- Email: [email protected]
The use of LPR (logic of plausible reasoning) to obtain information on innovative casting technologies
archives of civil and mechanical engineering 14 (2014) 25–31
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/acme
Original Research Article
The use of LPR (logic of plausible reasoning) to obtain information on innovative casting technologies S. Kluska-Nawareckaa, B. S´niez˙yn´skib, W. Paradab, M. Lustofinb, D. Wilk-Kołodziejczykc,n a
Foundry Research Institute in Cracow, Zakopiańska Street 73, Kraków, Poland AGH University of Science and Technology, Department of Computer Science, Mickiewicza 30, 30-059 Kraków, Poland c Andrzej Frycz Modrzewski Krakow University, Department of Applied Informatics, Gustawa Herlinga-Grudzińskiego 1, 30-705 Kraków, Poland b
art i cle i nfo
ab st rac t
Article history:
The paper presents a methodology for the application of a formalism of the logic of
Received 4 January 2013
plausible reasoning (LPR) to create knowledge about a specific problem area. In this case,
Accepted 30 May 2013
the methodology has been related to the task of obtaining information about the
Available online 5 June 2013
innovative casting technologies. In the search for documents, formulas created in terms
Keywords:
of LPR have a much greater expressive power than the commonly used keywords. The
Logic of plausible reasoning
discussion was illustrated with the results obtained using a pilot version of the original
Cast foundry
information tool.
Intelligent information system
& 2013 Politechnika Wroc"awska. Published by Elsevier Urban & Partner Sp. z o.o. All rights reserved.
Methodology for the application Innovative casting technologies
1.
Introduction
Currently, there are many search engines that allow us to obtain information from open sources. However, it appears that the use of a universal search engine usually leads to identification of a large number of documents that do not meet the user needs, and therefore require painstaking selection carried out in a “manual” mode, which is particularly troublesome in the case of making a large amount of expertise, or when quick decisions are to be made. As a result, the problems of effective information and knowledge retrieval from the available (often distributed) sources have recently been the subject of intensive work of many research centres [3,5,11,12]. The results obtained in this respect clearly show that a significant increase in search
efficiency can be achieved through the creation of solutions targeted at a specific area of knowledge and tailored to the specific sources, in which the acquisition is to be carried out. It is important for the search engines to be equipped with intelligent mechanisms (procedures) that allow for a possibly precise assessment of the contents of the documents obtained. This article presents a proposal for a solution where the search for documents is carried out using a knowledge base constructed with an LPR formalism [logic of plausible reasoning] [2]. The use of this formalism seems to be interesting because of its intuitive nature (expressing the user's way of thinking) and the ability to assess the degree of certainty of the results. The originality of the concept lies in the fact that the LPR was previously mainly used in typical expert systems
n
Corresponding author. Tel.: +48 505 353 700. E-mail address: [email protected] (D. Wilk-Kołodziejczyk).
1644-9665/$ - see front matter & 2013 Politechnika Wroc"awska. Published by Elsevier Urban & Partner Sp. z o.o. All rights reserved. http://dx.doi.org/10.1016/j.acme.2013.05.011
26
archives of civil and mechanical engineering 14 (2014) 25–31
Expert
Domain knowledge
Knowledge base
User
User interface
Inference
Information sources
Selected documents
Fig. 1 – Schematic diagram of the intelligent search engine.
[1,4–9]. There is only one work [10] in which LPR is used for knowledge representation and inference in recommendation system, but that work focuses on multistrategy learning. Here the problem is related with the representation of knowledge oriented at an intelligent search of documents (knowledge), focussed on a specific problem area and machine learning is not considered. The solution disclosed here relates to the sourcing of information on innovative foundry technologies, but the methodology itself used for the creation of a knowledge base, and the procedures for its use are universal in nature and may be applied to any problem area. In terms of utility, in the studies carried out currently, the application developed by the AGH Department of Computer Science and Technology has been used. Its functionality includes both the creation of a knowledge base expressed in terms of LPR, as well as the inference engine and user interface. Here (for reasons of volume) any broader description of the computing aspects of this application has been dropped, focussing attention on the presentation of the principles of creating a knowledge base and procedures for obtaining documents of the desired content. Schematic diagram illustrating the operating concept of the test system is shown in Fig. 1. Knowledge base, expressed in the form of rules, hierarchy, similarities and statements, is used to search the knowledge sources in compliance with the needs identified by the user through the interface. The result is a set of documents with the specified degree of reliability, corresponding to the user's intuition. This procedure will be further presented in relation to documents describing the innovative foundry technologies.
Table 1 shows fragment of a list of such innovations prepared with the assistance of foundry engineers from the Foundry Research Institute. The next phase that belongs to the expert competence is to formulate a list of attributes describing main entities. It is worth noting that the attributes can be single- or multivalued. In the former case, the attribute has the nature of a unique description of the document, while in the latter case it can take several values from the given set. In the process of searching, usually, only some of the attributes are known, while others are defined by the inference process conducted according to the LPR rules. The list of attributes for the innovations is presented in Table 2. Creating a knowledge base consists in generating for each of the main entities a set of the corresponding formulas. In the described application it is limited to the use of only certain selected types of LPR formulas and these are outlined below.
2.1.
