ENBIS-17 in Naples

9 – 14 September 2017; Naples (Italy) Abstract submission: 21 November 2016 – 10 May 2017

My abstracts

 

The following abstracts have been accepted for this event:

  • Statistical Challenges in Compact Models for LED Design

    Authors: Alessandro Di Bucchianico (Eindhoven University of Technology), Josephine Sari (Eindhoven University of Technology)
    Primary area of focus / application: Modelling
    Keywords: LED, Model order reduction, Compact models, Singular value decomposition, Dimension reduction, Nonlinear regression
    Submitted at 8-May-2017 13:07 by Alessandro Di Bucchianico
    Accepted (view paper)
    11-Sep-2017 11:10 Statistical Challenges in Compact Models for LED Design
    LED lamps are a game changer in the illumination industry, because of their superior technical properties with respect to energy efficiency and life span. Design of LED lamps is much more involved than for traditional lamps, since LED lamps offer many new technical illumination features. Since competition in this industry is fierce and electro-thermal-optical simulation of new LED designs is time consuming, it is essential to provide designers with models that allow for fast simulation so that design choices can be evaluated in a timely way.
    As part of the EU project Delphi4LED, so-called compact models are being derived based on combining physical laws with techniques from mathematical analysis and control theory under the name of Model Order Reduction. We will discuss data-driven versions of these techniques from a statistical point of view and show how they relate to common statistical approaches to dimension reduction. We will also discuss the statistical challenges that arise in fitting compact models to real data sets.
  • A Multivariate Risk-Adjusted EWMA Control Chart

    Authors: Stelios Psarakis (Athens University of Economics and Business, Dept of Statistics), Athanasios Sachlas (Athens University of Economics and Business.), Sotiris Bersimis (University of Piraeus)
    Primary area of focus / application: Process
    Keywords: Control charts, Risk-adjusted control charts, EWMA, Monitoring
    Submitted at 9-May-2017 12:49 by Stelios Psarakis
    Accepted
    Risk-adjusted control charts appeared in the bibliography, in the last two decades, to improve the monitoring mainly of medical processes. Risk-adjusted control charts take into consideration the varying health conditions of the patients. Biswas and Kalbfleisch (2008) outlined a risk-adjusted CUSUM procedure based on the Cox model for a failure time outcome while Sego et al. (2009) proposed a risk-adjusted survival time CUSUM chart for monitoring a continuous, time-to-event variable that may be right-censored. Although several authors have dealt with multivariate CUSUM charts and multivariate EWMA charts little or no work has been done to multivariate time-weighted risk-adjusted control charts. Motivated by the above mentioned papers, in this work we describe the multivariate extension of risk-adjusted EWMA control chart. The central idea is to propose methods for monitoring simultaneously more than one survival times. Since we target to more than one survival time, we use multivariate frailty models such as Gamma frailty.

    References
    P. Biswas and J.D. Kalbfleisch. A risk-adjusted cusum in continuous time based on the cox model. Statistics in Medicine, 27:3382-3406, 2008.
    L.H. Sego, M.R. Reynolds Jr, and W.H. Woodall. Risk-adjusted monitoring of survival times. Statistics in Medicine, 28:1386-1401, 2009.
  • CUSUM-Shewhart Charts for Monitoring Normal Variance

    Authors: Sven Knoth (Helmut Schmidt University, University of the Federal Armed Forces Hamburg)
    Primary area of focus / application: Process
    Secondary area of focus / application: Quality
    Keywords: Combo charts, ARL, Change detection, SPC, Variance change
    Submitted at 9-May-2017 12:49 by Sven Knoth
    Accepted (view paper)
    13-Sep-2017 09:00 CUSUM-Shewhart Charts for Monitoring Normal Variance
    Monitoring the normal variance experienced lows and highs in the SPC literature. Besides very common vehicles such as the R, S or S^2 Shewhart control charts, some more sophisticated tools such as EWMA (exponentially weighted moving average) and CUSUM (cumulative sum) charts derived from the mentioned statistics were introduced already decades ago. It is quite surprising that the analysis of combining the simple Shewhart with one of the more advanced charts gained not much interest. Except for the classic Yashchin (1985), no further studies of the subject seem to be available. One potential reason is that despite the simple operation of the combo scheme, the numerical ARL (average run length) analysis is a demanding task. Here, we want to provide some new insights following the more recent Knoth (2016). It is demonstrated that the CUSUM-Shewhart combo deploying the running sample variance S_i^2 provides a simple and powerful procedure to detect a wide range of potential changes.

    Knoth, S. (2016) "New results for two-sided CUSUM-Shewhart control charts", in Proceedings of the XII^th International Workshop on Intelligent Statistical Quality Control, pp. 269-287.

