ENBIS-14 in Linz

21 – 25 September 2014; Johannes Kepler University, Linz, Austria Abstract submission: 23 January – 22 June 2014

My abstracts

 

The following abstracts have been accepted for this event:

  • The Challenges of PCA-Based Statistical Process Monitoring: An Overview and Solutions

    Authors: Eric Schmitt (KU Leuven), Bart De Ketelaere (KU Leuven), Tiago Rato (University of Coimbra), Marco Reis (University of Coimbra)
    Primary area of focus / application: Process
    Keywords: Control charts, Time-Dependent data, High-Dimensional data, Principal component analysis, Methodology implementation
    Submitted at 31-May-2014 12:24 by Eric Schmitt
    Accepted
    23-Sep-2014 15:00 The Challenges of PCA-Based Statistical Process Monitoring: An Overview and Solutions
    Control charts are tools developed in statistical process monitoring (SPM) to identify when a process is out-of-control. When constructing control charts, one or more tuning parameters should be chosen to obtain the desired properties. For classical control charts, approaches for this are well documented, but for multivariate control charts this still remains an open field of research. Control charts for high-dimensional data, including those based on principal component analysis (PCA) are no exception, with more complex methods designed for time-dependent data being in particular need of further elaboration. We first provide an example based overview of classical PCA-based control charts and identify the current approaches to choose their tuning parameters. Based on that discussion, we will then show how violation of the basic assumptions w.r.t. the data (e.g. with respect to autocorrelation and/or stationarity) also violates the underlying assumptions of commonly used parameter selection approaches, resulting in poor monitoring performance. An example of this is the problem of specifying the control limits based on analytic expressions that can even fail in stationary scenarios. In the case of time-dependent data, those issues are even more accentuated, and identifying a model, even for data exhibiting normal operating conditions, can be a challenge. As a result, the user is often compelled to select the tuning parameters based on inefficient trial and error approaches. In our previous research into extensions of classical PCA, which was presented to ENBIS in 2013, we compared the performance of those methods while using manually selected parameters. In this follow-up we will highlight existing, but less well-known solutions to some of these parameter selection challenges when setting up the control charts, and propose solutions for others. By addressing such issues we intend to provide a more straightforward application of PCA-based control charts in research and in practice so that these methods may be more accessible to users.
  • The Use of Statistical Improvement Methods in Swedish Organizations

    Authors: Peder Lundkvist (Luleå University of Technology), Bjarne Bergquist (Luleå University of Technology), Erik Vanhatalo (Luleå University of Technology)
    Primary area of focus / application: Education & Thinking
    Secondary area of focus / application: Economics
    Keywords: Statistical methods, Statistical process control, Capability analysis, Design of Experiments, Survey, Swedish organizations
    Submitted at 31-May-2014 12:31 by Erik Vanhatalo
    Accepted
    22-Sep-2014 12:15 The Use of Statistical Improvement Methods in Swedish Organizations
    Statistical improvement methods have been around for many decades but prior studies in Sweden as well as international studies show that their use has been moderate at best. The purpose of this study is to investigate if and how the use of the three improvement tools statistical process control (SPC), capability analysis, and design of experiments (DoE) has changed over the last years in Sweden but also to investigate which aspects that are regarded as hindrances for the use of these methods.

    Our survey targeted alumni from the Master’s program in Industrial and Management Engineering at Luleå University of Technology (LTU) who are familiar with the methods from undergraduate courses. The respondents were asked to assess the use of the methods in their last two workplaces. In total 147 responses were obtained, giving a response rate of 36 %. These respondents assessed the use of the methods in 211 organizations in total.

    The results indicate that the use of statistical methods in Swedish organizations has increased somewhat compared to earlier studies, and in particular, the use has become more systematic. Organizations larger than 250 employees have a relatively higher use of the methods compared to smaller organizations and their use is generally higher in industry compared to the service sector. The results clearly show that SPC is the most commonly used of the three methods and that DoE is the least used method. Furthermore, the most important aspects that are reported to hinder the use of the methods are insufficient resources in terms of time and low commitment of top and middle managers.
  • Optimal Design of Discrete Choice Experiments with Partial Profiles

    Authors: Daniel Palhazi Cuervo (University of Antwerp), Peter Goos (University of Antwerp, University of Leuven), Roselinde Kessels (University of Antwerp), Kenneth Sörensen (University of Antwerp)
    Primary area of focus / application: Design and analysis of experiments
    Keywords: Discrete choice experiments, Partial profiles, D-optimality criterion, Utility-neutral designs, Locally optimal designs, Bayesian optimal designs
    Submitted at 31-May-2014 18:03 by Daniel Palhazi Cuervo
    Accepted
    22-Sep-2014 11:05 Optimal Design of Discrete Choice Experiments with Partial Profiles
    Discrete choice experiments are conducted to identify the attributes that drive people’s preferences when choosing between competing options of products or services. A large number of attributes might increase the complexity of the choice task, and might have a detrimental effect on the quality of the results obtained from the experiment. In order to reduce the cognitive effort required by the experiment, some authors recommend designs with partial profiles (where the levels of some attributes are held constant in each choice situation). There exist D-optimal designs for experimental scenarios where respondents are assumed to have no preference for any of the alternatives. In more realistic scenarios, however, experimenters have limited prior information about people’s preference for certain alternatives. For this cases, no optimal designs are known and experimenters must resort to algorithmic generation methods. We present an integrated algorithm for the generation of locally optimal designs and Bayesian optimal designs with partial profiles. This algorithm optimizes the set of constant attributes and the levels of the varying attributes simultaneously. The designs generated by the integrated algorithm outperform those produced by existing algorithms. Moreover, when respondents are assumed to be indifferent, the algorithm generates designs that match the quality of the known D-optimal designs.
  • Process Improvement via Design of Experiment: Applicative Examples in Technological Field

