Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=tcad20

Download by: [University of Alberta] Date: 05 November 2015, At: 11:41

Computer-Aided Design and Applications

ISSN: (Print) 1686-4360 (Online) Journal homepage: http://www.tandfonline.com/loi/tcad20

CAD/CAE associative features for cyclic fluidcontrol effect modeling

Lei Li & Yongsheng Ma

To cite this article: Lei Li & Yongsheng Ma (2015): CAD/CAE associative features forcyclic fluid control effect modeling, Computer-Aided Design and Applications, DOI:10.1080/16864360.2015.1084190

To link to this article: http://dx.doi.org/10.1080/16864360.2015.1084190

Published online: 28 Sep 2015.

Submit your article to this journal

Article views: 9

View related articles

View Crossmark data

http://www.tandfonline.com/action/journalInformation?journalCode=tcad20http://www.tandfonline.com/loi/tcad20http://www.tandfonline.com/action/showCitFormats?doi=10.1080/16864360.2015.1084190http://dx.doi.org/10.1080/16864360.2015.1084190http://www.tandfonline.com/action/authorSubmission?journalCode=tcad20&page=instructionshttp://www.tandfonline.com/action/authorSubmission?journalCode=tcad20&page=instructionshttp://www.tandfonline.com/doi/mlt/10.1080/16864360.2015.1084190http://www.tandfonline.com/doi/mlt/10.1080/16864360.2015.1084190http://crossmark.crossref.org/dialog/?doi=10.1080/16864360.2015.1084190&domain=pdf&date_stamp=2015-09-28http://crossmark.crossref.org/dialog/?doi=10.1080/16864360.2015.1084190&domain=pdf&date_stamp=2015-09-28

COMPUTER-AIDED DESIGN & APPLICATIONS, 2015http://dx.doi.org/10.1080/16864360.2015.1084190

CAD/CAE associative features for cyclic fluid control effect modeling

Lei Li and Yongsheng Ma

University of Alberta, Canada

ABSTRACTNowadays, CAE has been widely applied in industry as an auxiliary tool for decision making. How-ever, human intervention still is required in reasoning CAE results and in determining design changeeffect. Ideally, CAD and CAE should be seamlessly integrated where design changes, effect anal-ysis and function reasoning should be explicitly supported by automatic feature extraction in acyclic manner. This paper presents two CAD/CAE associative feature concepts, which form a robustmechanism aiming to support automatic CAD/CAE interactions with both geometric and seman-tic information. Firstly, a concept of CAE boundary feature is proposed to model and record thesimulation intents, and therefore, the CAE analysis cycles could be managed with the consistent ini-tialization setups, e.g. boundary conditions, mesh types, and physical models. Secondly, CAE resultsare processed by extracting the differential characteristic features, which are named as CAE effectfeatures. An effective feature reflects the sensitivity information related to each engineering changeand hence leads to the evaluation of design modifications. Consequently, the cyclic design modifi-cations can be tracked and reasoned; then further modifications are guided to converge into theoptimum. The research innovation lies in the proposed CAD/CAE integration scheme which hasbeen explored preliminarily to demonstrate an automated and efficient way to keep the CAD/CAEsemantic consistency over cycles of optimization. A case study about hydraulic valve developmentis presented.

KEYWORDSCAD/CAE associative feature;CAE boundary feature; CAEeffect feature; designautomation; CAD/CAEintegration

1. Introduction

CAD/CAE integration remains to be a demandingresearch topic in the past two decades, as interfacing sim-ulation and design tools for intelligent product develop-ment is still widely recognized as a technical gap withouta general solution. Although a number of approacheshave been studied, seamless CAD/CAE integration hasnot been fully realized. The majority of efforts focusedon the front process of preparing the CAD model forCAE analysis effectively and efficiently. The close loopCAD/CAE interactions remain to be a research issue, andthere are mainly two difficulties: how to synchronize theCAD and CAE models and how to interpret CAE resultsfor design modification evaluation. In order to overcomethese difficulties, this paper presents two new concepts.For the former difficulty, a CAE boundary feature con-cept is proposed to manage the geometric and semanticassociations between CAD and CAE models based onthe well-established associative feature concept [17]. ThisCAE boundary feature is defined as a software objectclass and its application is put forward as a robust tool tomaintain analysis setup information consistency during

CONTACT Yongsheng Ma yongsheng.ma@ualberta.ca

the cyclic CAD/CAE mesh information conversion. Forthe latter difficulty, another class, referred to as CAEeffect feature, is introduced by extracting the sensitiv-ity information from two consecutive CAE run results;such extracted information are necessary for identify-ing proper designmodification direction. By introducingthese concepts, a new CAD/CAE integration frameworkhas been developed which covers both the forward andreverse integrations and supports more automated cyclicproduct development. Before getting into the details, abrief literature review about CAD/CAE integration ispresented below.

From early 1980 s to 2000, CAD and CAE researchersfocused on the geometry conversion and simplification.During that period, semantic information related to theCAE geometry meshes was missed during the conver-sion, and redundant efforts were needed to recover thelost information, i.e. the boundary conditions. There-fore, the efficiency is low, especially for cyclic designprocesses.

To make up this deficiency, there are two commonapproaches proposed: one is to develop an integrated

2015 CAD Solutions, LLC, http://www.cadanda.com

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

http://orcid.org/[0000-0002-5259-407X]http://orcid.org/[0000-0002-6155-0167]mailto:yongsheng.ma@ualberta.cahttp://www.cadanda.com

2 L. LI AND Y. MA

system with both CAD and CAE modules, for whichthe feature-based techniques are supposed to maintainthe semantic information during model conversion; theother is to develop a unified feature model incorpo-rated with both CAD and CAE information, with whichboth CAD and CAE views could be extracted and theinformation consistency could be easily sustained [2].