Statements
Statement is called the expression: VðO; X; ZÞ
This means that the attribute X of object O has a value of Z.
2.2.
Hierarchies—generalisation and specialisation
Hierarchies are a way of describing the relationship between concepts, which in this case refers to the level of generality. Accordingly, a formula for generalisation is used: A GEN Ai
2. Methodology of the description of innovation The first phase in the expert activity is to identify main entities which can be searched for. Knowledge base is built around these entities. In our example domain, the search is for innovative foundry technologies [13], and so a set of technologies is formulated, which the documents sought should refer to.
ð1Þ
ð2Þ
where A is a more general term (dominant) with respect to Ai and for specialisation: Ai SPEC A
ð3Þ
where Ai is the clarification of the concept of A. Instead of these two forms the following may be used: HðAi ; A; CÞ
ð4Þ
where C represents a context in which the hierarchy is considered.
archives of civil and mechanical engineering 14 (2014) 25–31
27
Table 1 – Fragment of a list of innovations. No.
Name of innovation
1 2 3 4 5
Environment-friendly technology for the reclamation of bentonite moulding sands. New simplified method for mechanical reclamation of moulding and core sands. Environment-friendly technology for the recycling of waste core sands. Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine. New control and measurement apparatus adapted to the EU standards used for the bending strength determination in sands for the cold box process. An upgraded technology for the continuous preparation of moulding sands in a two-pan mixer. etc.
6 etc.
Table 2 – List of attributes that describe documents related to innovation. Name of attribute
Type
Description
Innovation name Article Target project Project title Term of project Project contractor Implementer Keyword Description
Single-valued Multi-valued Single-valued Single-valued Multi-valued Single-valued Single-valued Multi-valued Single-valued
Name of innovation; every innovation has got a name Numbers of articles, each article can also indicate a file or a link Reference number of a target project Title of project Years in which the project was implemented, recorded in “YYYY” format Name of company carrying out the project Project implementer Keywords related to innovation Description of innovation
As a clarification of the concept of the hierarchy, numerical factors are often incorporated determining typicality (τ) of the given relationship and the degree of dominance (δ) of the subordinate concept. Table 3 shows a fragment of the hierarchy statement established for the case of innovative technologies. In this table, the symbol H denotes hierarchical relationships, while words written in italics define variables for which values of the appropriate attributes are substituted.
2.3.
The similarity and dissimilarity
The similarity or dissimilarity of terms is written as Ai SIMAj : s
ð5Þ
Ai DISA2
ð6Þ
where Ai, Aj—terms subject to comparison, s—the degree of similarity or dissimilarity. Often the context in which the comparison is made is also referred to. We use the following notation to represent similarity: SðAi ; Aj ; CÞ : s
ð7Þ
of document searching, the rule expressing the implication can be written in a general form as: ½VðU; chose A; ZÞ; VðE; A; ZÞ-VðU; matches A; EÞ : 1; 0
ð8Þ
where variables used in statements have the following meaning: U—user, A—name of the attribute, Z—value of the attribute, E—main entity (defined in phase one), representing innovation in example domain. The implication expressed in the form of Eq. (8) has the following interpretation: from the statement that user (U) adopted for the attribute (A) the predetermined value (Z), and from the statement that main entity (E) is described with attribute (A) of the value (Z), the statement follows that main entity (E) matches (fits) user (U) in the context of the attribute (A) and to the extent estimated by the coefficient of the value 1.0. Selected implications describing the considered set of innovations are presented in Table 5. I is used instead of E to emphasise that entities represent innovations.
2.5.
Search mode and reliability of results
where C denotes context. Fragment of knowledge base regarding the similarity of concepts under consideration was compared in Table 4. As before, italics in this table refer to the variables for which the values of attributes have been substituted.
In ongoing experiments, two types of search procedures were used, i.e. fast mode and advanced mode. In the former case, the objective of the reasoning has the form of the following statements:
2.4.
VðE; A; ZÞ
Implications
A substantial part of the contents of the knowledge base is made of implications used in the process of reasoning. They define the relationships between the statements. In the case
ð9Þ
where E represents main entity defined in the first phase, A is attribute name and Z is its value. Inference is executed for every attribute chosen by the user and the goal is to find the entities which have attribute values provided.
