interactive systems for designing machine elements...

Management and Production Engineering Review

Volume 6 • Number 3 • September 2015 • pp. 21–34DOI: 10.1515/mper-2015-0023

INTERACTIVE SYSTEMS FOR DESIGNING MACHINE ELEMENTS

AND ASSEMBLIES

Wojciech Kacalak1, Maciej Majewski1, Keith Douglas Stuart2, Zbigniew Budniak1

1 Koszalin University of Technology, Faculty of Mechanical Engineering, Koszalin, Poland2 Polytechnic University of Valencia, Department of Applied Linguistics, Valencia, Spain

Corresponding author:

Maciej Majewski

Koszalin University of Technology

Faculty of Mechanical Engineering

Raclawicka 15-17, 75-620 Koszalin, Poland

phone: (+48) 94 3478-352

e-mail: [email protected]

Received: 6 June 2015 Abstract

Accepted: 27 July 2015 The article describes the development of fundamentals of machine elements and assembliesdesign processes automation using artificial intelligence, and descriptions of structural ele-ments’ features in a natural language. In the proposed interactive automated design systems,computational artificial intelligence methods allow communication by speech and naturallanguage, resulting in analyses of design engineer’s messages, analyses of constructions, en-coding and assessments of constructions, CAD system controlling and visualizations. Thesystem is equipped with several adaptive intelligent layers for human biometric identifica-tion, recognition of speech and handwriting, recognition of words, analyses and recognitionof messages, enabling interpretation of messages, and assessments of human reactions. Thearticle proposes a concept of intelligent processing for analysis of descriptions of machineelements’ structural features in a natural language. It also presents the developed method-ology for similarity analysis between structural features of designed machine elements andcorresponding antipatterns allowing normalization of parameters of the analysed structuralsolutions.

Keywords

design systems, interactive systems, intelligent interfaces, antipatterns, computer-aideddesign.

Non-repeatability of design processes

and uncertainty level of results

Many tasks related to designing machine and de-vice components are isolated cases. Any attempt atforecasting working properties of machine compo-nents and assemblies entails the necessity to face thefact, that the conditions of their manufacturing, in-stallation and operation are described with incom-plete and uncertain, as well as inaccurate informa-tion.The presented research involves the development

of intelligent interactive automated systems for de-signing machine elements and assemblies using de-scriptions of structural elements’ features in a nat-ural language. Realization of the automated designprocesses is in conditions of uncertainty and with

non-repeatable processes. We propose a new con-cept [1] which consists of a novel approach to thesesystems, with particular emphasis on their ability tobe truly flexible, adaptive, human error-tolerant, andsupportive both of design engineers and data process-ing systems.

The foundation of interactivity is bi-directionalcommunication between the data processing systemand the user. Results of evaluation, performed withartificial intelligence, of the designed solutions areused at as early as the design stage.

The aim of the research is to develop the basisfor new design processes of higher level of automa-tion, and object approach to problems and applica-tion of voice communication between the design engi-neer and data processing systems. Some of the mostsignificant disadvantages of the systems responsible

21


for creation of construction representation, presentin the current design systems, are the following:

• Creating designs by performing graphical opera-tions – using slow communication interfaces suchas: keyboard, tablet, and mouse – on basic com-ponents like graphical symbols and lines.

• Drawing still takes too much of a designer’sthought processing.

• Data completion and processing takes place onlayers – containing graphical elements of a cer-tain type – which makes it more difficult to takeadvantage of benefits of object-oriented approachtowards geometrical components of an item, e.g. ashaft’s step, or a particular cut-in.

• Data completion and processing taking place onlayers containing graphical elements of a certaintype (lines, circles, . . . ) makes it necessary fortechnological processes design software to performrecognition operations of basic geometric objects,based on analysis of their graphical representation(it has to reconstruct drawings in a technologicalaspect).

• Instead of storing information as objects repre-senting an item’s basic components, graphical in-formation is stored.

• Storing designs in formats characteristic of obso-lete vector graphics systems which use basic imagecomponents (lines) instead of objects, which canbe automatically rendered by a code interpreter.

• The disadvantage mentioned above can be ex-plained by analogy with comparison of differentmethods of data representation: technical draw-ings, description of structure of documents (XML)and webpages (HTML). The conclusion to bedrawn here is that instead of storing a vectorgraphics image, we simply can store item’s char-acteristics, and an interpreter will reconstruct theimage under any operating system, using universalsoftware. After making amendments to the itemits “image” will be stored again as a set of char-acteristics.

The comparison of the proposed new automateddesign system with the present system of carryingout design tasks is shown in Fig. 1.

The condition for effectiveness of automated sys-tems of machine components and composite ma-chines design is mainly the implementation of thefollowing mechanisms:

• Creating elementary objects, as components of anelement, based on description created or passed inreal time in a form of voice or symbols.