For the first approach, Kao et al. [13] introduce anintegrated system to generate thread rolling die-plategeometry. The design parameters are linked to Solid-Works through Microsoft Excel. Matin et al. [19] putforward a feature-based CAD/CAE integration systemfor mold design. With the commercial CAD softwarePro/Engineer, the specific mold design module guaran-tees rapid mold modification. The simulation moduledetermines the injection molding parameters. Lin andKuo [15] build an integrated CAD/CAE/CAM systemfor automobile stamping die development. In this sys-tem, STRIM is used as 3D surface construction CADsoftware, CATIA as CAD/CAE software, DYNAFORMas stamping formability analysis software andCADCEUSas CAM software. Johansson [12] develops a prototypesystem to integrate SolidWorks, ANSAandLS-Dyna sim-ulating the behavior of ski-racks during car collision. Forthe latter, Deng et al. [5] propose the CAD/CAE featureconcept to incorporate both CAD and CAE design infor-mation. CAD/CAE features are related to both design andanalysis processes. Geometric information from CADfeatures is assigned to CAD/CAE features. The designintent derived from analysis constitutes the CAE por-tion of the CAD/CAE features. The integration can beestablished once all the CAD/CAE features are created.Lee [14] develops a feature-based single model for CADandCAE integration. It is unique for themulti-resolutionand multi-abstraction modeling techniques making itcapable for a wide range of applications. Cuilliere andFrancois [4] establish a unified topological model to inte-grate CAD, FEA and topology optimization. This origi-nally developed environment and database organizationenables multi-source model utilization, automatic mesh-ing, reanalysis corresponding to remeshing, and topologyoptimization.

CAE analysis is carried out to check whether thedesign is satisfactory. If not, reasoning of the analy-sis result is significant in guiding the following designmodifications. This actually forms the reverse processof CAD/CAE integration. Different methods are avail-able to properly interpret the analysis result. Merkel andSchumacher [20] utilize response surface methods inthe CAE driven product development. Park and Dang[21] use commercial CAD and CAE software construct-ing an integrated system to achieve structural optimiza-tion with metamodeling techniques including response

surface method and radial basis function. Robinsonet al. [22] propose a method to obtain optimized designby adding new parameters or features to CAD modelincrementally.

For cyclic CAD/CAE integrated design process,CAD/CAEmodel synchronization is also important. Forthis purpose, the associative feature concept proposedby Ma and Tong [17] is the right mechanism to real-ize the seamless synchronization. An associative featureis defined as a set of semantic relationships which canbe both geometric and non-geometric among differentgeometries or applications. Associative feature is a newconcept which distinguished from traditional form fea-tures [18]. Associative feature can not only manage fixedrelations but also the developing ones which are derivedfrom them. Features which are based on volume of mate-rial are just specific types of associative features. Moreimportantly, different from form features, associative fea-tures can be independent of volume. This is the keycharacteristic explaining the reason why associative fea-tures can also reveal and manage the relations embeddedin the application system.

There have been commercial software tools provid-ing integration support. For example, SolidWorks has aComputational Fluid Dynamics (CFD) module embed-ded in the software environment. ANSYS Workbench iscapable to carry out modeling, meshing, simulation andoptimization in a single environment. TOSCA is inte-grated in this environment as the optimization module[1]. It has a wide range of application in solid and fluidmechanics. The optimization result is remeshed auto-matically to get validated. However, the CFD module inSolidWorks is not comprehensive due to its limitationin mesh type and solver. The usability of the modellerin ANSYS Workbench is not sufficiently powerful andflexible. Moreover, the optimization process is not fullyautomated. In these commercial and integrated pack-ages, high-level feature information is still not explicitlymanaged in the CAD/CAE interaction conversion cycles[25]. In general, there are some limitations both in theresearch and application of CAD/CAE integration. CAE-oriented design information cannot be fully described.The CAE calculation accuracy which has not obtainedenough emphasis is actually a critical factor in the forma-tion of integration. In addition, few researches are con-ducted in the fluid domain. As a result, further researchon this issue is needed.

The rest of this paper is organized as follows. The newfeature concepts are introduced in Section 2 to facilitatethe CAD/CAE integration. Based on the features broughtforward, Section 3 explains the mechanism how CADand CAE integrate in detail. A case study of hydraulicvalve is illustrated in Section 4 to show the effectiveness

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

COMPUTER-AIDED DESIGN & APPLICATIONS 3

of thismethod. At last, conclusions towards the presentedwork are made.

2. Feature concepts in CAD/CAE integration

As mentioned earlier, associative feature is capable ofestablishing and managing both geometric and non-geometric associations. Therefore, associative featureconcept is used in this work to interface the CAD andCAE tools, which synchronizes the different applicationmodels and guarantees the data consistency. The overallintegration scheme is shown in Fig. 1.

The design flow starts from the conceptual designwhich preliminarily satisfies the design requirements.Given the optimality, the conceptual design is still imma-ture and requires the simulation-based design modifica-tions. The conceptual design model conveys both geo-metric and non-geometric information; therefore, it issignificant to guarantee the complete information trans-fer between CAD and CAE tools. The CAD/CAE asso-ciative feature plays the roles in fulfilling this seamlessintegration. To be specific, all geometric entities from theCAD model and analysis model employ the one-to-onecorrespondence, and the semantic information useful forCAE analysis is also linked to analysis model. In this way,the CAE analysis setup could be automatically completedwithout redundant model preparations.

The accuracy of calculation is critical in the cyclicCAD/CAE integration because the analysis depends onthe previous calculation. Proper application of bound-ary conditions is very important to obtain accurateresult, because they will not only affect the type of mesh

generated but also the setup of the solver. Many researchworks are dedicated in the integration of CAD and CAE,but rarely focus on the accuracy. However, if the resultis not valid, the integration would be meaningless. As aresult, the CAE boundary feature is proposed to managethe relations among the geometry, boundary conditionsand specific mesh type. Fig. 2 shows the semantic asso-ciations in this CAD/CAE integration process. Based onthe associative feature concept [17], CAE boundary fea-ture can be defined as a class of features that contains themapping relations of geometrical dependencies betweenCAD entities and their associated CAE mesh repre-sentations, e.g. grids and tetrahedrons, as well as non-geometrical dependencies, such as inherited properties,like fluid properties, fluid space body face names, tags,constitutional structures, and conceptual constraints toapply CAE boundary conditions. For example, in compu-tational fluid dynamics, the velocity and pressure bound-ary are assigned to the inlet or outlet of the geometryaccording to the application situation. The walls formedby the faces of the fluid body geometry are subject tono-slip boundary condition. Particularly, mesh inflationshould be applied along solidwalls to pursue higher accu-racy because it is capable of estimating the steep gradientsin the boundary layer.