    Yashchin, E. (1985), "On the analysis and design of CUSUM-Shewhart control schemes", IBM Journal of Research and Development 29, 377-391.
  • Multistage Acceptance Sampling under Nonparametric Dependent Sampling Designs

    Authors: Andreas Sommer (Institut für Statistik und Wirtschaftsmathematik), Ansgar Steland (Institut für Statistik und Wirtschaftsmathematik)
    Primary area of focus / application: Quality
    Keywords: Acceptance sampling, Quality control, Panel model, Sampling plan
    Submitted at 9-May-2017 13:42 by Andreas Sommer
    Accepted
    13-Sep-2017 09:00 Multistage Acceptance Sampling under Nonparametric Dependent Sampling Designs
    An important problem arising in statistical quality control is to check if a lot or shipment of produced items is in agreement with the specifications. Consumer and producer both have an interest that the items meet certain quality criteria at the time of delivery and beyond. Thus, the lot should be reinspected later on. We extend the procedure of acceptance sampling by variables for an arbitrary number of inspection points under nonparametric models in the presence of additional data. In particular, we allow a time-dependent structure resulting in kind of a panel design with time-individual sample sizes. Approximations of the test statistics for independent as well as dependent sampling schemes are derived leading to asymptotically admissible sampling plans. Every sampling plan consists of the sample size and the corresponding critical value, which are itself random variables. Miscellaneous properties of the sampling plans are shown, in particular consistency and asymptotic normality. The stopping time of the procedure and the average outgoing quality are elaborated. In a simulation study, optimal choices of parameters as well as finite sample properties of the sampling plans are examined. Moreover, we propose a bootstrap procedure in order to improve the accuracy of the procedure in terms of the consumer's and producer's risk.
  • Semiparametric Statistical Analysis of the Blade Tip Timing Data for Detection of Turbine Rotor Speed Instabilities

    Authors: Marek Brabec (Institute of computer science, Czech Academy of Sciences), Pavel Prochazka (Institute of Thermomechanivs, Czech Academy of Sciences), Dusan Maturkanic (Institute of Thermomechanivs, Czech Academy of Sciences)
    Primary area of focus / application: Metrology & measurement systems analysis
    Secondary area of focus / application: Modelling
    Keywords: Semiparametric statistical modeling, GAM, Penalized splines, Time-varying regression, BTT, Turbine rotor vibrometry
    Submitted at 9-May-2017 14:37 by Marek Brabec
    Accepted
    11-Sep-2017 12:40 Semiparametric Statistical Analysis of the Blade Tip Timing Data for Detection of Turbine Rotor Speed Instabilities
    We will present a semiparametric statistical model for detecting instabilities in a turbine rotor speed. The modeling and detection uses data obtained from the now standard BTT (Blade Tip Timing) contactless measurement method. The model is based on time-varying coefficient model formulated as a GAM (Generalized Additive Model) with appropriately selected penalty. Our approach can be perceived as a fully formalized time-varying statistical extension of the traditional Fourier analysis. As such, it can reveal important rotor instabilities not readily apparent in the traditional approaches. After presenting the underlying statistical modeling framework, we will illustrate the performance of our methodology on experimental data measured on a test turbine via magneto-resistive BTT technology. The research is supported from the AV21 Strategy of the Academy of Sciences of the Czech Republic.
  • Heterogranular Multivariate Analytics for Detecting and Controlling the Root Causes of the Mismatching Machines in Semiconductor Manufacturing

    Authors: Aabir Chouichi (Ecole des Mines de Saint-Etienne ( Campus George Charpak)), Claude Yugma (Ecole des Mines de Saint-Etienne (Campus George Charpak)), Jakey Blue (Ecole des Mines de Saint-Etienne (Campus George Charpak)), Francois Pasqualini (STMicroelectronics Crolles)
    Primary area of focus / application: Mining
    Secondary area of focus / application: Process
    Keywords: Machine/Chamber matching, Multivariate analysis, Real time, Product measurements, Machine sensor readings, Run-to-run regulation, Semiconductor manufacturing
    Submitted at 9-May-2017 15:47 by Aabir CHOUICHI
    Accepted
    12-Sep-2017 18:20 Heterogranular Multivariate Analytics for Detecting and Controlling the Root Causes of the Mismatching Machines in Semiconductor Manufacturing
    In all manufacturing industries, parallel machines/chambers at a single production area are expected to have similar capability and, most importantly, to yield identical product quality. However, this is usually not the case in real practice.
    Maintaining a stable performance of parallel machines in the semiconductor industry is a real challenge given the fact that, in the massive production environment, the high product mix of very different recipes can be processed by the machines with objectives of maximizing the throughput and optimizing the machine utilization. Unsurprisingly, after going through very different recipe set-ups, machine conditions will not be identical. The implementation of Run-to-Run regulation permit to reduce the mismatching effect on quality product without solving the root causes.
    The aim of our research is to propose a methodology to detect and correct machine/chamber mismatching in real-time by exploiting all the available data, such as the product measurements, machine sensor readings and R2R regulators. These data are monitored separately to ensure the compliance with set limits. The core idea is to integrate them in order to identify the root causes of any significant differences among machines/chambers processing identical recipes.
    The current approach is first to detect existing differences between parallel machines/chambers by referring to the measurement data. Machine sensor readings are then analyzed to highlight the most influential indicators on the mismatching issue. The key is to catch the indicators, which are significantly correlated with the quality but independent of the R2R regulators, and to put on the effective control mechanism.