28
archives of civil and mechanical engineering 14 (2014) 25–31
Table 3 – Selected hierarchies in the description of innovation. Rule
Typicality and dominance
Description
H(article_no. article, name_innovation)
τ¼ 0.8 δ ¼ 0.5 τ¼ 0.8 δ ¼ 0.5 τ¼ 0.8 δ ¼ 0.5 τ¼ 0.8 δ ¼ 0.6 τ¼ 0.8 δ ¼ 0.5 τ¼ 0.8
Each article is a type of article in the context of the name of a specific innovation Each article is a type of article in the context of the project title of a specific innovation Name of the project contractor is a type of the innovation contractor in the context of the project title Each keyword associated with innovation is a type of the keyword in the context of the article Each keyword associated with innovation is a type of the keyword in the context of the project contractor Each keyword associated with innovation is a type of the keyword in the context of the project title
H(article_no. article, title_project) H(name_contractor, project contractor, project title) H(key_word, keyword, article) H(key_word, keyword, project contractor) H(key_word, keyword, project title)
δ ¼ 0.5
Table 4 – Selected formulas of similarities. Rule
Degree of similarity
Description
S(key_word_1, key_word_2, innovation) S(key_word_1, key_word_2, project title) S(key_word_1, key_word_2, project contractor) S(article_1, article_2, innovation) S(article_1, article_2, project contractor) S(name_contractor, name_implementer, project title)
s¼ 0.8
Keyword 1 is similar to keyword 2 in the context of the same innovation
s¼ 0.8
Keyword 1 is similar to keyword 2 in the context of the same project title
s¼ 0.8
Keyword 1 is similar to keyword 2 in the context of the same project contractor
s¼ 0.8 s¼ 0.8
Article 1 is similar to article 2 in the context of the same innovation Article 1 is similar to article 2 in the context of the same project contractor of the same innovation Contractor name is similar to the name of the implementer in the context of the project title of the same innovation
s¼ 0.8
Table 5 – Selected implications. Implication
Description
[V(U, ‘chose keyword, Z), V(I, keyword, Z)] -V(U, matches keyword, I):1.0 [V(U, chose completion date, Y), V(I, completion date, Y)]-V(U, matches completion date, I):1.0 [V(U, chose article, M), V(I, article, M)]-V(U, matches article, I):1.0 [V(U, chose project title, N), V(I, project title, N)]-V(U, matches project title, I):1.0 [V(U, chose innovation name, J), V(I, innovation name, J)]-V(U, matches innovation name, I):1.0
If user U chose keyword Z and innovation I has a link with keyword Z, then innovation I matches user U in the context of the keyword If user U chose completion date Y and innovation I has a link with completion date Y, then innovation I matches user U in the context of the completion date If user U chose article M and innovation I has a link with article M, then innovation I matches user U in the context of the article If user U chose project title N and innovation I has a link with project title N, then innovation I matches user U in the context of the project title If user U chose innovation name J and innovation I has a link with innovation name J, then innovation I matches user U in the context of the innovation name
In advanced mode the following expression is used: VðU; matches A; EÞ
ð10Þ
Here also inference is executed for every attribute chosen. Consequently, in the latter case, the entire knowledge base is sought. Inference process may use implication (8) to find matching entities and can infer statements describing user preferences and main entities. In the former case only fragments of the
knowledge base are used (attributes of main entities) which of course has a significant effect on the time taken for response. The certainty factor assumes values from the range (0,1] and is calculated from the following formula: q ¼ ðq1 þ q2 þ ⋯ þ qn Þ=n
ð11Þ
where: qi—certainty value for the i-th chosen attribute, n—number of attribute values selected by the user.
29
archives of civil and mechanical engineering 14 (2014) 25–31
The certainty factor q is the basis for the ranking with preference for responses with the largest values. The certainty of the attribute values (qi) are calculated by the inference module based on the knowledge base and certainties provided by the user in the question definition.
3.
Test results
Completed application was tested on a set of documents created from the descriptions of research and implementation projects carried out at the Foundry Research Institute. The extensive material obtained as a result of the conducted experiments includes exploration in fast and advanced mode, taking into account the values of the 1–9 attributes. The following are extracts from these experiments that adequately illustrate the nature of the obtained results. Comparing the documents referred to in Table. 6 (1 attribute) with the contents of Table 7 (3 attributes) it can be concluded that increasing the number of the included attribute values will generally result in their more accurate
selection but does not always translate into increased certainty of the result. Analysing the impact of the search mode, it is clear that the advanced mode usually leads to the selection of more documents than the fast mode, but the certainty of the result for the additionally indicated documents is usually lower. It is worth noting that increasing the number of attributes included in the search significantly affects the increase in computation time (an exponential relationship). In assessing the overall results of the tests, it can be concluded that they have demonstrated the correct operation of the tool constructed and allowed observing certain regularities found in the search for innovation.
4.