Generally, if the number of the meshes is increased,the computational costwill be higher accordingly.Here inour presented approach, the accuracy is not achieved bythe number of the meshes, but the type of meshes. As weused FluentTM, mesh inflation is applied along the wallboundary in order to obtain more accurate result. As aresult, this approach can achieve higher accuracy without

Figure 1. CAD/CAE Integration Scheme.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

4 L. LI AND Y. MA

Figure 2. Semantic Associations in CAD/CAE Integration.

reducing computational speed. This is because of the useof inflation mesh type, which is well-supported by theCAE boundary feature which maintains the consistencyduring iterations. The reason why we address the impor-tance of accuracy is that the optimal design will not beachieved without rounds of accurate iterations. The pro-posed approach provides an efficient way to increase andmaintain the CAE accuracy.

Smit and Bronsvoort [23] suggested an analysis viewwhich is an interface concept to interact withmesh gener-ation, boundary conditions, analysis model, and solutionmethods in a multi-view feature modeling environment.In the proposed approach, CAD model and CAE modelare associated through the CAE boundary feature. Incomparison, the CAE boundary feature expands its asso-ciations to the design model and the following optimiza-tion process in the integration loop. It is a robust tool tomaintain information consistency during the iterations.

Currently, human intervention in the CAE analysisstill plays an important role in decisionmaking. To obtaina better design, specialist will usually make the decisionon the modification and then process the change. Thisis a tedious process because the CAD model may bemodified for hundreds of times and CAE analysis should

be conducted accordingly. In order to realize automaticinterpretation of the CAE result, the concept of CAEeffect feature is proposed here.

As shown in Fig. 2, CAE effect feature is also defined asa class of features that represents the unique characteris-tics of interestedmeasure changes for a physical behaviorin the context of a CAE analysis scope; in other words,its applied instances explicitly express the influence onthe physical performance of a defined function due tothe incremental concept changes in the associated CADmodel. This concept is of significant importance to formthe loop of CAD/CAE integration. After the analysis ofthe initial design, the system will attempt a method forthe modification of design parameters. Then, the CADmodel and CAE scenario will be updated synchronously,resulting in newCAE analysis results. CAE effect featureswill be extracted from the difference between the newanalysis results and the previous ones. This cyclic processwill be conducted in an iterative way. As a result, moreCAE results will be obtained from different design mod-els and corresponding CAE scenarios, which contributeto find the trend of design propagation. With the set ofCAE effect features extracted, the interpretation can becarried out and the optimized design would be possibly

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

COMPUTER-AIDED DESIGN & APPLICATIONS 5

achieved with the help of physical modeling. In addi-tion, it should be noticed that engineers can still intervenethe system after the CAE effect features extraction. Basedon engineering knowledge, they can check whether themodification is promising and determine the next step ofmodification.

In summary, Fig. 2 expresses the functions of CAEboundary feature and CAE effect feature propertiesexplicitly. Therefore, the changes in CAD design can bereflected in CAE scenarios accurately. CAE effect featureis a powerful tool to manage the effect caused by differ-ent design regardless of the modification sequence. CAEeffect feature has a flexible data structure making it capa-ble to realize decision making. Engineers can interposethe justification based on their knowledge. By optimiza-tion method, design parameters will be updated until anoptimal design is obtained. Meanwhile, all the featuresof the design model can be updated synchronously. Theformation of this loop is the foundation for CAE effectfeature extraction and design optimization.

3. Integration of CAD and CAE

The mechanism of CAD/CAE integration is to be intro-duced in this section based on the background of fluiddynamic analysis. Because the hydraulic system is highlystandardized, the components such as control valves andpumps can be modeled by feature-based and parametricapproach. Parametric modeling can be easily achieved bythe built-in expression function library in aCADpackagelike SolidWorksTM and NXTM. By introducing appropri-ate constraints, a parametric model can be constructedin such a way it can be readily integrated with automatedoptimization loops [6].

Fluid space is abstracted from the established geom-etry by Boolean operations. Because face IDs in CADsystem may differ from that in another system [24], theIDs of fluid space faces will be assigned specific tags withattributions and boundary conditions attached. The tagis an identifier which can be recognized by both CADand CAE systems. It works as part of the CAE inter-face protocol. The information of entities with tags isstored in database for later processes. Tab. 1 shows themechanism how the information is transmitted betweendifferentmodules, inwhichm, n, p and q are the numbersof corresponding faces in CAD model.

Table 1. Information Transmission in CAD/CAE Conversion.

Tag Attribute Boundary condition

I1, I2, . . . . . . , Im Inlet Velocity or pressure inletO1, O2, . . . . . . , On Outlet Velocity or pressure outletW1, W2, . . . . . . , Wp Wall No-slip wallS1, S2, . . . . . . , Sq Symmetrical plane Symmetry

Based on the design of a control valve, Fig. 3 shows thedetailed CAD/CAE integration mechanism. The valvecan be parameterized by the dimensions in both axialand radial direction. The diameter of the valve core deter-mines the scale of the fluid space formed by the valve coreouter face, valve sleeve face and valve body inner face.Meanwhile, according to empirical formula, valve corediameter is decided by many factors. Such kind of infor-mation forms forward physical reasoning, which will beuseful in the optimization process. The majority of fluidspace faces are subject to the no-slip boundary conditionwhich is defined as walls. In order to reduce the compu-tation load, only half of the model is calculated due tothe symmetric property of geometry. Not all of the CAEfeatures are geometrical features, such as the parametersof the hydraulic oil and the value and type of boundaryconditions.