    Authors: Flaviana Tagliaferri (University of Naples Federico II), Biagio Palumbo (University of Naples Federico II), Martin Dix (Chemnitz University of Technology), Michael Kuhl (Fraunhofer - IWU)
    Primary area of focus / application: Design and analysis of experiments
    Secondary area of focus / application: Quality
    Keywords: Design of Experiment (DoE), Laser Assisted Machining (LAM), Water abrasive jet, Ceramic machining
    Submitted at 31-May-2014 20:19 by Biagio Palumbo
    Accepted
    22-Sep-2014 12:15 Process Improvement via Design of Experiment: Applicative Examples in Technological Field
    The main objective of this work is to highlight the strategic role that a systematic and sequential approach to experimentation plays in order to get competitive advantage and technological innovation for both manufacturing industry and research centres.
    The efficacy of this approach is demonstrated here by three applicative examples in technological field, developed at Chemnitz University of Technology where the integration between statistical and technological knowledge has allowed: to characterize and to optimize some manufacturing processes; to catalyze the innovation process; to promote the technological transfer. Moreover this approach allows to put into action a virtuous cycle of sequential learning.
    An accurate pre-design (i.e. pre-experimental planning phase) is the solid basis on which a statistical approach has to be built. The use of the pre-design guide sheets provides a way to systematize the process by which an experimentation team does the experimental plan. In fact, these sheets drive the experimenter to clearly define the objectives and scope of an experiment and to gather information needed to design an experiment.
  • Learning from Mistakes: Using Problems with a Small, Real Experiment to Teach Principles of DoE

    Authors: Jacqueline Asscher (Kinneret College on the Sea of Galilee)
    Primary area of focus / application: Education & Thinking
    Secondary area of focus / application: Design and analysis of experiments
    Keywords: Design of Experiments (DoE), Training, Teaching, Project based, Problems
    Submitted at 31-May-2014 20:25 by Jacqueline Asscher
    Accepted
    22-Sep-2014 12:35 Learning from Mistakes: Using Problems with a Small, Real Experiment to Teach Principles of DoE
    In the first week of a course in statistical design of experiments (DoE), participants run a small real experiment to check whether placing candles in the freezer and/or shielding them from drafts extends the length of time that they burn.
    The design of the experiment is identical for all participants, a simple two-way ANOVA with two levels for each factor and four replicates.
    The assignment is deliberately vague, leaving the course participants to decide how to choose materials, measure the burn time and organize the 16 runs. Typically problems arise: for example: "we'll put the red candles in the freezer and leave the yellow ones at room temperature"; "who knew these candles would burn so long – I want to go to sleep"; "I ran the experiment at dusk, turned the light on in the middle and forgot that the ceiling fan turns on with the light"; "the candles surrounded by other candles burned faster".
    This small experiment is useful for teaching the importance of a pilot experiment, as a point of reference for teaching the principles of DoE such as blocking, randomization and interaction, as a review of the material from a previous course in basic statistics and as preparation for a larger real experiment run by participants later in the course.
  • Hidden Markov Model for the Detection of a Degraded Operating Mode of Optronic Equipment

    Authors: Anne Gégout-Petit (Université de Lorraine), Camille Baysse (Thales Optronique), Saracco Jérôme (INRIA Bordeaux)
    Primary area of focus / application: Reliability
    Keywords: HMM, Degradation, Detection, Reliability
    Submitted at 31-May-2014 20:53 by Anne Gégout-Petit
    Accepted
    22-Sep-2014 16:40 Hidden Markov Model for the Detection of a Degraded Operating Mode of Optronic Equipment
    We use hidden Markov Model to detect as soon as possible a change of state of optronic equipment in order to propose maintenance before failure. For this, we use the dynamic of a variable called "cool down time" (cdt) that reflects the state of the cooling system. Indeed, it is an observation of the hidden state of the system.
    This one is modelled by a Markov chain and the cdt is a noisy function of it. Thanks to filtering equations, we obtain results on the probability that an appliance is in degraded state at time t, knowing the history of the Tmf until this moment.
    We have evaluated the numerical behavior of our approach on simulated data.
    This method can be implemented in a HUMS (Health and Usage Monitoring System) and used to propose adaptive maintenance conditional to the estimated state.