ENBIS-13 in Ankara

15 – 19 September 2013 Abstract submission: 5 February – 5 June 2013

My abstracts

 

The following abstracts have been accepted for this event:

  • Robust Design Experiment on a Resonant System

    Authors: Jan Åkerström (XRTECH), Bjarne Bergquist (Luleå Technical University)
    Primary area of focus / application: Design and analysis of experiments
    Keywords: Design of experiments, Mechanical design, Robust design, Two level fractional factorials, Resonant nonlinear systems, Multidimensional resonance margin, Multiphysics simulation
    Submitted at 30-May-2013 16:51 by Jan Åkerström
    Accepted
    17-Sep-2013 16:55 Robust Design Experiment on a Resonant System
    An assumption for factorial experimental design is that the coupling between the factors and responses are linear, or at least of low order. These assumptions do not generally hold, and may be seriously violated for some dynamic systems.
    Here, an experiment performed on the design of an exhaust gas economizer (an heat exchanger), where improvements in an welding production process resulted in out of tolerance component operating noise. To trace the root cause a factorial experiment was performed on eight design factors to screen which ones, if any, were causing the noise. Besides the standard dBA readings, an octave band sound pressure level measurement was made for each run. Factor effect calculations on these octave bands was used to identify root cause driving mechanisms, sound energy generation and dissipation paths. Parameter settings ensuring a stable noise free economizer was identified. A rudimentary fluid simulation (CFD) was made to validate the experiment root cause mechanism assumption of coinciding component resonance and vortex shredding frequencies. Design changes eliminating root causes reduced the generated sound levels from 120 dBA down to 94 dBA.
    An assumption for factorial experimental design is that the coupling between the factors and responses are linear, or at least of low order. These assumptions do not generally hold, and may be seriously violated for some dynamic systems.
    Here, an experiment performed on the design of an exhaust gas economizer (an heat exchanger) is described. A background to the experiment is that a redesign of the welding production process had resulted in severe component operating noise. A factorial experiment was performed on eight design factors to screen which factors, if any, were causing the noise. The measured responses were the standard dBA readings, and the octave band sound pressure profile.. Factor effect calculations on these octave bands was used to identify root cause driving mechanisms, sound energy generation and dissipation paths. Parameter settings ensuring a stable noise free economizer was identified. A rudimentary fluid simulation (CFD) was made to validate the root cause mechanism assumption of coinciding component resonance and vortex shredding frequencies. Design changes eliminating root causes reduced the generated sound levels from 120 dBA down to 94 dBA.
    The experiment is an example of a system with highly non-linearity behaviour caused by resonance mechanisms and an example of a bimodal robust or unstable system with occasional borderline cases spontaneously switching between stable and unstable states. In robust subspaces traditional linear analysis could be effectively performed.
    The experiment resulted in multiple design solutions. However, the sound profile response was ill suited for traditional experimental analysis methods focusing on scalar responses and assuming linearity throughout the factor space, and an alternative analysis method is proposed here.
  • The Effect of Autocorrelation on the Hotelling T2 Control Chart

    Authors: Erik Vanhatalo (Luleå University of Technology), Murat Kulahci (Technical University of Denmark), Bjarne Bergquist (Luleå University of Technology)
    Primary area of focus / application: Process
    Keywords: Statistical process control (SPC), Control charts, Autocorrelation, Multivariate data, Time series modeling, Simulation.
    Submitted at 30-May-2013 16:57 by Erik Vanhatalo
    Accepted
    16-Sep-2013 16:05 The Effect of Autocorrelation on the Hotelling T2 Control Chart
    One of the basic assumptions for traditional univariate and multivariate control charts is that the data is independent. However, many times the data are serially dependent (autocorrelated) and crosscorelated due to, for example, frequent sampling and process dynamics. It is evident from the literature that autocorrelation affects the false alarm rate and the shift detection capability of traditional control charts. However, we have not found a study of how the false alarm rate and the shift detection capability of the Hotelling T2 control chart is affected by various auto and cross correlation structures for different magnitudes of shifts in the process mean.

    Average Run Lengths (ARL) for the Hotelling T2 control chart for different shift sizes and various auto and cross correlation structures are compared using simulated data. We construct control charts using the original data as well as charts using the residuals from fitted multivariate time series models. The simulations are based on a first-order vector autoregressive structure, VAR(1).
  • Power Analysis of Methods for Analysing Unreplicated Factorial Experiments

    Authors: Bjarne Bergquist (Luleå University of Technology), Erik Vanhatalo (Luleå University of Technology)
    Primary area of focus / application: Design and analysis of experiments
    Keywords: Simulation, Power Analysis, Unreplicated Factorials, Governing Principles
    Submitted at 30-May-2013 16:58 by Bjarne Bergquist
    Accepted
    16-Sep-2013 14:45 Power Analysis of Methods for Analysing Unreplicated Factorial Experiments
    Several methods for formal analysis of unreplicated factorial type experiments have been proposed in the literature. Based on a simulation study, five formal methods found in the literature based on the effect sparsity principle have been studied. The simulation included 23 and 24 type factorials with one, two, or four active effects. The simulated signal-to-noise ratios for the effects were all between two and four, and the Type I and Type II errors of the analysis methods were analysed. Preliminary results show that Bayesian models are more powerful in these contexts, especially if informative priors based on the effect heredity and effect hierarchy principles are used.
  • Design and Analysis of Electrospinning Experiments