After all the parameters defined and geometry estab-lished, CAE boundary features will be established tolink geometry characteristic faces of the fluid body,meshes, corresponding boundary conditions and the spe-cific mesh generation method. The fluid body geometricfaces such as inlets, outlets, walls and symmetrical planeswith unique tags will be indexed from the database andassigned nameswith the type of boundary. Later inmesh-ing stage, CAE boundary features also direct the meshgeneration and refinement. For example, mesh inflationis applied along the wall boundary in order to obtainmore accurate result. CAE Solver invoked by the CAEmodule will automatically recognize the boundary withsuch kind of names and the assigned boundary con-ditions correspondingly if applicable. As a result, afterthe CAD model is updated, the CAE model, includ-ing both geometric and non-geometric information, canbe synchronized accordingly. In this process, the non-geometrical parameters remain unchanged. CAD/CAEfeature information sharing is achieved by the associa-tions embedded within the CAE boundary feature.

After the mapping from CAD module to CAE mod-ule, the meshing of fluid space should be conducted. Thisis a key process in this cycle because it will highly affectthe accuracy of result which is a critical factor in theintegration. Some detailed features may greatly increasethe time required for meshing and Finite Element Anal-ysis (FEA). Even, they may lead to the failure in meshgeneration [26], which will end up the whole process. Itis important to remove the irrelevant geometric detailswithin the error limit providing precondition for steadycalculation. Estimating the effects of removing negativefeatures in engineering analysis can be achieved [16]. Fer-randes et al. [8] put forward a method not only evaluatethe error of simplification but also repair the model if thesimplification triggers unacceptable error. Simplification

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

6 L. LI AND Y. MA

Figure 3. CAD/CAE Integration Mechanism.

features concept [11] was proposed to identify the detailsto be simplified and maintain consistency between CADmodel and FEA model. The assessment and control ofsimplification will increase the robustness, accuracy andefficiency in cyclic integration.

Under the control of CAE boundary feature, the fluidflow space is meshed. Each of the CAE mesh cell asso-ciated with the local fluid attributes, is modeled as anobject; the data type, such as the generic data structureof a mesh element associated with the local fluid flowproperties, is defined as an object class, named as a par-tial seed element in this proposed framework. The entireflow space with partial seed elements is processed by FEAin the solver. Finite Volume Method (FVM) is appliedbecause it is usually the most efficient for flow simu-lations comparing to Finite Difference Method (FDM)and Finite Element Method (FEM). The analysis resultsare checked to see whether any of the constraints inthe design requirements is violated. Optimization will beneeded if the original design does not meet the require-ment. Both of the CAD and CAE features will be mod-ified after the modification of the design parameters,such as the valve core diameter. The change of designparameters will be reflected on the fluid space directlydue to the associations. Subsequently, the remeshing offluid space will be carried out. The introduction of CAE

boundary feature is useful for the cyclic analysis setupbecause it enables consistent reassignment of boundaryconditions to the mesh geometry automatically in eachiteration after remeshing. For the model which remainsunchanged, adaptive remeshing strategies [10] are usedin CFD to control the error level on given numericalsolution by updating mesh according to the error estima-tor. For modified model, using the prior existing meshes,efficient remeshing can be done [27]. This approach pro-vides a robust way to regenerate the mesh after geome-try modification. A 3D automatic remeshing method [9]is well suited for remeshing models with small designparameters change.

To facilitate the measurement of flow field compari-son before and after an incremental change of design, thestates of the consecutive flow fields have to be memo-rized. The state of flow field before an interested designchange is introduced and stored as a reference state witha mesh space associated with a matrix of flow proper-ties. This reference flow state is called persistent flowspace with partial seed elements (PFSPSE). Then thenew simulated results, which are reflected by the newstate of flow field with another associated mesh space, isnamed as current flow space with partial seed elements(CFSPSE). Ideally, these two states are comparable auto-matically, because by updating the PFSPSEwith the input

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

COMPUTER-AIDED DESIGN & APPLICATIONS 7

of CFSPSE, cyclic CAD design change effect could beconsistently represented. PFSPSE can also be understoodas the pre-defined reference state, while CFSPSE is gener-ated according to the current design through the actionsof solver in each design iteration.

Theoretically, by working out the difference betweenPFSPSE and the updated CFSPSE, physical sensitivityanalysis could be carried out further, which potentiallyprovides the direction of design changes after the nec-essary human or artificial intelligent evaluation. Hencethe authors define the concept of CAE effect features,which are the CAE flow field differences of different

rounds of analysis results; they will reveal the sensitivitytowards design changes. This will provide the backwardphysical reasoning. Hence, by reasoning of a physicalphenomenon, engineers can figure out the method ofmodification and thus make progress for design opti-mization iteration by iteration.

It can be appreciated that the extraction of CAEeffect feature needs comparable mesh grid distributionsbetween the two checking simulated flow fields. How-ever, the mesh grids are commonly redistributed aftereach rounds of remeshing. To have a unified field com-parison reference mesh model, a mesh grid mapping

Figure 4. Grid-mapping: (a) Grid-mapping and Property Propagation (b) Grid Scaling, and (c) Grid Scaling and Interpolation.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

8 L. LI AND Y. MA

method is proposed in this work which is illustrated inFig. 4. The mesh grids generated in CFSPSE accordingto the mesh distribution and research interest. PFSPSEcan be updated by CFSPSE in a controlled manner. Theflow properties associated with the grids in CFSPSE andPFSPSE can be compared and unified through the so-called grid-mapping mechanism.

