ENBIS-15 in Prague

6 – 10 September 2015; Prague, Czech Republic Abstract submission: 1 February – 3 July 2015

My abstracts

 

The following abstracts have been accepted for this event:

  • Scheduling Preventive Maintenance on a Wind Turbine Using Statistical Process Control

    Authors: A. Di Bucchianico (Eindhoven University of Technology), S. Kapodistria (Eindhoven University of Technology)
    Primary area of focus / application: Reliability
    Secondary area of focus / application: Process
    Keywords: Preventive maintenance, Statistical Process Control, Wind turbine
    Submitted at 1-May-2015 15:38 by Alessandro Di Bucchianico
    Accepted (view paper)
    7-Sep-2015 10:20 Scheduling Preventive Maintenance on a Wind Turbine Using Statistical Process Control
    Inspired by a physical model for the power output of the wind turbine, we designed a statistical process monitoring procedure for scheduling preventive maintenance. The process is monitored with a control chart with the purpose of quickly detecting shifts to inferior operational states of the wind turbine due to the occurrence of unobservable assignable causes. At the same time, the information collected from the monitoring process is used to determine the overall operational state of the wind turbine, aka the degradation process of the asset. This degradation process moves in a continuous manner be-tween two extremes (perfect condition and failure) with random measurement errors. Although, these three models, the physical model, the statistical model and the stochastic model, are obviously related, they have been typically treated in the literature independently. We will highlight the underlying connections between the three models and present a general mathematical model that can be used for the optimal identification of what constitutes sufficient evidence of imminent failure, so as to perform preventive maintenance, taking into account a maintenance cost structure. This research is conducted within the national Dutch research projects CAMPI (Coordinated Advanced Maintenance and Logistics Planning for the Process Industries) and DAISY4OFFSHORE (Dynamic Asset Information System for Offshore Wind Farm Optimisation).
  • ENBIS Live - Open Problem Session

    Authors: Christian Ritter (Université Catholique de Louvain)
    Primary area of focus / application: Other: Open Problem Session
    Keywords: Industrial statistics, Consulting, Open problems
    Submitted at 3-May-2015 12:16 by Christian Ritter
    Accepted
    8-Sep-2015 17:00 Special Session: Problem Solving Session
    Do you have a burning problem in industrial and business statistics and need input?
    Would you like to help others who have such problems?
    Then don't miss the ENBIS live session.

    During the session, two or three participants will briefly describe their burning problems. I will help in formulating, asking stimulating questions, and summarizing the answers. Participants who attended a similar session at ENBIS2014 in Linz enjoyed it a lot. They said that by participating either as a presenter or in the audience, they got much deeper insights into the problems than by listening to usual talks.

    If you would like to be one of the persons presenting a problem, it would be nice if you could contact me (christian.ritter@ridaco.be) before the conference. If you just want to be part of the audience just come but then you must participate actively in the discussion.
  • A New Model for Duration and Segmentation of Maintenance Activities in Semiconductor Fabs

    Authors: Diamanta Benson-Karhi (The Open University of Israel), Itai Regev (Intel)
    Primary area of focus / application: Reliability
    Keywords: Availability, Downtime, Preventive maintenance, Productivity, Segmentation, Semi-conductor
    Submitted at 4-May-2015 09:33 by Diamanta Benson-Karhi
    Accepted
    8-Sep-2015 16:15 A New Model for Duration and Segmentation of Maintenance Activities in Semiconductor Fabs
    Preventive maintenance (PM) plays an essential role in maintaining effective control of the equipment in a semiconductor plant, ensuring that it performs within the control limits required in semiconductor manufacturing. Accordingly, PM execution, both duration-wise and quality-wise has significant impact on equipment performance. Despite the need to have a clear understanding on process inefficiencies or improvement opportunities in the PM execution, it is complicated to generate automated segmented view of the PM duration due to the manual textual data entry by human operators, thus factories are forced to refer to PM as ‘One PM’ instead of segmenting its parts. In this research, a model is proposed to describe expected PM duration and to divide the ‘One PM’ concept into smaller parts in such a way that does not require a change in human practice. The segmentation, done automatically by sequencing the production lots run on the tool (semiconductor equipment), results in a clear overview of time losses in the PM and of improvement opportunities.
  • Flexible Econometric Modelling Based on Sparse Finite Mixtures