• Eliminating defects and operator’s errors in theprocess of automatic interaction with the techni-cal documentation generation system.

• Automatic technological classification of machinecomponents using artificial neural networks.

• Fuzzy inferencing for normalization of structuralfeatures, as inputs for the automatic technologicalclassification system.

Fig. 1. Comparison of the proposed new automateddesign system with the present system of carrying out

design tasks.

In the design many advanced methods for prob-lem solving and task fulfilment are used, in which thedesigner’s explicit and implicit knowledge and skillsare required for determination of product features tobe created. The product features will be the outcomeof many complex processes and actions, and will beassessed by the user in various operational conditionsand various technical states.In the presented, new design system the proposed

methods for storing items’ characteristics allow anyadvanced integration of characteristics representa-tion of structures and technological processes, as wellas organization guidelines. However, there still willbe cases where a symbolic or verbal description mightbe ambiguous, and for such cases graphical data willremain important. Also, data coming from processesof shape and dimension reconstruction, in graphicaland numerical form, will serve as supporting data.In complex design tasks liberating the designer

from the necessity of manually using slow interfaceswill allow to eliminate the intermediate phase (com-posing an image using graphical symbols), which rel-egates object-oriented approach to elements to thelayer-oriented method, disadvantageous to furtherimplementations of a project.Many advanced methods for solving problems

and executing design tasks are used when designing

22 Volume 6 • Number 3 • September 2015


constructions. Explicit and tacit knowledge, as wellas designer’s skills are used to determine propertiesof a product, which will be later manufactured. Prop-erties of such a product will be the result of manycomplex processes, and work of many manufactur-ers, and users will test and judge those propertiesin different operational conditions and at differentgrades of technical condition. The time spent design-ing a component is the key element to the wholecreation process of new constructions, which shouldmeet certain strength and structural requirements,and also be faultless.

Machine components design is a very time-consuming and responsible task. It is the first phaseof the manufacturing process, and its aim is to antic-ipate the impact of external factors on the designedcomponent. Therefore designers must be able to cor-rectly combine knowledge of many fields. To their aidcome new design-aiding systems for both individualmachine components, as well as complex mechanicalsystems in a concept phase.

Currently, the demand from computer-aided de-sign is that it meet requirements [2] such as errorhandling, geometrical uncertainty, automatic errordetection, as well as intelligent design aiding [3]. Aspart of the last of the listed postulates, the authors,using their own artificial-intelligence-based solutionsrelated to bi-directional voice communication [4] be-tween technological devices and their operators, havebegun works the aim of which is to devise comple-mentary fundamentals for creating intelligent and in-teractive systems for designing machine componentsand assemblies, based on their attributes describedin a natural language.

Many research centers all over the world havebegun to work on features which would supplementCAD systems with simple mechanisms of speech in-put [5, 6] (in the form of a simple interface for recog-nizing selected basic shapes). Those attempts sharea simple postulate, which is to make it easier to workwith traditional systems (through a simple interfacefor limited control of a CAD system).

The objective character

of design features’ descriptions

A part of the proposed system is our symbolicallanguage used for defining structural properties of se-lected machine components during the design stage.Its purpose is to enable creation of descriptions offundamental objects, such as elements of a structure.

In order to ensure unambiguous description andcorrect semantics of the language, a method of dis-tinguishing individual elements constituting machine

parts has been defined. It has been specified thateach structural object has its own attributes (e.g.length, diameter, etc.) and is comprised of con-stituent objects. As an example, for a structuralobject being a machine shaft, constituent objectsare distinguished based on their position, dimensionsand shape, as well as their surface layer properties.Constituent objects are simple polyhedra and theyhave their own attributes: dimensions, shape and po-sition. In the given example a step of the shaft isa constituent object.

Any changes of dimensions, shape and surfacelayer properties are made with modifiers. A modi-fier also has its own attributes, which are defined inrelation to the modified constituent object (parentobject), to another constituent object, or to anoth-er modifier. A modifier cannot exist independently.Therefore, a modifier is an integral part of its parentobject. Examples of a modifier are chamfer, a hole,or a thread.

The main assumption of the symbolical nota-tion of structural features of machine parts’ elementsis that their description is object-oriented. Objectsare characterized by attributes and permitted oper-ations. Attributes are properties or variables associ-ated with the object, and operations are functions oroperations performed on object’s attributes for thepurpose of auto-modification or in order to influenceother objects. Access to an object and its attribut-es is possible only through a defined interface. Theinterface interacts with an object’s instance – whichcan be perceived as a living specimen (instance) ofa species (object). Functionality of an object is di-rectly connected with the attributes associated to it.