As shown in Fig. 4(a), when the changes of designare minor and the fluid space change does not causetopological change of the fluid space mesh, then by pat-tern comparison approach, flow states can be measuredquantitatively. To support such point-to-point compari-son, the grids in the PFSPSE and the CFSPSE must bereasonably corresponding to each other; that means theyneed to be consistently mapped. This research proposestwo types of grid mapping techniques, scaling and inter-polation. As shown in Fig. 4(b), a mesh unit edges in theCFSPSE can bemapped by the scaling factors in both hor-izontal and vertical directions to compare with PFSPSE;and vice versa. Further, in the case of mesh refinementoccurred automatically triggered by the mesher or intro-duced by the user, as shown in Fig. 4(c), interpolation ofgrids with the same topological pattern can be workedout to achieve the common and yet the refined meshgrid matrix. Comparing the flow field of CFSPSE withthe persistent flow field referencemodel which is the flowfield of PFSPSE, fluid flow changes can be quantitativelyrevealed by mesh grid-based effect matrix creation. If theuser allows, both the mesh of PFSPSE and the persistentflow field reference model can be updated according tonew analysis result, thus providing reasonable referencein the next iteration. Overall, the CAE effect features sup-port the CAE to CAD change processes and further helpto establish theCAD/CAE integration loop inwhich con-sistency maintenance [3] has to be used to update the

feature model in different view. The process iterates untilan optimal design is obtained.

The prospective system framework is shown in Fig. 5.This proposed system mainly consists of three modules,namely CAD module, CAE module and post-processingmodule. CAD and CAE modules can be from differentcommercial software tools which should be capable ofcomplex geometry and analysis. Most commercial CADsystems employ design by feature (DBF) method to buildmodel [7]. DBF is a paradigm in which the geometryis established using features. Typically, there is a featurelibrary in the CAD module to support model creationwith DBF. CAD module and CAE module are integratedby associative features [17].

Under the associative feature concept [17], CADmodel is mapped to analysis model which can be pro-cessed byCAEmodule. CorrespondingCAE features willbe extracted to prepare themodel for the following calcu-lation in the solver. In the post-processing module, CAEeffect feature are extracted with different CAE resultsfrom CAE module. Physical reasoning gained from CAEeffect feature is sufficient for the optimization of designparameters which will be updated in CADmodule. Userscan get access to all the modules and functions throughuser interface.

We acknowledge that not all the features identified bythe commercial software represent an interest of anal-ysis. As aforementioned, CAE features are predefinedwith clear semantic and ontological structures and geo-metric constituents. With such generic CAE features, abi-directional reasoning mechanism is proposed. Underthis scheme, the forward reasoning proposes potentialfeaturemodification options based on empirical formula.Correspondingly, the backward reasoning responds themodification with the changing effect derived. In this

Figure 5. Prospective CAD/CAE Integration System Framework.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

COMPUTER-AIDED DESIGN & APPLICATIONS 9

way, the sensitivity towards the proposed design changeis obtained. This mechanism could be the criteria toidentify the feature to be modified.

4. Case Study

The valve design case is studied in this section to provethe effectiveness of our proposed method. The concep-tual design is shown in Fig. 6(a), forwhich the valve sleeveand the valve core are coaxially assembled into the valvebody cavity. Functionally, the valve core reciprocates tocontrol the flow loop and rate, and the valve sleeve isapplied for replacement purpose because of the excessivewear. The CAD system applied here is Pro/ENGINEERwildfire.

Practically, the erosion caused by excessive hydraulicoil velocity will be significant. As a result, after the designprocess, analysis on the flow field inside the valve shouldbe conducted to check whether the maximum velocityexceeds the allowance. Here, the P-A valve chamber isstudied, which has a high flow rate as it controls themovement of the actuator. In order to process the FEA,the fluid space should be extracted. By using Booleanoperation, the fluid space could be obtained with tagsassigned, which is shown in Fig. 6(b). It is worth notingthat tags with attribute of wall will be assigned auto-matically after the setup of tags I, O and S because themajority of the faces are wall boundaries. Obviously, afterthe substantiation of the fluid space, it shares the sameface formed by the inner face of valve body, the faceof the valve sleeve and the outer face of the valve core.Hence, geometrically, CAD features andCAE features areassociated intimately.

In this case, the fluid dynamic analysis software Flu-ent is applied. The mesh shown in Fig. 6(c) is generatedby ANSYS Meshing module. Inflation mesh distributes

along the wall boundary to increase the simulation accu-racy in the corresponding areas. The overall informationof this model is listed as follows. Valve core diameter: 110mm; Medium:

Type: Hydraulic oilDensity: 870 kg/m3

Dynamic viscosity: 0.04 kg/ms; Analysis type: Velocity analysis; Boundary conditions:

Inlet P: Hydraulic oil velocity 16m/sOutlet A: Hydraulic oil pressure 20MPa;

Objective: Maximum hydraulic oil velocity < 27m/s.Fig. 7(a) shows the result of the FEA. It is obvious that

the maximum velocity is approximately 34m/s, whichexceeds the limitation. This result indicates that with thespecified design parameters, the valve could not providethe desired flow rate with a proper maximum hydraulicoil velocity inside the valve. Hence, further modificationis needed to meet the requirement.

According to empirical formula, the valve core diame-ter is determined by flow rate. This provides the forwardphysical reasoning for optimization. The modification ofvalve core diameter may be an effective method to opti-mize the design. In order to support the consistent designthroughout the concurrent engineering life cycle, fluidspace faces should be associated with the faces of valvebody, valve sleeve and valve core. If the design parametersare modified, the fluid space can be updated accordingly.With the single change in the diameter of the valve core,the corresponding modifications on CAD and CAE fea-tures are supposed to be conducted automatically. Conse-quently, the inconvenience induced by the modificationon the model will be eliminated.