    Authors: Antonio Pievatolo (CNR-IMATI), Ettore Lanzarone (CNR-IMATI), Laura Martìn Fernandez (CNR-IMATI), Fabrizio Ruggeri (CNR-IMATI)
    Primary area of focus / application: Design and analysis of experiments
    Keywords: Electrospinning, Design of experiments, Proportional odds models, Operating region
    Submitted at 30-May-2013 19:51 by Antonio Pievatolo
    Accepted (view paper)
    16-Sep-2013 15:25 Design and Analysis of Electrospinning Experiments
    Electrospinning is a method for producing fibres on the nano scale, using an electrical charge. A polymeric solution is pushed through a spinneret at a certain distance from a collector, then by a known process determined by the electrical charge a liquid jet forms which changes into a filament which is finally deposited on the collector. Factor such as solution density, voltage, distance, solution flow rate affect the quality of the deposited fibres, which is measured on a qualitative ordinal scale. We examine the result of a preliminary experiment and of a second planned experiment with a single spinneret, using a proportional odds model, to understand the relationship between the experimental setting and the fibre quality. The purpose is to find a stable operating region and to plan further experiments with multiple spinnerets.
  • Forecasting Future Trends in Turkey Housing Market by Using a Vector Error Correction Model

    Authors: Esra Akdeniz Duran (Istanbul Medeniyet University)
    Primary area of focus / application: Modelling
    Keywords: Turkey House Price index, Real estate, Macroeconomic indicators, Forecasting, Cointegration, Vector Error Correction Model
    Submitted at 30-May-2013 22:10 by Esra Akdeniz Duran
    Accepted
    17-Sep-2013 09:20 Forecasting Future Trends In Turkey Housing Market By Using A Vector Error Correction Model
    The housing market plays a vital role in a nation’s economy. Reidin.com country-wide house price index is intended to be a trusty indicator of the housing market state in Turkey. The primary purpose of this paper is to forecast Reidin.com house price index using a wide range of macroeconomic indicators such as volume of mortgage loans, consumer price index, monetary aggregate, Istanbul Stock Exchange REIT index, dolar/TL exchange rate, consumer confidence index, mortgage loan interest rate, current account balance, for the period June 2007 through January 2013. The relation between the housing market and macroeconomic activity has drawn special attention because housing investment is considered to be an important leading indicator of economic activity. We find that house price index and macroeconomic variables have cointegrating relations thus we utilized vector error correction model (VECM). Dynamic causal relationships between house price index and the macroeconomic indicators are also investigated. The impulse response functions and variance decomposition models revealed that shocks to consumer confidence index, dolar/TL exchange rate and current account balance have noticeable effects on changes in the Turkish housing market. The proposed model forecasts an increasing trend for the following 6 months period and 3-month forecasts are very accurate with the realized index values. To our knowledge, this paper is the first attempt to forecast country-wide house price index based on macroeconomic variables. The results of this study would help policy makers and property investors for creating more effective property management strategies in Turkey.
    References:
    Engle, Robert F., Granger, Clive W. J. (1987). Co-integration and error correction: Representation, estimation and testing, Econometrica, 55(2), 251-276.
    Hepşen, A. , Vatansever, M. (2011). Forecasting future trends in Dubai housing market by using Box-Jenkins autoregressive integrated moving average, International Journal of Housing Markets and Analysis, 4(3), 210-223.
    Hepşen, A. , Vatansever, M. (2012). Relationship between residential property price index and macroeconomic indicators in Dubai housing market, International Journal of Strategic Property Management, 16 (1), 71-84.
    Lütkepohl H. (2005). New Introduction to Multiple Time Series Analysis , Springer.
    Lütkepohl H., Kraetzig, M. (2004). Applied Time Series Econometrics, Cambridge University Press.
    Pfaff, B. (2008). VAR, SVAR and SVEC Models: Impletation Within R Package vars, Journal of Statistical Software, 27 (4).
    Pfaff, B. (2008). Analysis of Integrated and Cointegrated Time Series with R, Springer.
    Sari, R., Ewing, B. T., Aydin, B. (2007). Macroeconomic Variables and the Housing Market in Turkey, Emerging Markets Finance and Trade, 43 (5), 5-19.
  • Case Based Reasoning Applications to Financial Decision Making for IPO Trading

    Authors: Suat Kasap (Hacettepe University)
    Primary area of focus / application: Mining
    Keywords: Case based reasoning, Data mining, Forecasting, financial applications
    Submitted at 30-May-2013 22:50 by Suat Kasap
    Accepted
    16-Sep-2013 12:55 Case Based Reasoning Applications to Financial Decision Making for IPO Trading
    In this study, we investigate the case based reasoning (CBR) and its applications to financial decision making problems. More specifically, a CBR model to assist the investor to determine stock trend is developed. There are several ways to play the breakout. First is to buy in anticipation of the breakout, second is buying as soon as the breakout occurs and lastly is waiting for a pullback to the breakout point, which now acts as support. These strategies are examined and tested for the IPO stocks in the year 2008 -2012 for NASDAQ US stock markets. The first sale of stock by a company to the public is known as the Initial Public Offering (IPO). It's hard enough to analyze the stock of an established company. An IPO company is even trickier to analyze since there won't be a lot of historical information. There are two ways to make money with IPOs. First, you can jump in early and hope the stock has a big increase quickly and then dump your shares for a quick profit. The other way is to watch what is coming out and see if the stock is fairly priced. Find a price that you think is reasonable and grab the stock if you can get it at that price. By using the technical analysis charts and the historical data, the lower and higher prices and corresponding breakouts are obtained. The stocks that give profit are classified into some groups. With using and analyzing these classes required parameters are determined to simulate the IPO market.