A disadvantage of the – currently used – struc-tural description, usually in the form of a tree, lessoften graphs, is lack of description of relations be-tween objects. When it comes to creation of tech-nological and organizational processes, object-baseddescription is important due to the manner of datanotation.

Object-oriented notation for structural featuresrequires creation of classes representing sets of ob-jects, as well as methods (functions) allowing to per-form an array of operations on particular objects,such as creation, deletion, modification, moving, ro-tation, etc. Objects of one class share many simi-larities, but they can insignificantly differ from eachother. A class defines which operations and attribut-es are associated with an object, but values assignedto attributes may vary between instances.

Complex structures can be described in manyways. It can be done using an XML-like language,a specialized language similar to HTML, or a sym-

Volume 6 • Number 3 • September 2015 23


bolical language. We have defined assumptions forcreation of a language, whereby a structural objectcan be easily divided into constituent objects, andmodifiers can be distinguished as elements for alter-ing dimensions, shape and properties of constituentobjects. Their symbolical notation is also simple andeasy to read.

Symbolical language is a form of machine nota-tion used by a system, which generates, and inter-prets such a notation for automatic real-time cre-ation of designed components. For the convenienceof designing engineers, a higher-level language canbe used, such as a constructional-XML.

Design tasks automation system’s

structure

The presented research proposes a concept andarchitecture of intelligent and interactive automat-ed machine components and assemblies design sys-tems [1] using descriptions of structural elements’

features in a natural language for design processesin conditions of uncertainty and with non-repeatableprocesses. Execution of design processes using thissystem (Fig. 2) allows:

• to have a higher level of design by unleashing de-signers’ full creative potential and liberating themfrom carrying out the process of creating graphicalrepresentations of designed components,

• to speed up the design process, especially in thecase of complex components,

• easier modification and rating of multiple solutionvariants,

• to automatize the most labour-intensive tasks inmachine components design,

• to create an object-oriented language for describ-ing structures, and also to create methods of sym-bolical notation,

• to improve the processing and data archiving op-erations,

• to create an artificial intelligence system aidingthe design processes.

Fig. 2. Concept of design processes using interactive automated systems.



Fig. 3. The architecture of interactive automated machine components and assemblies design systems.



Implementation of intelligent interaction systemsaims at increasing efficiency and comfort of design-ers and speeding up creation of new designs. Devel-opment of fundamentals of this new and promisingdesign methodology comprises execution of impor-tant and labour-intensive tasks:• development of fundamentals of effective conver-sion of voice messages into a symbolic object-oriented notation,

• improvement of designed methods of handwritingand graphical symbols recognition,

• research on effectiveness and correctness of de-sign’s features during its development, occurringautomatically with continuous integration of in-coming information regarding its attributes.Research on creation of an intelligent and inter-

active automated design system comprise:• creation of fundamental principles of structure el-ements features’ notation development,

• creation of fundamentals of a new language for no-tation, archiving and processing of data related toconstruction,

• description (a hypertext-based object-orientedconstruction description language),

• creation of improved methods for voice messagesrecognition and processing,

• development of created handwriting recognitionalgorithms,

• creation and testing of fundamental proceduresfor using artificial intelligence to create structures’symbolic notation based on their natural-languagedescription,

• verification of created methods for structure no-tation creation based on their description in anatural language for the following classes of ma-chine components: shafts, axles, spindles, cog-wheels, disks and more,

• examination of quality of generated projects withuncertainty rating for achieving results matchingtheir originals, and results of dependent tasks us-ing incomplete and uncertain data,

• creation of fundamentals of a new symbolic nota-tion language, and an object-oriented constructiondescription language,

• determination of direction for further research.The proposed interactive automated design sys-

tem (Fig. 3) [1] contains many specialized modulesand it is divided into the following subsystems [7–14]: subsystem for communication between design-ers and the intelligent CAD system, subsystem fordesign engineers’ voice messages content analysis,construction analysis subsystem, construction nota-tion subsystem, construction rating subsystem, sub-system for visualization and CAD system control,

design process optimization subsystem, constructiondecoding subsystem.

In this system, artificial intelligence methods al-low communication by speech and natural language,resulting in analyses of design engineer’s messages,analyses of constructions, encoding and assessmentsof constructions, CAD system controlling and visual-izations. The system is equipped with several adap-tive intelligent layers for human biometric identifica-tion, recognition of speech and handwriting, recog-nition of words, analyses and recognition of mes-sages [15], enabling interpretation of messages, andassessments of human reactions.

Designing with natural language

as the communication interface

between the designer and the system

The developed fundamentals of machine elementsand assemblies design processes automation using ar-tificial intelligence contain a concept of a system foranalysis of incoming information on structure’s fea-tures.