Extracting the CAE effect feature is an important pro-cedure to get an optimal design. The valve core diameter

(a) (b) (c)

T: Return outlet; B: Inlet B; P (I1): Pressure Inlet tagged #1; A (O1): Outlet A tagged #1; S1: Symmetry plane tagged #1; Wi: The ith indexed wall

Figure 6. Forward Integration Process: (a) Valve Structure, (b) Fluid Space Abstraction (Sectional View), and (c) Mesh Generation.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

10 L. LI AND Y. MA

(a) (b)

Figure 7. Velocity Distribution: (a) Original Design, and (b) Modified Design.

(a) (b)

Figure 8. Maximum Velocity: (a) with Different Valve Core Diameter, and (b) with Different Valve Sleeve Groove Diameter.

is increased by 1mm per step. The velocity distribu-tion after the first valve core modification is shown inFig. 7(b). The maximum velocity appears to be 32.9m/s,which decreases in comparison with the original design.It can be concluded that the change of valve core diame-ter associates with many other design features and CAEfeatures, thus definitely affects the valve performance. Itis a reasonable direction to increase the diameter fur-ther to see the developing trend. After 9 iterations, themaximum velocities with different valve core diametersare obtained. The maximum hydraulic oil velocity of thisiterative process is shown in Fig. 8(a).

Based on engineering knowledge, the increase of valvecore diameter is at a cost of material and dynamic per-formance. In addition, the decreasing rate of maximum

velocity is not that fast with the increase of valve corediameter. As a result, the increase of valve core diam-eter is not an efficient way to optimize the maximumvelocity. It can be noticed that in the radial direction,the valve sleeve groove diameter, which is related to thescale of fluid space, is not a matching dimension. As ourobjective is sensitive to the radial dimension, it is pre-dicted to be another approach to reduce the maximumvelocity by increasing the valve sleeve groove diameterdirectly. Fig. 8(b) shows the analysis results with dif-ferent valve sleeve groove diameters. All these analysesare carried out based on a fixed valve core diameter, i.e.120mm. The modification starts with the initial valvesleeve groove diameter of 144mm. After this, the valvesleeve groove diameter is increased by 1mm per step.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

COMPUTER-AIDED DESIGN & APPLICATIONS 11

(a) (b) (c)

Figure 9. CAE Effect Feature Extraction.

Finally, the maximum velocity falls below 27m/s leadingto the convergence. By extracting the CAE effect fea-ture, the sensitivity towards different parameter changeis obtained providing physical reasoning back to CADmodification. Thus, a bi-directional reasoning mecha-nism is established.

In addition to optimize the physical property of sin-gle point of interest, CAE effect feature also demonstratesthe effect on the whole fluid domain by grid mappingmethod. A set of characteristic points are assigned tothe locations where interested, which is shown in Fig. 9.AssumeFig. 9(a) shows the PFSPSE. Fig. 9(b) is generatedin the process of valve core diameter change while othergeometrical features remain unchanged such as the axialdimensions and outer radial boundary. Then the gridmapping between the flow states before and after thechange can be done by scaling in each mesh cell in anoutward direction. Then the current flow field coupledwith grids of CFSPSE is compared with persistent flowfield reference model demarcated by PFSPSE. The cor-responding scaling can be shown as between cell g(1,x)of Fig. 9(a) and cell g(2,x) of Fig. 9(b). The error of thisproposed mapping is negligible if the step of modifica-tion is very small. If the change causes remeshing of thefluid space as shown in Fig. 9(c), then the effect featureextraction needs regional interpolation between PFSPSEandCFSPSEmesh grids as well as the flowproperty prop-agation. The corresponding interpolation unification canbe worked out as between region g(2, y) of Fig. 9(b) andregion g(3, y) of Fig. 9(c).

The difference in the velocity of corresponding pointsis calculated after the first modification. Divided by thechange in valve core diameter, the velocity change rate

can be derived further, which is shown in the followingmatrix. This is the mathematical form of CAE effect fea-ture, which shows the area and intensity of influence.In this matrix, along with the increase of valve corediameter, positive values indicate the velocity increasesand the negative ones for the velocity decreases. More-over, each points sensitivity towards the change of valvecore diameter is reflected by the value, which is of greatsignificance to the further optimization. For instance,local large velocity is one of the causes of hydraulicshock. The extraction of CAE effect feature can iden-tify the region related to hydraulic shock and pro-vide efficient method to mitigate even eliminate thisphenomenon.

VxDc

=

0.08 0.08 0.08 0.08 0.08 0.20 0.20 0.080.25 0.25 1.33 1.330.25 0.40 0.45 0.40

0.03 0.08 0.36 0.70 0.95 0.40 0.08 0.030.40 0.40 0.40 2.060.55 0.45 0.40 2.10

0.08 0.19 0.45 1.05 2.45 1.3 0.14 0.080.50 0.50 0.40 0.400.50 0.50 0.40 0.40

0.20 0.60 0.60 0.20 0.20 0.40 0.40 0.201.00 1.00 1.05 1.05

0.65 0.65 0.70 0.70

As is shown in the matrix, the majority of the pointsvelocity decreases with the increased valve core diameter,and the changes are more remarkable in the areas wherethe flow channel changes greatly. In each of the iterations,the matrix is checked for the validity of the modificationmethod.

5. Conclusions

This paper explores a mechanism of CAD/CAE inte-gration based on the concept of associative feature. An

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

12 L. LI AND Y. MA

overall feature mapping framework for persistent asso-ciations for CAD/CAE interactions has been suggested.This framework can effectively represent cyclic knowl-edge of product design, CAE evaluation, change justifica-tion and optimization, and design evolvement. Througha case study, it is clear that the associations between CADmodel and CAE analysis can be achieved. It can be con-cluded that associative feature is an effective mechanismof managing not only geometric associations, but alsosemantic portions of the features.

The concept of CAE effect feature is proposed for thefirst time, enabling the realization of the integrated loopof CAD/CAE sessions with effect comparison over thechanges. In addition, the interpretation of CAE effectwith features provides a consistent way to measure theoptimization results and navigate its procedure whichso far cannot be solved by CAE methods. This work iscurrently focused on the CAD/CAE interaction mecha-nism related to fluid dynamics; the example illustrates thedesign process of a hydraulic valve. As an objective, themaximum velocity in the valve flow field is studied andthe CAE effect features are extracted which leads to anoptimized design.

