five-tuple granularity is not required by many appli-cations processing the monitoring data.Such a typical ap-plication is accounting,where the total traffic volume sent and received by a customer’s host has to be determined for billing purposes.Reducing the amount of monitoring data is also required if the data is to be stored.Note that already in midsize enterprise networks,one or more Gigabytes of flow records are easily generated in a single day 1.Flow aggregation techniques reduce the amount of mon-itoring data by discarding unneeded flow information and merging multiple flow records with similar properties.As a result,the aggregated flow information describes the same traffic observation at a coarser level of granularity and with fewer details.The following list gives some examples:•A common way to reduce the amount of monitoring data is to reduce the time resolution by merging sub-sequent records of the same flow into one record cov-ering a larger time interval.This technique is often deployed if the monitoring data has to be archived.•In case of client-server applications,information about the requested service is mostly contained in the port number at the server side.The client port number is usually dynamically assigned by the operating system of the client host.Hence,client port numbers can be discarded from the flow information without loss of valuable information.•As mentioned above,accounting applications usually require information about the traffic volume sent and received by a connected client or exchanged at a gate-way between two peering rmation about every single flow is not required.Especially,source1Inthe backbone network of the University of T¨u bingen,about 10GB per day of Netflow data are generated.and/or destination addresses and port numbers are not considered and can be discarded.Currently,flow aggregation techniques are largely de-ployed at the analyzer.Depending on the needs of the ap-plication,the monitoring data is compressed and unneeded information is discarded.However,flow aggregation per-formed at the analyzer does not reduce the amount of moni-toring data generated by the monitoring probe and transmit-ted to the analyzer.To circumvent this problem,we propose a configurable aggregation scheme that is deployed at the monitoring probe before the monitoring data is exported.2.2.IPFIX Aggregation at the MonitoringProbeIn an RFC summarizing the IPFIX requirements[16], the IPFIX working group at the IETF has identified the ne-cessity of implementingflow aggregation functionality in the monitoring probes.This idea is picked up in the IPFIX architecture draft[17]where an example shows howflows can be aggregated after masking the source and/or destina-tion IP addresses.In[5],we presented how the required functionality can be realized and implemented.We further proposed an extension to the IPFIX protocol that allows ex-porting aggregatedflow information in an efficient way.In the following subsections,our approach will be explained in more detail.2.3.Aggregation ProcessIn accordance with[16],the proposed aggregation func-tionality can be realized either as an additional process that is installed between the metering process(es)and the ex-porting process of a monitoring probe,or within a separate device that we call a concentrator(seefigure1).Such a concentrator receivesflow records from other monitoring probes,performs theflow aggregation,and exports the ag-gregatedflow information to an analyzer.Both options are depicted infigure2.In order to differentiate between regularflow records and records containing aggregated information,we intro-duce the termflow aggregate.In general,aflow aggregate comprises a set of individual IP-five-tupleflows that share some common properties,the so calledflow keys.The cor-respondingflow record contains information that identifies and describes theflow aggregate as a whole.Which and how individualflows are aggregated intoflow aggregates is specified by aggregation rules.An aggre-gation rule specifies whichfields(flow keys and non-flow keys)are kept and whichfields are discarded in order not to appear in the resultingflow record any more.For IP ad-dresses,a mask operation can be applied in order limit theexported monitoring data(IPFIX Protocol)EP: Exporting ProcessAP: Aggregation ProcessMP: Metering ProcessCP: Collecting Processexported monitoring data(IPFIX Protocol)exported monitoring data(IPFIX Protocol) Figure2.Architecture of IPFIX aggregation.Aggregation can be employed within a mon-itoring probe(left)or in form of a separateconcentrator(right)distinction between differentflow aggregates to network ad-dresses.Sometimes,the application of an aggregation rule is restricted toflows that match a given criteria.For ex-ample,to aggregateflows originating from a specific host or network,a selection pattern for the correspondingfield must be specified.Fields carrying measurement information are aggregated in a special way in order to get the corresponding values for theflow aggregate.As an example,packet and byte coun-ters of the individualflows are summed up,while start and end time of theflow measurement are set to the minimum and maximum of all aggregatedflows respectively.An aggregation rule defines the treatment of the consid-eredfields in separate lines according to the following struc-ture:1.Afield