A variation of the used natural language (English,Polish) corresponds to the glossary of nomenclatureused during the design process, and to the librariesdeveloped for the purpose of processing speech (aspart of the voice user interface). It is possible to de-velop interfaces based on text, graphics, and symbols,depending on the user’s preferences.

The system performs intelligent processing foranalysis of descriptions of structural features of ma-chine elements and assemblies in a natural language.This proposed intelligent semantic-based system fortext corpus analysis is shown in abbreviated form inFig. 4A. It consists of a text corpus processing sub-system and an intelligent processing subsystem fordescription analysis.

In the text corpus processing subsystem, wordsare isolated from text extracted from the text cor-pus, which are developed into various combinationsof word clusters based on the statistical models ofword sequences. The developed word clusters rep-resenting appropriate N-gram models are processedfurther for training hybrid probabilistic neural net-works [22, 23] with learning patterns of words andclusters.

In the intelligent processing subsystem, wordclusters are retrieved from descriptions of structuralelements’ features in a natural language using a pars-er. In the next step, word clusters are extracted bythe parser using lexical and grammar patterns. Theseparated words are processed for letter strings iso-



Fig. 4. Diagram of the proposed system for analysis of incoming information on structure’s features, (B) inputs of theword recognition module, (C) inputs of the word cluster recognition module.

lated in segments as possible cluster word compo-nents. This analysis has been carried out using Ham-ming neural networks which are described in detail in[15–19]. The output data of the analysis consists ofprocessed word segments [20, 21]. Individual wordsegments treated here as isolated possible compo-nents of the cluster words are inputs (Fig. 4B) of hy-brid probabilistic neural networks [22, 23] for recog-nizing words. The networks use learning files contain-ing words and are trained to recognize words as wordcluster components, with words represented by out-put neurons. The intelligent cluster word recognitionmethod allows for recognition of words with similarmeanings but different lexico-grammatical patterns.In the next stage, the words are transferred to theword cluster syntax analysis module. The modulecreates words in segments as word cluster compo-nents properly, which are coded as vectors. Thenthey are processed by the module for word clus-ter segment analysis using hybrid binary neural net-works [23].The analyzed word cluster segments become in-

puts of the word cluster recognition module usinghybrid probabilistic neural networks (Fig. 4C). Themodule uses multilayer probabilistic neural networks,either to recognize the cluster and find its meaningor else it fails to recognize it. The neural networks

of this module use learning files containing patternsof possible meaningful word clusters. The intelligentanalysis and processing allow for recognition of anycombination of meaningful word clusters with similarmeanings but different lexico-grammatical patterns.The overall detailed results of the intelligent analysisare subject to processing for text corpus character-istics and its linguistic description including: statis-tical analysis, checking occurrences, and validatinglinguistic rules.The proposed intelligent system for analysis of in-

coming information on structure’s features containshybrid probabilistic neural networks which are de-scribed in detail in [22–24] (Fig. 5). The networkconsists of five layers: cluster processing, cluster in-put, cluster pattern, summation and output layers.This pattern classifier can become effective tools forsolving classification problems of lexico-grammaticalstructures, where the objective is to assign cases ofclusters of letters or words to one of a number ofdiscrete cluster classes. The classifier places each ob-served vector of cluster data x in one of the prede-fined cluster classes ki, i = 1, 2, ..., K where K is thenumber of possible classes in which x can belong. Theeffectiveness of the cluster classifier is limited by thenumber of data elements that vector x can have andthe number of possible cluster classes K.



Fig. 5. The hybrid probabilistic neural networks for recognition of clusters of letters or words [23], (B) Neuron of thepattern layer, (C) Neuron of the output layer.

The probabilistic neural network offers a way tointerpret the network’s structure in the form of aprobabilistic density function. It approximates theprobability that vector x belongs to a particular classki as a sum of weighted Gaussian distributions cen-tred at each cluster training sample. The output ofthe model is an estimate of the cluster class mem-bership probabilities.

Antipatterns in assessment

of design solutions’ quality

The role of antipatterns

in the machine design process

The antipattern concept, which is successfully ap-plied in software engineering, can serve as inspira-tion to using it in other engineering-related tasks.The most favourable scenario would be a task basedon heuristic concept creation, in which the designerwouldn’t be limited to using patterns, but instead he

would be expected to come up with innovative solu-tions characterized by a high level of non-obviousnessin relation to the established knowledge.

Issues connected with optimal design have beenelaborated in many publications on subjects likedesign method selection [25, 26]. Methods such asprogramming tasks, equality constraints with excessvariables, dynamic programming, and optimizationof variable-topology structures are useful when itcomes to accepting decisions regarding the value ofa structural feature. When making such decisions,stress and distortions, as well as correspondence ofgeometries and topologies from a finite set of com-binations with other structural elements are takeninto account. However, usefulness of the mentionedmethods in conception development is limited.