The consistency and accuracy of integration is main-tained by introducing another new CAE feature, CAEboundary feature. CAEboundary featuresmodel and rep-resent boundary conditions during iterations and achieveassociated mesh regeneration for solver setup. At thisstage, the authors are encouraged by the generic capa-bility of the integration method and are not aware ofany limitation of the proposed approach so long theoptimization space keeps the similar topology. Furtherresearch needs to be conducted to reveal the efficiencyand deficiency.

As to future work, under this proposed associativeCAD/CAE integration scheme, prototyping of automatedfunctions is to be carried out to provide implementa-tion proof. Further, it is worthwhile to study the accurateextraction of CAE effect features with less computationalcost. A CAE analysis interpretation method is expectedto conduct optimization of the physical field, and thuscontrolling the corresponding phenomena.

Acknowledgement

The authors would like to acknowledge the financial supportfrom China Scholarship Council, University of Alberta Doc-toral Recruitment Scholarship, NSERC Discovery Grant andRGL Collaboration Fund.

ORCID

Lei Li http://orcid.org/0000-0002-5259-407XYongsheng Ma http://orcid.org/0000-0002-6155-0167

References

[1] ANSYS, http://www.ansys.com/staticassets/ANSYS/staticassets/resourcelibrary/presentation/integrated-optimization-system-fedesign.pdf

[2] Bronsvoort, W. F.; Jansen, F. W.: Feature modelling andconversionkey concepts to concurrent engineering,Computers in Industry, 21(1), 1993, 6186. http://dx.doi.org/10.1016/0166-3615(93)90045-3

[3] Bronsvoort, W. F.; Noort, A.: Multiple-view feature mod-elling for integral product development, Computer-AidedDesign, 36(10), 2004, 929946. http://dx.doi.org/10.1016/j.cad.2003.09.008

[4] Cuillire, J. C.; Francois, V.: Integration of CAD, FEAand topology optimization through a unified topologicalmodel, Computer-Aided Design and Applications, 11(5),2014, 493508. http://dx.doi.org/10.1080/16864360.2014.902677

[5] Deng, Y.-M.; Lam, Y. C.; Tor, S. B.; Britton, G. A.:A CAD-CAE integrated injection molding design sys-tem, Engineering with Computers, 18(1), 2002, 8092.http://dx.doi.org/10.1007/s003660200007

[6] Delap, D.; Hogge, J.; Jensen, C. G.: CAD-centric creationand optimization of a gas turbine flowpath module withmultiple parameterizations. Computer-Aided Design andApplications, 3(14), 2006, 175184. http://dx.doi.org/10.1080/16864360.2006.10738454

[7] De Martino, T.; Giannini, F.: Feature-based ProductModelling in Concurrent Engineering, In Globaliza-tion of Manufacturing in the Digital CommunicationsEra of the 21st Century, Springer US, 1998, 351362.http://dx.doi.org/10.1007/978-0-387-35351-7_28

[8] Ferrandes, R.; Marin, P. M.; Lon, J. C.; Giannini, F.: Aposteriori evaluation of simplification details for finite ele-ment model preparation, Computers & Structures, 87(1),2009, 7380. http://dx.doi.org/10.1016/j.compstruc.2008.08.009

[9] Franois, V.; Cuilliere, J. C.: 3D automatic remeshingapplied to model modification, Computer-Aided Design,32(7), 2000, 433444. http://dx.doi.org/10.1016/S0010-4485(00)00028-2

[10] Gunette, R.; Fortin, A.; Kane, A.; Htu, J. F.: An adaptiveremeshing strategy for viscoelastic fluid flow simulations,Journal of Non-Newtonian Fluid Mechanics, 153(1),2008, 3445. http://dx.doi.org/10.1016/j.jnnfm.2007.11.009

[11] Hamri, O.; Lon, J. C.; Giannini, F.; Falcidieno, B.: Com-puter aided design and finite element simulation consis-tency, Strojniki vestnik-Journal of Mechanical Engineer-ing, 56(11), 2010, 728743.

[12] Johansson, J.: A feature and script based integration ofCAD and FEA to support design of variant rich prod-ucts, Computer-Aided Design and Applications, 11(5),2014, 552559. http://dx.doi.org/10.1080/16864360.2014.902687

[13] Kao, Y. C.; Cheng, H. Y.; She, C. H.: Development ofan integrated CAD/CAE/CAM system on taper-tippedthread-rolling die-plates, Journal of Materials ProcessingTechnology, 177(1), 2006, 98103. http://dx.doi.org/10.1016/j.jmatprotec.2006.04.082

[14] Lee, S. H.: A CADCAE integration approach usingfeature-based multi-resolution and multi-abstractionmodelling techniques, Computer-Aided Design, 37(9),

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

http://orcid.org/0000-0002-5259-407Xhttp://orcid.org/0000-0002-6155-0167http://www.ansys.com/staticassets/ANSYS/staticassets/resourcelibrary/presentation/integrated-optimization-system-fedesign.pdfhttp://www.ansys.com/staticassets/ANSYS/staticassets/resourcelibrary/presentation/integrated-optimization-system-fedesign.pdfhttp://www.ansys.com/staticassets/ANSYS/staticassets/resourcelibrary/presentation/integrated-optimization-system-fedesign.pdfhttp://dx.doi.org/10.1016/0166-3615(93)90045-3http://dx.doi.org/10.1016/0166-3615(93)90045-3http://dx.doi.org/10.1016/j.cad.2003.09.008http://dx.doi.org/10.1016/j.cad.2003.09.008http://dx.doi.org/10.1080/16864360.2014.902677http://dx.doi.org/10.1080/16864360.2014.902677http://dx.doi.org/10.1007/s003660200007http://dx.doi.org/10.1080/16864360.2006.10738454http://dx.doi.org/10.1080/16864360.2006.10738454http://dx.doi.org/10.1007/978-0-387-35351-7{_}28http://dx.doi.org/10.1016/j.compstruc.2008.08.009http://dx.doi.org/10.1016/j.compstruc.2008.08.009http://dx.doi.org/10.1016/S0010-4485(00)00028-2http://dx.doi.org/10.1016/S0010-4485(00)00028-2http://dx.doi.org/10.1016/j.jnnfm.2007.11.009http://dx.doi.org/10.1016/j.jnnfm.2007.11.009http://dx.doi.org/10.1080/16864360.2014.902687http://dx.doi.org/10.1080/16864360.2014.902687http://dx.doi.org/10.1016/j.jmatprotec.2006.04.082http://dx.doi.org/10.1016/j.jmatprotec.2006.04.082

