web-based data mining for quenching data analysis
DESCRIPTION
Web-based Data Mining for Quenching Data Analysis. Aparna S. Varde, Makiko Takahashi, Mohammed Maniruzzaman, Richard D. Sisson Jr. Center for Heat Treating Excellence Worcester Polytechnic Institute Worcester, MA, USA. Introduction. - PowerPoint PPT PresentationTRANSCRIPT
Web-based Data Mining for Quenching Data Analysis
Aparna S. Varde, Makiko Takahashi, Mohammed Maniruzzaman, Richard D. Sisson Jr.
Center for Heat Treating ExcellenceWorcester Polytechnic Institute
Worcester, MA, USA.
Introduction
Web-based Data Mining Tool “QuenchMiner” being developed at CHTE, WPI
Purpose: Analysis of experimental data generated during quenching in the heat treating of materials
Supports CHTE Quench Probe System that gathers experimental time-temperature data
Functions Existing CHTE Database QuenchPAD on the Web Advanced Features e.g. querying complex data Decision Support System (DSS)
Phase I: Query Processing
Flat Files
Web Interface
Conversion Unit
Query Processor
Integrated Store (Relational Database System)
QuenchPAD
Complex Data
Raw Data
User Input User Output
SQL Query
SQL Result
QuenchPAD on the Web for worldwide access
Integral Store for complex data, flat files, raw data
Advanced Features for queries, graphs etc.
Phase II: Decision Support System
Web Interface
Semantic Analyzer
Decision-maker
Knowledge Base
Integral Store (RDBMS)
Data Miner
User Scenario Output to User
Analytical Output Sample Decisions
DataBackground Information
ExtractionRule-building
User Case Studies and Analysis
Data Mining to acquire knowledge, build rules
Decision-making using rules and cases
Data Mining
Discovering interesting patterns/trends in large data sets for guiding future decisions
Most Important step of Knowledge Discovery in Databases (KDD)
Data Mining Techniques: Association Rules, Decision Trees etc.
Rules and action paths fed into Knowledge Base to help decision-making
Association Rules
Statement of the type “X => Y”, where X and Y are events or conditions
Examples: High carbon content => More potential for distortion Excessive agitation => Excessively high cooling rate Use of Water Quenchant => Faster heat extraction
Rules built from analysis of data using statistical measures, probability and domain knowledge
Rules serve as basis for Decision Trees
Decision Trees
Representation of paths of action taken on occurrence of certain events
Example tree for sub-case of distortion
Suggests action, based on part geometry, to minimize distortion during quenching
SuspendVertically in the Quenchant
Geometry
ThinAndLong
Has SharpCorners
Variable Cross Section
AddRounds to the Ends
AdjustCR toThickestSection
Demo of QuenchMiner
Authorized users may get this from http://mpi.wpi.edu
Query Processing screens with results
DSS screens with sample analysis and decisions
Screen-dumps of Demo shown here
Current Status
QuenchMiner Query Processing (Phase I): Alpha Version with real data in Demo
QuenchMiner DSS (Phase II): Prototype with sample data in Demo
Integral Store (Data Mart) has been built
Knowledge Base (Rules and Decisions) is being built
Conclusions
QuenchMiner does Web-based Data Mining for the CHTE Quench Probe System
It Performs Query Processing for Simple and Complex data types
It will serve as a Decision Support System for CHTE member companies
Future Issues: Introducing Artificial Intelligence to build an Expert System