The set of methods used for solving problemsconnected with designing machine components com-prises many topics related to data analysis and ex-ploration [27–31]. This set of methods can be ex-panded with many more details, such as: multidi-



mensional scaling, tree-based classifiers, neural clas-sifiers, search algorithms (including heuristic ones),data-clustering algorithms, and content-based searchmethods.The drawback of using design patterns is that

not only do they impede innovation, but also theyincrease the odds of a scenario, where poor applica-tion of a pattern produces additional design errors,or contributes to delayed detection of such errors andmay result in repeated application of solutions worsethan the possible correct ones, coming at the samecost. Antipatterns are generic definitions of possibleinstances of incorrect solutions to design problems.Work on theoretical and experimental basics of an-tipattern definition methodology, their features andexceptional matching factor measure evaluation clas-

sification ability are important lines of research, im-portant to the automated systems of aided designprocess, which aspire to being prevalent in the nearfuture.Antipatterns can comprise multiple attributes:

(AFD) the function served by the designed compo-nent or structural system, (ASG) the component’sstructural group, (AIS) the definition of an incor-rect solution, (AEC) the error cause, (AER) the er-ror result, (AEI) the error’s importance, (AES) othererror similarity, and design solution features (AF1),(AF2)..(AFn). The distinct features of antipatternsare their geometrical and physical properties. An-tipatterns can also be classified based on the rootcauses of flawed designs, the most common of whichare presented in Table 1.

Table 1Root causes of designs with errors.

The name of an ERROR’s cause for each symbol

EPI a lack of correct evaluation of PRODUCT’S IMPORTANCE for the future manufacturing program

EOE a lack of project’s economical OUTCOME EVALUATION

ETP an underestimation of TECHNOLOGICAL PROBLEMS’ impact on the product’s quality

EMR a lack of analysis of MAINTENANCE REQUIREMENTS

ECP a lack of analysis of CUSTOMER PREFERENCES in relation to quality, modernity, durability, ease ofoperation, and visual design

EDG DOMINANCE of GRAPHICAL tasks over conceptual analysis in the process of structure design

EDT DOMINANCE of TRUST in results of massive data processing over trust in knowledge

ESE a deferral of SOLUTION EVALUATION, with overemphasized trust in ease of maintenance

EAF an introduction of an ABUNDANT FUNCTIONALITY

ECF incorrectly COMBINED FUNCTIONS in the designed elements

EAE existence of ABUNDANT ELEMENTS, which do not serve any functions

ELH a LACK of HARMONY in functionality, incorrect division of functions into elements of the structure

ELV a LACK of a VERIFICATION (competent and multicriterial) at each stage, and a lack of quality evaluation,supposed to be carried out with the required procedures

EDR DEADLINE RUSH due to short time

EAT incorrectly ASSIGNED TASKS within the team

EDV either a neglect or awareness of DEFERRED VERIFICATION of structural design effects and unclearevaluation of responsibility

EAS stubborn ADHERENCE to SOLUTIONS (known and otherwise correct)

EIM INCOMPATIBLE METHODOLOGIES, preferences, and design criteria among team members

EOC application of OLD COMPONENTS and norms in the designed structure

ELE a LACK of EVALUATION of a need for scalable, modular and unifiable solutions

ENE an excessive (implying a lack of integration) or an insufficient (implying differentiation) NUMBER of ELE-MENTS

EDS too easily DEFORMABLE STRUCTURE

ECM incorrect CHOICE of MATERIALS

EDI DIFFICULT INSTALLATION, limited or costly parts replacement ability

EQI an insufficient emphasis on the QUALITY of INTERACTING surfaces and the properties of outer layers

EAA an excessively limited ABILITY to ADAPT structural features to operating conditions

EED a lack of EVALUATION of DYNAMICS of operating conditions

EEO a lack of EVALUATION of OCCURRENCES, which are not characteristic of the typical operation (condi-tions pacing)



Evaluation measures of correspondence

between the designed structure

and antipatterns

Determination of measures for evaluation of cor-respondence between the designed structure and an-tipatterns requires normalization of compared fea-tures. Antipatterns’ linguistic attributes can beused for selection by (AFD), (ASG), (AIS), (AEC),(AER), (AEI), (AES), (AF1), (AF2) ... (AFn). In-put attributes require normalization to fit in the (0,1)range. Below are listed examples of attributes usedfor antipatterns selection:

• (AFD) – the function served by the designed el-ement or aggregate of components – rotating el-ement, housing, hood, frame, cantilever, adaptor,guide, cylinder, valve, handle, floor, pipe, springetc.,

• (ASG) – component’s structural groups, forAFD=’rotating’ possible options are: straight orstep shaft, spindle, disk, ring, roller, ball, and ad-ditionally: crankshaft, axle, toothed wheel etc.,

• (AIS) – incorrect solution’s definition, forAFD=’rotating’ and ASG=’shaft’ such a defini-tion can be: easy deformation, excessive differencebetween steps’ diameters, incorrect diameter toler-ance, high level of steps, lack of exit zone, incorrectdimensions of splineway etc.

