ALGORITMI IN TEHNIKE PODATKOVNEGA RUDARJENJA NA BAZI PROCESNIH PARAMETROV

Colja, Sara

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	ALGORITMI IN TEHNIKE PODATKOVNEGA RUDARJENJA NA BAZI PROCESNIH PARAMETROV
Authors:	ID Colja, Sara (Author) ID Benkovič, Dominik (Mentor) More about this mentor... ID Cvahte, Peter (Comentor)
Files:	UNI_Colja_Sara_2011.pdf (1,27 MB) MD5: 424CB9341DF94C67EC1DC7A7378B5302 PID: 20.500.12556/dkum/306bf7c9-7792-4aff-bad3-c66b9f72ac6d
Language:	Slovenian
Work type:	Undergraduate thesis
Organization:	FNM - Faculty of Natural Sciences and Mathematics
Abstract:	V diplomskem delu je predstavljen proces podatkovnega rudarjenja, njegovi algoritmi, tehnike in uporaba v praksi. V prvem delu se seznanimo s teorijo podatkovnega rudarjenja. Omenjene so tehnike podatkovnega rudarjenja in nekateri najbolj znani algoritmi. Podrobneje je predstavljen algoritem nevronskih mrež, ki se uporabi v praktičnem primeru. V drugem delu je po korakih splošne metode podatkovnega rudarjenja, predstavljene v prvem delu, predstavljen konkreten poslovni problem, ki ga rešujemo s podatkovnim rudarjenjem. Na bazah podatkov podjetja Impol sta zgrajena modela za iskanje povezav med kemijsko sestavo zlitine EN AW-7075 (interna oznaka PD30) in njenimi mehanskimi lastnostmi. Po združitvi različnih baz in agregiranju podatkov je bilo uporabljenih 675 množic zgodovinskih podatkov za zlitino PD30. Model je bil zgrajen z orodjem SPSS Modeler, s feed-forward nevronsko mrežo in vzvratnim širjenjem napake. Naučeni nevronski mreži napovedujeta mehanske lastnosti napetost tečenja (R0,2), natezna trdnost (Rm) in raztezek (A), kot funkcijo procesnih parametrov. Točnost napovedi modela nevronske mreže za napetost tečenja je 84,8%, točnost napovedi modela za natezno trdnost in raztezek pa 91,8%. S predstavljenima modeloma nevronskih mrež je pokazano, da lahko podjetje Impol razvije model za ocenjevanje končnih mehanskih lastnosti, kot funkcijo procesnih parametrov. S tem je omogočena optimizacija procesne poti glede na produktivnost in kvaliteto.
Keywords:	podatkovno rudarjenje, nevronska mreža, odločitvena drevesa, vzvratno širjenje napake, metoda padajočih gradientov
Place of publishing:	Maribor
Publisher:	[S. Colja]
Year of publishing:	2011
PID:	20.500.12556/DKUM-18116
UDC:	51(043.2)
COBISS.SI-ID:	18312968
NUK URN:	URN:SI:UM:DK:PFZHWQLH
Publication date in DKUM:	04.05.2011
Views:	4106
Downloads:	519
Metadata:
Categories:	FNM
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Secondary language

Language:	English
Title:	DATA MINING TECHNIQUES AND ALGORITHMS ON DATABASE OF PROCESS PARAMETERS
Abstract:	In the graduation thesis we presented the data mining process, data mining techniques and algorithms on database of process parameters. The thesis begins with a short introduction of data mining process. We described the best know techniques and algorithms of data mining. In more details we presented algorithms of neural networks, wich we used to work on practical business problem. In the second part is described through the generic data mining method an alternative approach to the physical modeling, the artificial intelligence approach, based on the neural networks. Data for data mining process were collected in company Impol. After merge different database and aggregate data, 675 sets of complete history data were collected for alloy EN AW-7075 (internal use Impol as PD30). For building the neural network model we used SPSS tool, with one of most popular architecture Multilayer Feedforward with Backpropagation learning. This neural networks are capable of predicting yield strength (R0,2), tensile strength (Rm) and elongation (A) as function of process parameters. The accuracy of neural network model for yield strength is 84,8%, the accuracy of neural network model for tensile strength and elongation is 91,8%. With the represented models of neural network we show, that the company Impol can develop a model for estimation of the final product properties as a function of the process parameters. This allows optimizing the process path with respect to productivity and quality in the perspective.
Keywords:	data mining, neural networks, decision trees, backpropagation learning, method of gradient descent

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back