Conclusions
This paper presents the concept of using LPR to search for the text documents with a given theme. It seems that the proposed methodology for the description of documents
Table 6 – Searching for innovations based on a single-attribute value. Attribute
Search mode
Searched innovations
Certainty of result
Attribute name: Description; Value: Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine; Certainty: 0.75
Fast
Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine to make moulds in bentonite sands Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine to make moulds in bentonite sands Implementation of the reclamation process of selfsetting sands according to a new vibration-fluidised bed method Development and implementation of a technology for the manufacture of a new grade of foundry resin with reduced content of free formaldehyde and its application in selected domestic foundries Environment-friendly technology for the reclamation of bentonite moulding sands New simplified method for mechanical reclamation of moulding and core sands Environment-friendly technology for the reclamation of bentonite moulding sands New simplified method for mechanical reclamation of moulding and core sands Environment-friendly technology for the reclamation of bentonite moulding sands Environment-friendly technology for the reclamation of bentonite moulding sands Environment-friendly technology for the reclamation of bentonite moulding sands Implementation of the reclamation of self-setting sands according to a new vibration- fluidised bed method Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine
0.6
Advanced
Attribute name: Article; Value: article_5; Certainty: 1.0
Fast
Advanced
Attribute name: Innovation name; Value: Environment-friendly technology for the reclamation of bentonite moulding sands; Certainty: 0.75
Fast Advanced
0.499
0.224
0.224
0.224 1.0 1.0 1.0 1.0 1.0 0.75 0.375
0.375
30
archives of civil and mechanical engineering 14 (2014) 25–31
Table 7 – Searching for innovations based on the values of three attributes. Attributes
Type of search
Innovations search
Certainty of result
Attribute name: Innovation name; Value: New technology for casting cylinder sleeves in spinning metal moulds;
Fast
New technology for casting cylinder sleeves in spinning metal moulds New simplified method for mechanical reclamation of moulding and core sands Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine to make moulds in bentonite sands … New technology for casting cylinder sleeves in spinning metal moulds New simplified method for mechanical reclamation of moulding and core sands Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine Development of design and implementation in production of a new versatile 20 dcm3 capacity core shooter to make cores by various technologies Implementation of the reclamation process of self-setting sands according to a new vibration- fluidised bed method Implementation of processes for moulding sand preparation, treatment and circulation and for casting automotive parts from ductile iron on a pilot line using prototype stand for impulse-squeeze moulding …
0.4666
Certainty: 1.0 Attribute name: Project contractor; Value: FERREX Sp. zo.o. Poznań; Certainty: 1.0
Attribute name: Project title; Value: Two-post FT-65 impulse-squeeze moulding machine; Certainty: 1.0
Advanced
greatly increases the power of its expression in relation to commonly used search based only on keywords. The attributes used in the described approach are a certain generalisation of keywords, and moreover, use of LPR formalism allows taking into account the context in which these attributes are present. As a problem area for which the proposed methodology has been used, the search for innovation in casting technologies was chosen as characterising well enough the specific nature of this class of the tasks. The designed pilot solution for the information tool makes it possible to carry out the experimental verification of the effectiveness of the approach. The tests performed are of course preliminary, but create successful outcomes for future work in this area. In the future it is expected to examine the effectiveness of information retrieval with a much larger collection of source documents, and use data sources of a distributed character, the Internet included.
Acknowledgements Scientific work financed from funds for the scientific research as an international project. Decision no. 820/N-Czechy/2010/0
references
[1] James L. Beck, Bayesian system identification based on probability logic, Structural Control and Health Monitoring1545-225517 (7) (2010) 825–847.
0.4666 0.3333
0.1333 0.4666 0.4666 0.3333
0.2999
0.2999 0.2999
0.1333
[2] A. Collins, R.S. Michalski, The logic of plausible reasoning: a core theory, Congnitive Science 13 (1) (1989) 1–49. [3] J. David, M. Vrožina, Z. Jančíková, Determination of crystallizer service life on continuous steel casting by means of the knowledge system, WSEAS Transactions on Circuits and Systems 10 (10) (2011) 359–369 ISSN:1109-2734 359. [4] S. Kluska-Nawarecka, Z. Górny, D. Wilk-Kołodziejczyk, A. Smolarek-Grzyb, The logic of plausible reasoning in the diagnosis of castings, Archives of Metallurgy and Materials1733-3490375–380. [5] I. Olejarczyk, A. Adrian, H. Adrian, B. Mrzygłód, Algorithm for controlling of quench hardening process of constructional steels, Archives of Metallurgy and Materials17333490171–179 Bibliogr. s. 179. [6] M. Saad KhanAn, T.R. Soomro, Overview of human plausible reasoning, International Journal of Computer Applications 45 (18) (2012) 0975–8887. [7] B. Śnieżyński, Basic semantics of the logic of plausible reasoning, in: Mohand-Saïd Hacid et al.(Ed.), Lecture Notes in Computer Science; vol. 2366. Lecture Notes in Artificial Intelligence, Springer-Verlag, Berlin etc., 2002, pp. 176–184 Bibliogr. s. 184, Abstr.3. [8] B. Śnieżyński, Proof searching algorithm for the logic of plausible reasoning, in: A.Kłopotek Mieczysław, T. Sławomir (Eds.), Wierchoń, Krzyszt of Trojanowski; Polish Academy of Sciences, Instytute of Computer Science—Berlin; Heidelberg; New York, Springer-Verlag, 2003, pp. 393–398 (Advances in Soft Computing; Series 4240). [9] D. Wilk-Kołodziejczyk, Budowa algorytmów wiedzy dla potrzeb diagnostyki wad wyrobów metalowych. Kraków 2009. [10] B. Śnieżyński, Recommendation system using multistrategy inference and learning, in: S. Piotr, Janusz Szczepaniak, Kacprzyk, Adam Niewiadomski (Eds.), Advances in Web Intelligence, volume 3528 of Lecture Notes in Computer Science, Springer, Berlin Heidelberg, 2005, pp. 421–426.
archives of civil and mechanical engineering 14 (2014) 25–31
[11] A.O. Erkimbaev, A.B. Zhizhchenko, Zitserman V. Yu., G.A. Kobzev, E.E. Son, A.N. Sotnikov, Integration of databases on substance properties: approaches and technologies, Automatic Documentation and Mathematical Linguistics 46 (4) (2012) 170–176 Allerton Press, Inc..
31
[12] T. Priyavrat, Competitive foundry through integration of TIPS, Indian Foundry Journal 58 (1) (2012) 37 January. [13] J. Prat, R. Stötzel, I. Yilmaz, Verminderte Blattrippenbildung auch ohne feuerfeste Schlichte, Neues Sandadditiv, red, Giesserei-Verlag, Düsseldorf, 2012.