Evaluation measures of correspondence will de-pend on the set of parameters typical of the afore-mentioned attributes, and they should be subjectedto normalization, so that they fall within the rangeof (0,1) using fuzzy sets. Error’s importance (AEI)should be – after normalization – attached to theset of attributes describing structural features. Nor-malization by the shape of the classification functionshould take into account the meaning of the valueto the result of normalization (flow direction). Thedeveloped methodology for design solutions evalua-tion with antipatterns has been presented in Fig. 6.An application example of the design solutions eval-uation methodology is a seven-step shaft (n = 7),presented in Fig. 7.

Fig. 6. A step shaft.

Fig. 7. Design solutions and antipattern correspondenceevaluation methodology.

The selected shaft antipattern has been, for thedesigned element’s function AFD=’rotating’, hasbeen chosen from the structural group ASG=’stepshaft’. The incorrect solution AIS1=’easy deforma-tion’ is characterized by shaft buckling due to anaxial force. The cause of the error AEC1=’highSL=L/dekw’ is such, that a flexible shaft, due to ax-ial compression, may be subject to buckling (bend-ing). In the case of the given example importanceAEI1 is high and it comprises similarity to other er-rors AES1: stability loss, which leads to inevitablephysical destruction, which usually involves other in-teracting elements, small stiffness, large deviations,noise, vibrations, unstable operation, etc.Another incorrect solution is AIS2=’excessive di-

ametral steps’, the cause of which AEC2=’high Sd’may lead to excessive stress cumulation caused byindentation. The Sd value is given by the formula:

Sd =√

Ss · Sm, (1)

where

Ss = max ∈

{

dj

dj−1

,dk

dk+1

}

, (2)

j ∈ {2, 3, ..i.., n}, k ∈ {1, 2, ..i.., n− 1}, (3)

di ∈ {d1, d2, ...., dn}, (4)



Sm =dmax

dmin

, (5)

dmax = max {d1, d2, ...., dn}, (6)

dmin = min {d1, d2, ...., dn}. (7)

To lessen the indentation’s impact it is advisedto assume Sm = dmax/dmin ≤1,2 and also to intro-duce a possibly large intermediate radius or a conicaltransition. The importance of the AEI2 error is con-siderable and it comprises similarities to other errorsAES2: stress cumulation, fatigue endurance, difficultinstallation, etc.More examples of applied methodology for de-

sign solution and antipattern correspondence eval-uation are presented in Figs. 8, 9 and 10. Selected

Fig. 8. A design of a shaft manufactured with machinecutting with tool entrance and exit zones: a) antipattern,

b) correct design.

Fig. 9. Design of a shaft with friction fit’s mountingsurfaces of different lengths.

shaft antipatterns for the designed element’s functionAFD=’rotating’ have been chosen from the struc-tural group ASG=’step shaft’.

Fig. 10. A machine shaft with highlighted errors:

a) an antipattern, b) correct design.

In the example presented in Fig. 8 AIS=’tool ex-it zone’ is an incorrect solution, the cause of whichAEC=’tool exit zone’s shape’ is the fact, that theshaft doesn’t provide a zone for the tool to freelymove around and exit certain zones during the man-ufacturing process. In the case of the given exampleimportance of AEI is high and it comprises similar-ity to other errors AES: manufacturing errors, pro-ducibility, poor installation, etc.Figure 7 presents a shaft designed with tool’s

reach and exit ability in mind, in accordance with theproducibility criterion and not displaying the afore-mentioned antipattern characteristics.In the example presented in Fig. 9 the incorrect

solution is AIS1=’installation of fitted elements’, thecause of which AEC1=’the Lp fitted surface’s length’is such, that the shaft has a mounting surface withthe diameter d = F85p6 and length Lp = 203 mm,which is much longer than the width of the elementmounted on it, e.g. a toothed wheel. The cause of theerror (AEC1) results in difficult installation – thatis forcing the element through a significant length.In the case of the given example the importance ofthe error AEI1 is average and it comprises similar-ity to other errors AES1, such as: causing damageof the mounting surface, costs of forcing the shaft



through, etc. At the same time, the shaft has an errorAIS2=’imprecise axial fitting of the mounted compo-nents’. The cause of the error AEC2=’filleting radiusR’ indicates the possible imprecision of componentsaxial mounting (lack of support for the fitted com-ponent from the end collar), in which the size of themounted component’s phase F is smaller than theradius of the shaft step’s filleting R. The error’s im-portance AEI2 is average, and its similarities sharedwith other errors AES2 are: noise, unstable opera-tion, vibrations, increased wearing out of the fittedelements’ surfaces, etc.