modifier(discard,keep,mask/n,or aggregate)specifies how thisfield is treated and implicitly defines if thefield appears in the resultingflow record or not;2.the consideredfield identified by the corresponding in-formation element ID[15];3.an optional pattern that restricts the aggregation toflows or packets that match this pattern.An aggregation rule is only applicable to an incomingflow or packet if all specifiedfields are present and if all indicated patterns match.Multiple aggregation rules can be specified to generate flow aggregates with different properties.Incomingflow records are usually checked against each rule and may thus contribute to variousflow aggregates at a time.In some cases,however,it is preferred that an incomingflow record that contributes to oneflow aggregate is excluded fromotherflow aggregates.This can be achieved by organiz-ing aggregation rules in chains or trees that define in which order the rules are to be checked.According to the rule se-mantic defined in[5],only thefirst matching rule in a chain or tree is applied.2.4.Aggregation ExampleWe demonstrate the principles and advantages of our flow aggregation mechanisms with the following aggrega-tion rule example:1Aggregation rule2discard protocolIdentifier in TCP3discard sourceTransportPort4mask/24sourceIpv4Address5discard destinationTransportPort in80,4436keep destinationIpv4Address in10.10.0.0/167aggregate packetDeltaCount8aggregate octetDeltaCount9aggregate flowStartMilliSeconds10aggregate flowEndMilliSecondsThefirst line indicates the beginning of the aggregation rule.Then(lines2and3),twofields(transport protocol and source port)are specified that have to be present in each flow record to be aggregated.However,the resultingflow record will not include thesefields because of the modi-fier discard.The pattern in TCP restricts the rule to TCPflows.The next line(4)requests the source address to be present in received records.This addressfield will be masked to24bit before aggregating(mask/24).The discard modifier is used again in line(5)but an addi-tional pattern is specified which must be matched by re-ceivedflows.Similarly,in(6)a pattern is provided but the keep modifier defines that thisfield appears in theflow record.Finally(6-10),the aggregate modifier is used to collect statistics of theflow such as the number of packet-s/octets and theflow start and end time.Figure3shows some non-aggregatedflow data.If the presented aggregation rule is applied,theflow aggregates shown infigure4will be created.Both,the number of flows as well as the number offields in eachflow is reduced. This reduction obviously depends on the kind of monitoring data(address/port distribution etc.)and the configuration of the aggregation rules.The proposed aggregation scheme provides aflexible mechanism to adapt the granularity and quantity of the aggregation depending on the particular sce-nario.Some sample measures are provided in section3. 2.5.Export of Flow Aggregate InformationSince IPFIX is a unidirectional protocol that does not allow querying specific information from the monitor-ingprobe,the exported monitoring data should be self-explanatory,i.e.the analyzer should be able to understandFigure3.Non-aggregatedflow data consist-ing of protocol information,source and desti-nation address,source and destination port,and statistical dataFigure 4.Resultingflow aggregates corre-sponding to the given rule setand interpret the received information without any addi-tional knowledge.In case offlow aggregates,the analyzer has to be informed about the corresponding aggregation rules.Otherwise,the circumstances under which theflow records were generated would not be clear.Especially,if aggregation rules contain patterns that restrict their applica-tion to a selection of incomingflow records,the analyzer has to be informed that the receivedflow information is biased.Moreover,if rules are organized in chains or tree as explained in the previous subsection,the order in which rules are checked also has to be disclosed.We solved this problem by specifying a new template type called data template.A data template describes the structure offlow records and includes additional informa-tion about the patterns used in the aggregation rules.In practice,a separate data template is used for each aggre-gation rule.A specialfield in the data template serves as a backward pointer to a preceding rule if the aggregation rules are organized in a chain or tree.Detailed information about the data template can be found in[5].3.Implementation and Evaluation3.1.Implementation in VermontWe used our modular IPFIX/PSAMP probe Vermont2to implement the presented aggregation mechanism.The re-sulting architecture of Vermont is shown infiing this architecture,it is possible to use the aggregation mecha-2Versatile Monitoring Toolkit,/CaptureP S A M PI P F I XI P F I XR a w P k t sN e t f l o w .v 9Figure 5.Architecture of Vermont including the aggregation modulenism in both ways as mentioned before:directly on a mon-itoring probe before exporting the flow data or in form of a concentrator that receives flow records (either Netflow.v9or IPFIX)and exports flow aggregate information.The first path is to capture packets using the pcap library,optionally filtering the collected data,and then sending the packets to the aggregation module using the ‘hook’as