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/acme
Original Research Article
The use of LPR (logic of plausible reasoning) to obtain information on innovative casting technologies S. Kluska-Nawareckaa, B. S´niez˙yn´skib, W. Paradab, M. Lustofinb, D. Wilk-Kołodziejczykc,n a
Foundry Research Institute in Cracow, Zakopiańska Street 73, Kraków, Poland AGH University of Science and Technology, Department of Computer Science, Mickiewicza 30, 30-059 Kraków, Poland c Andrzej Frycz Modrzewski Krakow University, Department of Applied Informatics, Gustawa Herlinga-Grudzińskiego 1, 30-705 Kraków, Poland b
art i cle i nfo
ab st rac t
Article history:
The paper presents a methodology for the application of a formalism of the logic of
Received 4 January 2013
plausible reasoning (LPR) to create knowledge about a specific problem area. In this case,
Accepted 30 May 2013
the methodology has been related to the task of obtaining information about the
Available online 5 June 2013
innovative casting technologies. In the search for documents, formulas created in terms
Keywords:
of LPR have a much greater expressive power than the commonly used keywords. The
Logic of plausible reasoning
discussion was illustrated with the results obtained using a pilot version of the original
Cast foundry
information tool.
Intelligent information system
& 2013 Politechnika Wroc"awska. Published by Elsevier Urban & Partner Sp. z o.o. All rights reserved.
Methodology for the application Innovative casting technologies
1.
Introduction
Currently, there are many search engines that allow us to obtain information from open sources. However, it appears that the use of a universal search engine usually leads to identification of a large number of documents that do not meet the user needs, and therefore require painstaking selection carried out in a “manual” mode, which is particularly troublesome in the case of making a large amount of expertise, or when quick decisions are to be made. As a result, the problems of effective information and knowledge retrieval from the available (often distributed) sources have recently been the subject of intensive work of many research centres [3,5,11,12]. The results obtained in this respect clearly show that a significant increase in search
efficiency can be achieved through the creation of solutions targeted at a specific area of knowledge and tailored to the specific sources, in which the acquisition is to be carried out. It is important for the search engines to be equipped with intelligent mechanisms (procedures) that allow for a possibly precise assessment of the contents of the documents obtained. This article presents a proposal for a solution where the search for documents is carried out using a knowledge base constructed with an LPR formalism [logic of plausible reasoning] [2]. The use of this formalism seems to be interesting because of its intuitive nature (expressing the user's way of thinking) and the ability to assess the degree of certainty of the results. The originality of the concept lies in the fact that the LPR was previously mainly used in typical expert systems
n
Corresponding author. Tel.: +48 505 353 700. E-mail address: [email protected] (D. Wilk-Kołodziejczyk).
1644-9665/$ - see front matter & 2013 Politechnika Wroc"awska. Published by Elsevier Urban & Partner Sp. z o.o. All rights reserved. http://dx.doi.org/10.1016/j.acme.2013.05.011
26
archives of civil and mechanical engineering 14 (2014) 25–31
Expert
Domain knowledge
Knowledge base
User
User interface
Inference
Information sources
Selected documents
Fig. 1 – Schematic diagram of the intelligent search engine.
[1,4–9]. There is only one work [10] in which LPR is used for knowledge representation and inference in recommendation system, but that work focuses on multistrategy learning. Here the problem is related with the representation of knowledge oriented at an intelligent search of documents (knowledge), focussed on a specific problem area and machine learning is not considered. The solution disclosed here relates to the sourcing of information on innovative foundry technologies, but the methodology itself used for the creation of a knowledge base, and the procedures for its use are universal in nature and may be applied to any problem area. In terms of utility, in the studies carried out currently, the application developed by the AGH Department of Computer Science and Technology has been used. Its functionality includes both the creation of a knowledge base expressed in terms of LPR, as well as the inference engine and user interface. Here (for reasons of volume) any broader description of the computing aspects of this application has been dropped, focussing attention on the presentation of the principles of creating a knowledge base and procedures for obtaining documents of the desired content. Schematic diagram illustrating the operating concept of the test system is shown in Fig. 1. Knowledge base, expressed in the form of rules, hierarchy, similarities and statements, is used to search the knowledge sources in compliance with the needs identified by the user through the interface. The result is a set of documents with the specified degree of reliability, corresponding to the user's intuition. This procedure will be further presented in relation to documents describing the innovative foundry technologies.
Table 1 shows fragment of a list of such innovations prepared with the assistance of foundry engineers from the Foundry Research Institute. The next phase that belongs to the expert competence is to formulate a list of attributes describing main entities. It is worth noting that the attributes can be single- or multivalued. In the former case, the attribute has the nature of a unique description of the document, while in the latter case it can take several values from the given set. In the process of searching, usually, only some of the attributes are known, while others are defined by the inference process conducted according to the LPR rules. The list of attributes for the innovations is presented in Table 2. Creating a knowledge base consists in generating for each of the main entities a set of the corresponding formulas. In the described application it is limited to the use of only certain selected types of LPR formulas and these are outlined below.
2.1.
Statements
Statement is called the expression: VðO; X; ZÞ
This means that the attribute X of object O has a value of Z.
2.2.
