Restricted randomization and multiple responses in industrial experiments

This article discusses the design and analysis of an experiment performed in a continuous process (CP). Three types of iron ore pellets are tested on two levels of a process variable in an experimental blast furnace process, using a full factorial design with replicates. A multivariate approach to the analysis of the experiment in the form of principal component analysis combined with analysis of variance is proposed. The analysis method also considers the split-plot-like structure of the experiment. The article exemplifies how a factorial design combined with multivariate analysis can be used to perform product development experiments in a CP. CPs also demand special considerations when planning, performing and analyzing experiments. The article highlights and discusses such issues and considerations, for example, the dynamic characteristic of CPs, a strategy to handle disturbances during experimentation and the need for process control during experimentation.

Section: Conclusion and Discussionmentioning

confidence: 95%

Using factorial design and multivariate analysis when experimenting in a continuous process

Vanhatalo

Vännman

2008

“…reduces the negative effect (power loss) of choosing too few components for testing. Another case study, explaining 50-50 MANOVA in further details, is given in Bjerke et al 8 .…”

Section: Statistical Tools: 50-50 Manovamentioning

confidence: 99%

Analysing multivariate data from designed experiments: a case study of photo‐oxidation in sour cream

Bjerke

Larsen

Tellefsen

2007

Self Cite

Exposure to light induces photo-oxidation in dairy products, causing undesired offtaste/rancidity. In this case study we present the results of a designed experiment in four factors: illumination source, type of packaging, distance to illumination source and exposure time, in order to observe the resulting oxidation of low-fat sour cream, measured by fluorescence spectroscopy. Such spectroscopic data yield highly correlated response variables. A novel multivariate method, 50-50 MANOVA, is applied for the statistical analysis. The main objective is to present and interpret complex and comprehensive data results to researchers and practitioners in an informative way. The experimentation revealed that several factors and interactions must be taken into consideration when designing optimal environment for sour cream storage. Copyright

“…However, factors that involve the press setup will be difficult and time-consuming to change, whereas factors that can be manipulated within a press run will be easier to change. This gives the experiment a split-plot aspect [22][23][24][27][28][29][30][31][32][33][34][35][36][37] .…”

Section: Randomization Schemementioning

confidence: 99%

Interactive data mining informs designed experiments

Gaudard¹,

Ramsey²,

Stephens³

2008

We illustrate how a Six Sigma project team can apply recursive partitioning to a historical data set to narrow down a list of potential experimental factors and then construct an experimental design using information from the partition analysis. The paper illustrates the value of analyzing historical manufacturing data to inform the choice of factors and levels for statistically designed experiments.