shown in the figure.Alternatively,a collector module can decode Net-flow and IPFIX data and put the received flow information directly into the aggregation module.The export of IPFIX conform flow data is provided by a separate exporter library.We verified the interoperability of Vermont with other IP-FIX/Netflow exporters and collectors at an IPFIX interop-erability event in 2005[18].3.2.Experimental AnalysisEven though the compression degree of the aggregation mechanism strongly depends on the properties of the mon-itored traffic as well as on the aggregation rules in use,we did some measurements in order to estimate the capabilities of our aggregation mechanism.The experiments were done using a standard PC (2.8GHz)running FreeBSD.Vermont was used for monitoring network data received at a dedi-cated mirror port of a Gigabit switch.The first experiment was done in front of our workgroup server where usually only small traffic volumes are observed (e-mail transfers and remote logins).The second experiment was done at the Internet gateway of the University of Erlangen (actually 622MBit/s).Both experiments lasted for 30minutes in or-der to keep the amount of monitoring data at a manageable size.We evaluated the aggregation efficiency using fivedif-+portsFigure 6.Aggregation ratio in comparison to flow accounting without aggregation.R1:workgroup server,R2:Internet connectionferent aggregation schemes:•Standard IPFIX export (IP-five-tuple flows)without aggregation (labeled IPFIX)•Aggregation of flows from the same source network (src/24)•Aggregation of flows from the same source net-work destined to the same destination network (src/24+dst/24)•Aggregation of flows without considering port infor-mation (ports)•Combination of the last two aggregation schemes (src/24+dst/24+ports)During the experiments,Vermont exported each record in an individual IPFIX packet,which is not very efficient with respect to encapsulation overhead but allowed us to determine the aggregation ratio directly from packet counts.The aggregation ratio of all mentioned aggregation schemes and for both experimental setups is shown in figure 6.Ob-viously,the aggregation of network addresses seems to be inefficient with respect to data compression (the aggrega-tion ratio is about 0.97).This effect occurs due to different values in the source and destination ports.If these values are ignored,the aggregation ratio increases (ports,approxi-mately 0.6),especially if combined with aggregation of ad-dress information (src/24+dst/24+ports).In the last case,we achieve an aggregation ratio of 0.48on our Internet con-nection and down to 0.09in the workgroup server scenario.Figure7.Rate offlow data as sent from themonitor to an analyzer:workgroup server(the average observed input rate is32pack-ets/s)The two plots shown infigures7and8depict the packet rate of the monitoring data sent to the analyzer.Bothfig-ures demonstrate the reduced amount of monitoring data sent from the monitoring probe to the analyzer.Not shown is the reduction of the load of the monitor itself.Because manyflows are mapped to a singleflow aggregate,the cor-responding hash table sizes are reduced.As a consequence, the memory requirements of the monitoring probe can be decreased and the search for existingflow records requires fewer steps,such fewer CPU cycles.3.3.Application ScenariosIn this subsection,we show how the aggregation mech-anism can be practically deployed in three different scenar-ios:accounting,anomaly detection,and TCP SYNflood detection.Typical accounting solutions intend to associate con-sumed bandwidth or time to individual end systems.Ex-act descriptions about all monitoredflows are not required. Therefore,the following rule set is appropriate to aggre-gate transmitted octets from/to each host in the monitored network(10.10.0.0/16)resulting in twoflow aggregates for each host(incoming/outgoing):1Aggregation rule2keep sourceIpv4Address in10.10.0.0/163aggregate octetDeltaCount4aggregate flowStartMilliSeconds5aggregate flowEndMilliSeconds6Aggregation rule7keep destinationIpv4Address in10.10.0.0/168aggregate octetDeltaCount9aggregate flowStartMilliSeconds10aggregate flowEndMilliSecondsFigure8.Rate offlow data as sent from themonitor to an analyzer:Internet connection(the average observed input rate is58’500packets/s)Using this rule set,the aggregation process provides an efficient preprocessing of the monitored data.Therefore,flow aggregation also reduces the computational resources at the analyzer,i.e.the accounting process in this scenario.Attack and anomaly detection have different objectives. Usually,the general behavior of all observedflows is con-sidered and not individualflows.For example,the total number of observed packets might be of interest(rule1), the number of monitored ICMP packets(rule2),the dis-tribution of traffic to different subnets(rule3,4),and the number of packets sent to request a specific service,e.g. web services(rule5).1Aggregation rule12aggregate packetDeltaCount3aggregate flowStartMilliSeconds4aggregate flowEndMilliSeconds5Aggregation rule26discard protocolIdentifier in ICMP7aggregate packetDeltaCount8aggregate flowStartMilliSeconds9aggregate flowEndMilliSeconds10Aggregation rule311mask/24sourceIpv4Address in10.10.0.0/1612aggregate packetDeltaCount13aggregate flowStartMilliSeconds14aggregate