Hierarchies—generalisation and specialisation
Hierarchies are a way of describing the relationship between concepts, which in this case refers to the level of generality. Accordingly, a formula for generalisation is used: A GEN Ai
2. Methodology of the description of innovation The first phase in the expert activity is to identify main entities which can be searched for. Knowledge base is built around these entities. In our example domain, the search is for innovative foundry technologies [13], and so a set of technologies is formulated, which the documents sought should refer to.
ð1Þ
ð2Þ
where A is a more general term (dominant) with respect to Ai and for specialisation: Ai SPEC A
ð3Þ
where Ai is the clarification of the concept of A. Instead of these two forms the following may be used: HðAi ; A; CÞ
ð4Þ
where C represents a context in which the hierarchy is considered.
archives of civil and mechanical engineering 14 (2014) 25–31
27
Table 1 – Fragment of a list of innovations. No.
Name of innovation
1 2 3 4 5
Environment-friendly technology for the reclamation of bentonite moulding sands. New simplified method for mechanical reclamation of moulding and core sands. Environment-friendly technology for the recycling of waste core sands. Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine. New control and measurement apparatus adapted to the EU standards used for the bending strength determination in sands for the cold box process. An upgraded technology for the continuous preparation of moulding sands in a two-pan mixer. etc.
6 etc.
Table 2 – List of attributes that describe documents related to innovation. Name of attribute
Type
Description
Innovation name Article Target project Project title Term of project Project contractor Implementer Keyword Description
Single-valued Multi-valued Single-valued Single-valued Multi-valued Single-valued Single-valued Multi-valued Single-valued
Name of innovation; every innovation has got a name Numbers of articles, each article can also indicate a file or a link Reference number of a target project Title of project Years in which the project was implemented, recorded in “YYYY” format Name of company carrying out the project Project implementer Keywords related to innovation Description of innovation
As a clarification of the concept of the hierarchy, numerical factors are often incorporated determining typicality (τ) of the given relationship and the degree of dominance (δ) of the subordinate concept. Table 3 shows a fragment of the hierarchy statement established for the case of innovative technologies. In this table, the symbol H denotes hierarchical relationships, while words written in italics define variables for which values of the appropriate attributes are substituted.
2.3.
The similarity and dissimilarity
The similarity or dissimilarity of terms is written as Ai SIMAj : s
ð5Þ
Ai DISA2
ð6Þ
where Ai, Aj—terms subject to comparison, s—the degree of similarity or dissimilarity. Often the context in which the comparison is made is also referred to. We use the following notation to represent similarity: SðAi ; Aj ; CÞ : s
ð7Þ
of document searching, the rule expressing the implication can be written in a general form as: ½VðU; chose A; ZÞ; VðE; A; ZÞ-VðU; matches A; EÞ : 1; 0
ð8Þ
where variables used in statements have the following meaning: U—user, A—name of the attribute, Z—value of the attribute, E—main entity (defined in phase one), representing innovation in example domain. The implication expressed in the form of Eq. (8) has the following interpretation: from the statement that user (U) adopted for the attribute (A) the predetermined value (Z), and from the statement that main entity (E) is described with attribute (A) of the value (Z), the statement follows that main entity (E) matches (fits) user (U) in the context of the attribute (A) and to the extent estimated by the coefficient of the value 1.0. Selected implications describing the considered set of innovations are presented in Table 5. I is used instead of E to emphasise that entities represent innovations.
2.5.
Search mode and reliability of results
where C denotes context. Fragment of knowledge base regarding the similarity of concepts under consideration was compared in Table 4. As before, italics in this table refer to the variables for which the values of attributes have been substituted.
In ongoing experiments, two types of search procedures were used, i.e. fast mode and advanced mode. In the former case, the objective of the reasoning has the form of the following statements:
2.4.
VðE; A; ZÞ
Implications
A substantial part of the contents of the knowledge base is made of implications used in the process of reasoning. They define the relationships between the statements. In the case
ð9Þ
where E represents main entity defined in the first phase, A is attribute name and Z is its value. Inference is executed for every attribute chosen by the user and the goal is to find the entities which have attribute values provided.