    Authors: Sylvia Fruhwirth-Schnatter (WU Vienna University of Economics and Business)
    Primary area of focus / application: Other:
    Secondary area of focus / application: Other:
    Keywords: Finite mixtures, Sparsity, Selecting number of compenents, Identification
    Submitted at 4-May-2015 11:39 by Sylvia Fruhwirth-Schnatter
    Accepted
    8-Sep-2015 12:40 Flexible Econometric Modelling Based on Sparse Finite Mixtures
    There has been a tremendous increase in applied work both in statistics as well as in econometrics using infinite mixtures based in particular on Dirichlet process priors. However, going to infinity makes the mathematics much more complicated, in particular, for applied statisticians and econometricians.

    This talk investigates the concept of sparse finite mixture modelling, which is based on a shrinkage prior on the weights that removes all redundant components automatically. The choice of the hyperparameters of this prior is based on recent asymptotic results by Rousseau and Mengersen (2011). The sparsity prior provides an automatic tool to select the number K of components and avoids the cumbersome computation of the marginal likelihood for each K.
    Furthermore, it is shown how the label switching problem could be solved using the framework of sparse finite mixtures. In contrast to infinite mixtures, this allows identification of component-specific parameters and classification.

    This approach is applied to various issues in statistical modelling such as choosing the number of latent classes in a latent class model, model-based clustering based on finite mixtures of normals, switching regression models, and choosing a flexible link function in binary data modelling. Finally, the sparse finite mixture approach is compared to infinite mixtures based on Dirichlet process priors.
  • Projection Properties of Blocked Non-Regular Designs

    Authors: Shahrukh Hussain (Norwegian University of Science and Technology, Department of Mathematics), John Sølve Tyssedal (Norwegian University of Science and Technology, Department of Mathematics)
    Primary area of focus / application: Design and analysis of experiments
    Secondary area of focus / application: Design and analysis of experiments
    Keywords: Non-regular designs, Screening, Blocking, Projectivity
    Submitted at 4-May-2015 14:52 by Shahrukh Hussain
    Accepted
    8-Sep-2015 10:30 Projection Properties of Blocked Non-Regular Designs
    The paper is concerned with projection properties of two-level minimum resolution IV (MinResIV) designs and Plackett-Burman (PB) designs when blocked. MinResIV designs with 10-20 runs and PB designs with 12 and 20 runs are investigated. These designs are popular screening designs due to their run size efficiency and their property of being projectivity P=3 designs which means that it is possible to estimate all main effects and all interactions of any three factors. One purpose of the investigation is to find out if this property can be maintained when blocked, and also to find the best blocking arrangement so that the designs can maintain P=3 or near P=3 when blocked. For orthogonally blocked design we used method of replacement as suggested by Wu & Hamada (2000) for regular designs. For non-orthogonal blocked design, we show how the loss of information due to blocking arrangements is related to an algorithm defined by Jacroux (2009) for two-level nonregular designs. We present an alternative algorithm based on the Ds-efficiency of projections and investigate what kind of projective properties are preserved under various blocking regimes.
  • Teaching Statistical Thinking - A MOOC Supported by JMP

    Authors: Volker Kraft (JMP)
    Primary area of focus / application: Other: ENBIS Goes MOOC (by Jean-Michel Poggi)
    Keywords: MOOC, E-Learning, Distance-learning, Teaching statistics
    Submitted at 11-May-2015 21:10 by Volker Kraft
    Accepted
    8-Sep-2015 15:55 Teaching Statistical Thinking - A MOOC Supported by JMP
    During winter semester 2014/2015 Duke University (Durham, North Carolina, USA) launched their first installment of a MOOC on Teaching Statistics: “Teaching Statistical Thinking: Part 1 Descriptive Statistics.” The course was hosted by Coursera Inc., and JMP was the featured software of this course and used in their analysis modules.

    This MOOC is designed with high school teachers in mind, but it should also enable any global citizen to understand the basic descriptive statistics presented by the media from breakfast to bedtime. Part I of this course covers descriptive statistics for a single variable and for the relationship between two variables. The course is organized around "core principle videos" that discuss the statistical content, along with satellite videos discussing resources, pedagogy, and the analysis of data using JMP. The course is taught in three units over five weeks, with more than 5000 registrants who attended the first
    class. Updates and extensions of this MOOC are planned for the future.

    We will share and discuss the experience from this course.