COMPUTER-AIDED DESIGN & APPLICATIONS 13

2005, 941955. http://dx.doi.org/10.1016/j.cad.2004.09.021

[15] Lin, B. T.; Kuo, C. C.: Application of an integratedCAD/CAE/CAM system for stamping dies for automo-biles, The International Journal of AdvancedManufactur-ingTechnology, 35(910), 2008, 10001013. http://dx.doi.org/10.1007/s00170-006-0785-y

[16] Li, M.; Gao, S.; Martin, R. R.: Estimating the effectsof removing negative features on engineering analy-sis, Computer-Aided Design, 43(11), 2011, 14021412.http://dx.doi.org/10.1016/j.cad.2011.08.013

[17] Ma, Y.-S.; Tong, T.: Associative feature modeling for con-current engineering integration, Computers in Indus-try, 51(1), 2003, 5171. http://dx.doi.org/10.1016/S0166-3615(03)00025-3

[18] Ma, Y.-S.; Britton, G. A.; Tor, S. B.; Jin, L. Y.: Associativeassembly design features: concept, implementation andapplication, The International Journal of Advanced Man-ufacturing Technology, 32(56), 2007, 434444. http://dx.doi.org/10.1007/s00170-005-0371-8

[19] Matin, I.; Hadzistevic, M.; Hodolic, J.; Vukelic, D.; Lukic,D.: A CAD/CAE-integrated injection mold design sys-tem for plastic products, The International Journal ofAdvanced Manufacturing Technology, 63(58), 2012,595607. http://dx.doi.org/10.1007/s00170-012-3926-5

[20] Merkel, M.; Schumacher, A.: An automated optimiza-tion process for a CAE driven product development,Journal of Mechanical Design, 125(4), 2003, 694700.http://dx.doi.org/10.1115/1.1631570

[21] Park, H. S.; Dang, X. P.: Structural optimization basedon CADCAE integration and metamodeling tech-niques, Computer-Aided Design, 42(10), 2010, 889902.http://dx.doi.org/10.1016/j.cad.2010.06.003

[22] Robinson, T. T.; Armstrong, C. G.; Chua, H. S.: Strategiesfor adding features to CAD models in order to optimizeperformance, Structural andMultidisciplinary Optimiza-tion, 46(3), 2012, 415424. http://dx.doi.org/10.1007/s00158-012-0770-z

[23] Smit, M. S.; Bronsvoort, W. F.: Integration of designand analysis models, Computer-Aided Design and Appli-cations, 6(6), 2009, 795808. http://dx.doi.org/10.3722/cadaps.2009.795-808

[24] Tierney, C. M.; Nolan, D. C.; Robinson, T. T.; Armstrong,C. G.: Managing equivalent representations of designand analysis models, Computer-Aided Design and Appli-cations, 11(2), 2014, 193205. http://dx.doi.org/10.1080/16864360.2014.846091

[25] Uddin, M. M.; Ma, Y.-S.: A feature-based engineeringmethodology for cyclic modeling and analysis processesin plastic product development, Computer-Aided Designand Applications, to be published in 12(6), 2015.

[26] White, D. R.; Saigal, S.; Owen, S. J.: Meshing complexity ofsingle part CADmodels, Proceedings of the 12th Interna-tional Meshing Roundtable Conference, 2003, 121134.

[27] Zhu, H.; Gao, S.; Li, M.; Pan, W.: Adaptive tetrahedralremeshing for modified solid models, Graphical Mod-els, 74(4), 2012, 7686. http://dx.doi.org/10.1016/j.gmod.2012.03.005

Dow

nloa

ded

by [

Uni

vers

ity o

f A

lber

ta]

at 1

1:41

05

Nov

embe

r 20

15

http://dx.doi.org/10.1016/j.cad.2004.09.021http://dx.doi.org/10.1016/j.cad.2004.09.021http://dx.doi.org/10.1007/s00170-006-0785-yhttp://dx.doi.org/10.1007/s00170-006-0785-yhttp://dx.doi.org/10.1016/j.cad.2011.08.013http://dx.doi.org/10.1016/S0166-3615(03)00025-3http://dx.doi.org/10.1016/S0166-3615(03)00025-3http://dx.doi.org/10.1007/s00170-005-0371-8http://dx.doi.org/10.1007/s00170-005-0371-8http://dx.doi.org/10.1007/s00170-012-3926-5http://dx.doi.org/10.1115/1.1631570http://dx.doi.org/10.1016/j.cad.2010.06.003http://dx.doi.org/10.1007/s00158-012-0770-zhttp://dx.doi.org/10.1007/s00158-012-0770-zhttp://dx.doi.org/10.3722/cadaps.2009.795-808http://dx.doi.org/10.3722/cadaps.2009.795-808http://dx.doi.org/10.1080/16864360.2014.846091http://dx.doi.org/10.1080/16864360.2014.846091http://dx.doi.org/10.1016/j.gmod.2012.03.005http://dx.doi.org/10.1016/j.gmod.2012.03.005

1. Introduction2. Feature concepts in CAD/CAE integration3. Integration of CAD and CAE4. Case Study5. ConclusionsAcknowledgementORCIDReferences