flowEndMilliSeconds15Aggregation rule416mask/24destinationIpv4Address in10.10.0.0/16 17aggregate packetDeltaCount18aggregate flowStartMilliSeconds19aggregate flowEndMilliSeconds20Aggregation rule521discard destinationTransportPort in80,44322aggregate packetDeltaCount23aggregate flowStartMilliSeconds24aggregate flowEndMilliSecondsSimilarly to the accounting scenario,preprocessing is provided by the aggregation process.The resultingflowdata can be directly feed into the corresponding anomaly detection engines of the attack or intrusion detection sys-tem.In the European project Diadem Firewall3,we employ the following rule set on a monitoring probe for count-ing TCP SYN packets per destination and TCP SYN/ACK packets per source:1Aggregation rule2discard protocolIdentifier in TCP3keep destinationTransportPort4keep destinationIpv4Address5discard tcpControlBits in SYN6aggregate packetDeltaCount7aggregate flowStartSeconds8aggregate flowEndSeconds9Aggregation rule10discard protocolidentifier in TCP11keep sourceIpv4Address12keep sourceTransportPort13discard tcpControlBits in SYN,ACK14aggregate packetDeltaCount15aggregate flowStartSeconds16aggregate flowEndSecondsNote that although protocol and TCP control bitfields are discarded from theflow records,the values of the pat-terns(TCP,SYN,SYN/ACK)are exported with the data templates.The resulting packet counters serve as input to a detection algorithm for TCP SYNflood attacks[20].Under normal conditions,every SYN packet sent to a certain des-tination should be answered by a SYN/ACK packet,apart from some SYN packets that are directed to closed ports or non-existing hosts.During a TCP SYNflood attack,many SYN packets are directed to a single open port at the vic-tim host.Returned SYN/ACK packets remain unanswered resulting in many half-open TCP connections.A victim host without specific protection mechanism is able to han-dle a limited number of half-open TCP connections only.If this limit is reached,new SYN packets are not answered by SYN/ACK packets any more,which is reflected by a high difference in the exported counter values.The number of flow records resulting from the above rule set depends on the number of attacked(address,port)pairs but not on the number of attack sources.This is important because address spoofing is usually applied in a SYNflood,causing a huge number of different IP-five-tuples.4.Discussion and ConclusionsIn this paper,we proposed an aggregation mechanism for efficient andflexible use inflow accounting scenarios.This mechanism is controlled by aggregation rules that allow adapting to particular application requirements.We spec-ified the necessary functionalities as enhancements of the 3Diadem Firewall is partly funded by the European Commission(FP6 IST-2002-002154).Homepage:http://www.diadem-fi IPFIX architecture and protocol.The aggregation mecha-nism can be deployed in two ways:either at the monitoring probe to reduce the number offlow directly at the observa-tion point,or in a separated device called concentrator.The second option allows merging and aggregatingflow infor-mation produced at different observation points in the net-work into a single stream that is sent to the analyzer.The aggregation mechanism supports multiple active aggrega-tion rules at a time as well as patterns that provide afiltering functionality.We exemplarily evaluated the aggregation mechanism by using an implementation of the aggregation module for the monitoring toolkit Vermont.The experiments shows the data compression capabilities of different aggregation rules. Furthermore,we discussed three application scenarios and presented appropriate aggregation rule sets.References[1]K.G.Anagnostakis,S.Ioannidis,tchev,J.Ioannidis,M.Greenwald,and J.M.Smith.Efficient Packet Monitoring for Network Management.In8th IEEE Network Operations and Management Symposium(NOMS),Florence,Italy,Apr.2002.[2] B.Claise.Cisco Systems NetFlow Services Export Version9.RFC3954,Oct.2004.[3] B.Claise.IPFIX Protocol Specification.Internet-Draft,work in progress,draft-ietf-ipfix-protocol-22.txt,June2006.[4]L.Deri.nProbe:an Open Source NetFlow Probe for GigabitNetworks.In TERENA Networking Conference(TNC2003), Zagreb,Croatia,May2003.[5] F.Dressler,C.Sommer,and G.M¨u nz.IPFIX Aggrega-tion.Internet-Draft,work in progress,draft-dressler-ipfix-aggregation-03.txt,June2006.[6] B.Fortz,J.Rexford,and M.Thorup.Traffic Engineeringwith Traditional IP Routing Protocols.IEEE Communica-tions Magazine,40(10):118–124,Oct.2002.[7] F.Ghys and ponent-Based Charging ina Next-generation Multimedia Network.IEEE Communica-tions Magazine,41(1):99–102,Jan.2003.[8]N.Hohn and D.Veitch.Inverting sampled traffic.In3rd ACM SIGCOMM Conference on Internet Measurement, pages222–233,Miami Beach,FL,USA,Oct.2003.[9]Y.Hu,D.-M.Chiu,and J.C.Lui.Adaptive Flow Aggre-gation-A New Solution for Robust Flow Monitoring under Security Attacks.In IEEE/IFIP Network Operations and Management Symposium(IEEE/IFIP NOMS2006),pages 424–435,Vancouver,Canada,Apr.2006.[10]R.Kemmerer and G.Vigna.Intrusion Detection:A BriefHistory and Overview.IEEE Computer-Special Issue on Security and Privacy,pages27–30,Apr.2002.[11]mpert, C.Sommer,G.M¨u nz,and F.Dressler.Vermont-A Versatile Monitoring Toolkit for IPFIX and PSAMP.In IEEE/IST Workshop on Monitoring,Attack Detection and Mitigation(MonAM2006),Tuebingen,Ger-many,Sept.2006.。