ENBIS-12 in Ljubljana

9 – 13 September 2012 Abstract submission: 15 January – 10 May 2012

My abstracts

 

The following abstracts have been accepted for this event:

  • Extracting a Product’s Mission Profile from Historical Data Sources

    Authors: Tom Delesie (Philips Innovative Applications), Joost Van Der Eyden (Philips Innovative Applications), Jan Pieter Dekker (Philips Innovative Applications), Piet Watté (Philips Innovative Applications)
    Primary area of focus / application: Reliability
    Keywords: Reliability Centered Maintenance, Markov Chain, Mission profile, Field Return
    Submitted at 13-Apr-2012 16:08 by Tom Delesie
    Accepted
    Lamps used in entertainment applications have a high demand on reliability. It is evident that entertainment lamps need to operate at all times during performances. For a good reliability centered maintenance schedule, it is therefore necessary to have an estimate of the mission profile, or average burning hours per unit of time.

    To correctly estimate the maintenance routine new products, the design engineer requires a sufficiently accurate mission profile of the product in the field. From historical data of similar products, it is possible to come to good estimate for the mission profile, allowing for a probability density function to be constructed.

    In this paper, a case is presented in which the available historical data provides an excellent data source to estimate the mission profile of the product. Field return data is used to construct a probability density function and through Markov Chain simulation, an estimate for the mean and standard deviation of the mission profile is calculated.
  • Reasons for Not Using Factorial Experimental Designs

    Authors: Bjarne Bergquist (Luleå University of Technology)
    Primary area of focus / application: Education & Thinking
    Keywords: Design of Experiments, implementation, industrial Use, process industry, statistical thinking
    Submitted at 13-Apr-2012 17:09 by Bjarne Bergquist
    Accepted
    10-Sep-2012 15:15 Reasons for Not Using Factorial Experimental Designs
    Factorial designs and other types of designs found in text books on how to perform experiments are considered mainstay of industrial experimentation, at least it seems so in the scholarly discussions. However, many studies have shown that the use of such methods is very limited, and it does not seem to increase. In this study, process engineers, product developers, industrial research specialists working in the mining industry as well as laboratory researchers have been interviewed to find reasons for the low use. The findings point to that altough a more systematic approach may be useful in some instances, many hindrances are hard to overcome.
  • A Time Series Analysis Approach to Analyze Two-Level Factorial Designs Affected by Disturbances

    Authors: Peder Lundkvist (Luleå University), Erik Vanhatalo (Luleå University)
    Primary area of focus / application: Design and analysis of experiments
    Keywords: experimental design and analysis, factorial experiments, blast furnace experiments, times series analysis, transfer function-noise modeling
    Submitted at 13-Apr-2012 17:27 by Erik Vanhatalo
    Accepted
    10-Sep-2012 11:50 A Time Series Analysis Approach to Analyze Two-Level Factorial Designs Affected by Disturbances
    Industrial experiments are sometimes subjected to disturbances which may force the experimenter to deviate from the experimental plan. Although the loss of one or a few of the experimental runs in a larger design is not a critical problem, the same loss in a design with few runs can make the analysis difficult. This problem may be especially important for many process industries where the experimental cost makes larger designs problematic and disturbances during the often long experimental campaigns are common.

    The purpose of the presentation is to propose and illustrate a method to analyze a small two-level factorial design performed in a continuous process where operational problems affected several of the experimental runs and the resulting response time series.

    The presentation outlines a time series analysis approach to analyze a two-level factorial design performed in a blast furnace where operational problems affected several of the experimental runs. In particular, the presentation illustrates how transfer function-noise modeling can be used to analyze a two-level factorial experiment after first filtering out the disturbances from the original time series response. The results are compared with those from a more ‘traditional’ analysis using averages of the response in each run as the single response in an analysis of variance (ANOVA).
  • Data Analysis for Condition Based Railway Maintenance

    Authors: Bjarne Bergquist (Luleå University of Technology), Peter Söderholm (The Swedish Transport Administration)
    Primary area of focus / application: Reliability
    Keywords: Condition Based Maintenance, Time Series Analysis, Metrology, Model Building
    Submitted at 13-Apr-2012 17:28 by Bjarne Bergquist
    Accepted
    10-Sep-2012 17:10 Data Analysis for Condition Based Railway Maintenance
    Condition monitoring for railway systems are important to optimize maintenance actions. A current method is to plan actions when a measurement reached a critical level, but it may be difficult to perform the maintenance action due to traffic concerns. Good prediction models would thereforemean less transportation delays and more consistent workflow for the maintenance entrepreneurs. A requirement for good maintenance prediction models are that the analysis methods are appropriate and that the data is of good quality. The data is multivariate and highly dependent, which puts restrictions on what statistical methods that may be useful for predictions. Here, we demonstrate how time series analysis may be used for prediction of track deterioration.
  • On the Use of Partial versus Marginal Correlations in SPC

    Authors: Tiago M. Rato (University of Coimbra), Marco S. Reis (University of Coimbra)
    Primary area of focus / application: Process
    Keywords: Statistical Process Control, Multivariate Systems, Partial Correlations, Marginal Correlations
    Submitted at 13-Apr-2012 21:40 by Marco P. Seabra dos Reis
    Accepted (view paper)
    10-Sep-2012 16:50 On the Use of Partial versus Marginal Correlations in SPC
    Current multivariate statistical process control methods are built over frameworks based on marginal correlation, such as the Hotelling-T2 and PCA control (Hotelling 1931; Jackson 1959). Therefore, by design, these methods are unable to discern local changes in the process structure. On the other hand, partial correlation coefficients may retain some of this information, which may then be employed for deriving more sensitive SPC schemes, able to detect fine perturbations in the local variables’ structure. Furthermore, such schemes will also be superior in the task of diagnosing the root causes after a special event is detected. In this work, we have studied several statistics for monitoring changes in the partial correlation coefficients, and therefore detect changes in the process structure. These statistics were applied on linear multivariate systems and their performances were compared with their marginal counterparts. The results obtained showed that the partial correlation based statistics were indeed able to detect changes in the systems structure and presented higher sensitivity regarding the traditional monitoring statistics tested.
    References
    Hotelling, H. (1931). "The Generalization of Student's Ratio." Annals of Mathematical Statistics 2(3): 360-378.
    Jackson, J. E. (1959). "Quality Control Methods for Several Related Variables." Technometrics 1(4): 359-377.
  • Rolling the Improvement Wheel in a Pharmaceutical Filling Plant

    Authors: Antje Christensen (Novo Nordisk)
    Primary area of focus / application: Six Sigma
    Keywords: Lean Six Sigma, Process Improvement, Problem Solving, Shop Floor Management
    Submitted at 14-Apr-2012 02:09 by Antje Christensen
    Accepted (view paper)
    10-Sep-2012 10:20 Rolling the Improvement Wheel in a Pharmaceutical Filling Plant
    The biggest risk to a Lean Six Sigma project is not the risk of not finding the right solution, but the risk of not anchoring it well in the line of business. At Novo Nordisk, we address this issue by integrating the Lean Six Sigma portfolio with initiatives to improve shop floor leadership, shop floor problem solving, improved documentation format, and improved training setup. The company’s Improvement Wheel will be presented and exemplified in a case from a pharmaceutical filling facility.