An example of an attributes set for the afore-mentioned cases can be as follows: (AEIN) – er-ror importance normalized value, (LVN) – normal-ized length value L to the diameter equivalent dekw ,(DVN) – maximal Sm diameter value normalized tothe smallest one, (STN) – normalized ratio of sur-face’s roughness parameter St to this surface’s sizetolerance T , etc.

The measure of correspondence with an antipat-tern for the given examples can be the sum ofabsolute distance values between the parametersof an antipattern and the designed element, thesum of squared distances or their geometrical dis-tance.

Before performing the evaluation of correspon-dence between the structural features of a designedmachine elements and antipatterns, it is possible toverify the ability to classify of the used parame-ters. The requirement would be to have a set ofdesign solutions, for which we want to determinethe differentiation ability with each of the parame-ters.

Presented antipatterns comprise a generic defini-tion of possible incorrect design solutions. The devel-oped methodology for evaluation of correspondencebetween a design solution and an antipattern, illus-trated with step shafts, allows to normalize parame-ter characteristics of selected features of analysed de-sign solutions, and, then, to determine the antipat-tern’s and the analysed structure’s matching factor.Work on theoretical and experimental basics of an-tipattern definition methodology, their features andexceptional matching factor measure evaluation clas-sification ability are important lines of research, im-portant to the automated systems of aided designprocess, which aspire to being prevalent in the nearfuture.

Conclusions

Works aiming to develop basics of automa-tion of processes in designing machine elements

and assemblies with the use of artificial intelli-gence in uncertainty and unrepeatability of process-es have been started. In automated design systems,which use a natural-language description of struc-tural features and an intelligent interface of naturalspeech and hand-drawn sketches, application of de-sign antipatterns – especially in combination withartificial-intelligence methods – carries a lot of mean-ing to effectiveness and development of such sys-tems.The devised concepts for creating a design and

object-oriented notation of information about itscharacteristics, in comparison to the known and usedsolutions, have benefits resulting from lack of con-straints related to graphical operations performedby the designer, as well as from the advantages ofobject-oriented data notation and lack of constraints(amount and type of data) related to the structureof classes representing object types.The benefits of using antipatterns are lack of

constraints when designing new solutions and less-er likelihood of making errors typical of antipat-terns.The devised solutions can be implemented in

commercial products, which may considerably accel-erate the processes of designing simple machine com-ponents, make more feasible the creation of morecomplex designs, as well as decrease the impact oflimitations related to the amount of work that canbe invested into potential corrections and analy-ses during the design creation process. However,in the case of elements characterized by complexshapes, difficult to define in linguistic terms, graph-ical interface-based methods will still find their ap-plication.

This project was financed from the funds of theNational Science Centre (Poland) allocated on thebasis of the decision number DEC-2012/05/B/ST8/02802.

References

[1] Kacalak W., Majewski M., New Intelligent Interac-tive Automated Systems for Design of Machine Ele-ments and Assemblies, ICONIP 2012, Doha, Qatar,in Lecture Notes in Computer Science, 7666, IV,115–122, 2012.

[2] Zeng Y., Horvath I., Fundamentals of next genera-tion CAD/E systems, Computer-Aided Design, 44,10, 875–878, 2012.

[3] Goel A.,Vattam S., Wiltgen B., Helms M., Cogni-tive, collaborative, conceptual and creative – Four



characteristics of the next generation of knowledge-based CAD systems: A study in biologically inspireddesign, Computer-Aided Design, 44, 10, 879–900,2012.

[4] Piegl L.A., Ten challenges in computer-aided design,Computer-Aided Design, 37, 4, 461–470, 2005.

[5] Kou X.Y., Xue S.K., Tan S.T., Knowledge-guidedinference for voice-enabled CAD, Computer-AidedDesign, 42, 6, 545–557, 2010.

[6] Kou X.Y., Tan S.T., Design by Talking with Com-puters, Computer-Aided Design and Applications,5, 1–4, 266–277, 2008.

[7] Kacalak W., Majewski M., Intelligent system for au-tomatic recognition and evaluation of speech com-mands, ICONIP 2006, Hong Kong, China, in Lec-ture Notes in Computer Science, 4232, 298–305,2006.

[8] Majewski M., Kacalak W., Intelligent system fornatural language processing, ICIC 2006, Kunming,China, Lecture Notes in Computer Science, 4114,742–747, 2006.

[9] Kacalak W., Majewski M., Natural LanguageHuman-Robot Interface Using Evolvable FuzzyNeural Networks for Mobile Technology, ICIC 2009,Ulsan, South Korea, Lecture Notes in Computer Sci-ence, 5754, 480–489, 2009.