28
archives of civil and mechanical engineering 14 (2014) 25–31
Table 3 – Selected hierarchies in the description of innovation. Rule
Typicality and dominance
Description
H(article_no. article, name_innovation)
τ¼ 0.8 δ ¼ 0.5 τ¼ 0.8 δ ¼ 0.5 τ¼ 0.8 δ ¼ 0.5 τ¼ 0.8 δ ¼ 0.6 τ¼ 0.8 δ ¼ 0.5 τ¼ 0.8
Each article is a type of article in the context of the name of a specific innovation Each article is a type of article in the context of the project title of a specific innovation Name of the project contractor is a type of the innovation contractor in the context of the project title Each keyword associated with innovation is a type of the keyword in the context of the article Each keyword associated with innovation is a type of the keyword in the context of the project contractor Each keyword associated with innovation is a type of the keyword in the context of the project title
H(article_no. article, title_project) H(name_contractor, project contractor, project title) H(key_word, keyword, article) H(key_word, keyword, project contractor) H(key_word, keyword, project title)
δ ¼ 0.5
Table 4 – Selected formulas of similarities. Rule
Degree of similarity
Description
S(key_word_1, key_word_2, innovation) S(key_word_1, key_word_2, project title) S(key_word_1, key_word_2, project contractor) S(article_1, article_2, innovation) S(article_1, article_2, project contractor) S(name_contractor, name_implementer, project title)
s¼ 0.8
Keyword 1 is similar to keyword 2 in the context of the same innovation
s¼ 0.8
Keyword 1 is similar to keyword 2 in the context of the same project title
s¼ 0.8
Keyword 1 is similar to keyword 2 in the context of the same project contractor
s¼ 0.8 s¼ 0.8
Article 1 is similar to article 2 in the context of the same innovation Article 1 is similar to article 2 in the context of the same project contractor of the same innovation Contractor name is similar to the name of the implementer in the context of the project title of the same innovation
s¼ 0.8
Table 5 – Selected implications. Implication
Description
[V(U, ‘chose keyword, Z), V(I, keyword, Z)] -V(U, matches keyword, I):1.0 [V(U, chose completion date, Y), V(I, completion date, Y)]-V(U, matches completion date, I):1.0 [V(U, chose article, M), V(I, article, M)]-V(U, matches article, I):1.0 [V(U, chose project title, N), V(I, project title, N)]-V(U, matches project title, I):1.0 [V(U, chose innovation name, J), V(I, innovation name, J)]-V(U, matches innovation name, I):1.0
If user U chose keyword Z and innovation I has a link with keyword Z, then innovation I matches user U in the context of the keyword If user U chose completion date Y and innovation I has a link with completion date Y, then innovation I matches user U in the context of the completion date If user U chose article M and innovation I has a link with article M, then innovation I matches user U in the context of the article If user U chose project title N and innovation I has a link with project title N, then innovation I matches user U in the context of the project title If user U chose innovation name J and innovation I has a link with innovation name J, then innovation I matches user U in the context of the innovation name
In advanced mode the following expression is used: VðU; matches A; EÞ
ð10Þ
Here also inference is executed for every attribute chosen. Consequently, in the latter case, the entire knowledge base is sought. Inference process may use implication (8) to find matching entities and can infer statements describing user preferences and main entities. In the former case only fragments of the
knowledge base are used (attributes of main entities) which of course has a significant effect on the time taken for response. The certainty factor assumes values from the range (0,1] and is calculated from the following formula: q ¼ ðq1 þ q2 þ ⋯ þ qn Þ=n
ð11Þ
where: qi—certainty value for the i-th chosen attribute, n—number of attribute values selected by the user.
29
archives of civil and mechanical engineering 14 (2014) 25–31
The certainty factor q is the basis for the ranking with preference for responses with the largest values. The certainty of the attribute values (qi) are calculated by the inference module based on the knowledge base and certainties provided by the user in the question definition.
3.
Test results
Completed application was tested on a set of documents created from the descriptions of research and implementation projects carried out at the Foundry Research Institute. The extensive material obtained as a result of the conducted experiments includes exploration in fast and advanced mode, taking into account the values of the 1–9 attributes. The following are extracts from these experiments that adequately illustrate the nature of the obtained results. Comparing the documents referred to in Table. 6 (1 attribute) with the contents of Table 7 (3 attributes) it can be concluded that increasing the number of the included attribute values will generally result in their more accurate
selection but does not always translate into increased certainty of the result. Analysing the impact of the search mode, it is clear that the advanced mode usually leads to the selection of more documents than the fast mode, but the certainty of the result for the additionally indicated documents is usually lower. It is worth noting that increasing the number of attributes included in the search significantly affects the increase in computation time (an exponential relationship). In assessing the overall results of the tests, it can be concluded that they have demonstrated the correct operation of the tool constructed and allowed observing certain regularities found in the search for innovation.
4.
Conclusions
This paper presents the concept of using LPR to search for the text documents with a given theme. It seems that the proposed methodology for the description of documents
Table 6 – Searching for innovations based on a single-attribute value. Attribute
Search mode
Searched innovations
Certainty of result
Attribute name: Description; Value: Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine; Certainty: 0.75
Fast
Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine to make moulds in bentonite sands Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine to make moulds in bentonite sands Implementation of the reclamation process of selfsetting sands according to a new vibration-fluidised bed method Development and implementation of a technology for the manufacture of a new grade of foundry resin with reduced content of free formaldehyde and its application in selected domestic foundries Environment-friendly technology for the reclamation of bentonite moulding sands New simplified method for mechanical reclamation of moulding and core sands Environment-friendly technology for the reclamation of bentonite moulding sands New simplified method for mechanical reclamation of moulding and core sands Environment-friendly technology for the reclamation of bentonite moulding sands Environment-friendly technology for the reclamation of bentonite moulding sands Environment-friendly technology for the reclamation of bentonite moulding sands Implementation of the reclamation of self-setting sands according to a new vibration- fluidised bed method Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine
0.6
Advanced
Attribute name: Article; Value: article_5; Certainty: 1.0
Fast
Advanced
Attribute name: Innovation name; Value: Environment-friendly technology for the reclamation of bentonite moulding sands; Certainty: 0.75
Fast Advanced
0.499
0.224
0.224
0.224 1.0 1.0 1.0 1.0 1.0 0.75 0.375
0.375
30
archives of civil and mechanical engineering 14 (2014) 25–31
Table 7 – Searching for innovations based on the values of three attributes. Attributes
Type of search
Innovations search
Certainty of result
Attribute name: Innovation name; Value: New technology for casting cylinder sleeves in spinning metal moulds;
Fast
New technology for casting cylinder sleeves in spinning metal moulds New simplified method for mechanical reclamation of moulding and core sands Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine to make moulds in bentonite sands … New technology for casting cylinder sleeves in spinning metal moulds New simplified method for mechanical reclamation of moulding and core sands Development and implementation in production of an innovative design of the FT-65 impulse-squeeze moulding machine Development of design and implementation in production of a new versatile 20 dcm3 capacity core shooter to make cores by various technologies Implementation of the reclamation process of self-setting sands according to a new vibration- fluidised bed method Implementation of processes for moulding sand preparation, treatment and circulation and for casting automotive parts from ductile iron on a pilot line using prototype stand for impulse-squeeze moulding …
0.4666
Certainty: 1.0 Attribute name: Project contractor; Value: FERREX Sp. zo.o. Poznań; Certainty: 1.0
Attribute name: Project title; Value: Two-post FT-65 impulse-squeeze moulding machine; Certainty: 1.0
Advanced
greatly increases the power of its expression in relation to commonly used search based only on keywords. The attributes used in the described approach are a certain generalisation of keywords, and moreover, use of LPR formalism allows taking into account the context in which these attributes are present. As a problem area for which the proposed methodology has been used, the search for innovation in casting technologies was chosen as characterising well enough the specific nature of this class of the tasks. The designed pilot solution for the information tool makes it possible to carry out the experimental verification of the effectiveness of the approach. The tests performed are of course preliminary, but create successful outcomes for future work in this area. In the future it is expected to examine the effectiveness of information retrieval with a much larger collection of source documents, and use data sources of a distributed character, the Internet included.
Acknowledgements Scientific work financed from funds for the scientific research as an international project. Decision no. 820/N-Czechy/2010/0
references
[1] James L. Beck, Bayesian system identification based on probability logic, Structural Control and Health Monitoring1545-225517 (7) (2010) 825–847.
0.4666 0.3333
0.1333 0.4666 0.4666 0.3333
0.2999
0.2999 0.2999
0.1333
[2] A. Collins, R.S. Michalski, The logic of plausible reasoning: a core theory, Congnitive Science 13 (1) (1989) 1–49. [3] J. David, M. Vrožina, Z. Jančíková, Determination of crystallizer service life on continuous steel casting by means of the knowledge system, WSEAS Transactions on Circuits and Systems 10 (10) (2011) 359–369 ISSN:1109-2734 359. [4] S. Kluska-Nawarecka, Z. Górny, D. Wilk-Kołodziejczyk, A. Smolarek-Grzyb, The logic of plausible reasoning in the diagnosis of castings, Archives of Metallurgy and Materials1733-3490375–380. [5] I. Olejarczyk, A. Adrian, H. Adrian, B. Mrzygłód, Algorithm for controlling of quench hardening process of constructional steels, Archives of Metallurgy and Materials17333490171–179 Bibliogr. s. 179. [6] M. Saad KhanAn, T.R. Soomro, Overview of human plausible reasoning, International Journal of Computer Applications 45 (18) (2012) 0975–8887. [7] B. Śnieżyński, Basic semantics of the logic of plausible reasoning, in: Mohand-Saïd Hacid et al.(Ed.), Lecture Notes in Computer Science; vol. 2366. Lecture Notes in Artificial Intelligence, Springer-Verlag, Berlin etc., 2002, pp. 176–184 Bibliogr. s. 184, Abstr.3. [8] B. Śnieżyński, Proof searching algorithm for the logic of plausible reasoning, in: A.Kłopotek Mieczysław, T. Sławomir (Eds.), Wierchoń, Krzyszt of Trojanowski; Polish Academy of Sciences, Instytute of Computer Science—Berlin; Heidelberg; New York, Springer-Verlag, 2003, pp. 393–398 (Advances in Soft Computing; Series 4240). [9] D. Wilk-Kołodziejczyk, Budowa algorytmów wiedzy dla potrzeb diagnostyki wad wyrobów metalowych. Kraków 2009. [10] B. Śnieżyński, Recommendation system using multistrategy inference and learning, in: S. Piotr, Janusz Szczepaniak, Kacprzyk, Adam Niewiadomski (Eds.), Advances in Web Intelligence, volume 3528 of Lecture Notes in Computer Science, Springer, Berlin Heidelberg, 2005, pp. 421–426.
archives of civil and mechanical engineering 14 (2014) 25–31
[11] A.O. Erkimbaev, A.B. Zhizhchenko, Zitserman V. Yu., G.A. Kobzev, E.E. Son, A.N. Sotnikov, Integration of databases on substance properties: approaches and technologies, Automatic Documentation and Mathematical Linguistics 46 (4) (2012) 170–176 Allerton Press, Inc..
31
[12] T. Priyavrat, Competitive foundry through integration of TIPS, Indian Foundry Journal 58 (1) (2012) 37 January. [13] J. Prat, R. Stötzel, I. Yilmaz, Verminderte Blattrippenbildung auch ohne feuerfeste Schlichte, Neues Sandadditiv, red, Giesserei-Verlag, Düsseldorf, 2012.