[10] Kacalak W., Majewski M., E-learning systems withartificial intelligence in engineering, ICIC 2009, Ul-san, South Korea, Lecture Notes in Computer Sci-ence, 5754, 918–927, 2009.

[11] Kacalak W., Majewski M., Zurada J.M., Intelligente-learning systems for evaluation of user’s knowl-edge and skills with efficient information processing,ICAISC 2010, Zakopane, Poland, Lecture Notes inComputer Science, 6114, 508–515, 2010.

[12] Lipinski D., Majewski M., System for Monitoringand Optimization of Micro- and Nano-MachiningProcesses Using Intelligent Voice and Visual Com-munication, IDEAL 2013, Hefei, China, LectureNotes in Computer Science, 8206, 16–23, 2013.

[13] Lipinski D., Majewski M., Interactive Hybrid Sys-tems for Monitoring and Optimization of Micro-and Nano-machining Processes, ICONIP 2013,Daegu, Korea, Lecture Notes in Computer Science,8227, 360–367, 2013.

[14] Lipinski D., Majewski M., Intelligent Monitoringand Optimization of Micro- and Nano-MachiningProcesses, Mechatronics 2014, Lodz, Poland, Ad-vances in Intelligent Systems and Computing, 317,101–110, 2015.

[15] Majewski M., Zurada J.M., Sentence recognitionusing artificial neural networks, Knowledge-BasedSystems, 21, 7, 629–635, 2008.

[16] Kacalak W., Stuart K., Majewski M., Intelligentnatural language processing, ICNC 2006, Xi’an, Chi-na, Lecture Notes in Computer Science, 4221, 584–587, 2006.

[17] Stuart K., Majewski M., Selected problems ofknowledge discovery using artificial neural networks,ISNN 2007, Nanjing, China, Lecture Notes in Com-puter Science, 4493, 1049–1057, 2007.

[18] Kacalak W., Stuart K., Majewski M., Selected prob-lems of intelligent handwriting recognition, IFSA2007, Cancun, Mexico, Advances in Soft Comput-ing, 41, 298–305, 2007.

[19] Stuart K., Majewski M., A new method for intelli-gent knowledge discovery, IFSA 2007, Cancun, Mex-ico, Advances in Soft Computing, 42, 721–729, 2007.

[20] Stuart K., Majewski M., Artificial creativity in lin-guistics using evolvable fuzzy neural networks, ICES2008, Prague, Czech Republic, Lecture Notes inComputer Science, 5216, 437–442, 2008.

[21] Stuart K., Majewski M., Evolvable neuro-fuzzy sys-tem for artificial creativity in linguistics, ICIC 2008,Shanghai, China, Lecture Notes in Computer Sci-ence, 5227, 46–53, 2008.

[22] Stuart K., Majewski M., Botella-Trelis A., Select-ed problems of intelligent corpus analysis throughprobabilistic neural networks, ISNN 2010, Shanghai,China, Lecture Notes in Computer Science, 6064,268–275, 2010.

[23] Stuart K., Majewski M., Botella-Trelis A., Intel-ligent semantic-based system for corpus analysisthrough hybrid probabilistic neural networks, ISNN2011, Guilin, China, Lecture Notes in Computer Sci-ence, 2011, 6675, 83-92.

[24] Specht D.F., Probabilistic neural networks, NeuralNetworks, 3, 1, 109–118, 1990.

[25] Brown W.J., McCormick H.W., Thomas S.W., An-tiPatterns in Project Management, New York: JohnWiley & Sons, 2000.

[26] Riel A.J., Object-Oriented Design Heuristics, Read-ing, MA: Addison-Wesley, 1996.

[27] Dostatni E., Diakun J., Grajewski D., WichniarekR., Karwasz A., Functionality assessment of ecode-sign support system, Management and ProductionEngineering Review, 6, 1, 10–15, 2015.

[28] Zhong H., Levalle R.R., Moghaddam M, Nof S.Y.,Collaborative Intelligence - Definition and Measured



Impacts on Internetworked e-Work, Managementand Production Engineering Review, 6, 1, 67–78,2015.

[29] Wasilewska B., Knosala R.,Methods for StimulatingCreativity in Production Engineering, Managementand Production Engineering Review, 5, 4, 76–84,2014.

[30] Jakubowski J., Peterka J., Design for manufactura-bility in virtual environment using knowledge engi-neering, Management and Production EngineeringReview, 5, 1, 3–10, 2014.

[31] Ahram T.Z., Karwowski W., Engineering sustain-able complex systems, Management and ProductionEngineering Review, 4, 4, 4–14, 2013.


interactive systems for designing machine elements...

Documents