A data mining based approach for process identification using historical data

Oulhiq, Ridouane; Benjelloun, Khalid; Kali, Yassine; Saad, Maarouf

doi:10.1080/02286203.2021.1905375

Cited by 5 publications

(1 citation statement)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The authors of [38] wrote that using observational data as a replacement for experimental data is risky because of the absence of controllable factors, spurious correlation, and the rise of potential multicollinearity or nonorthogonality. This finding was similar to the problem of semiconductor production in the work of [3] and the slurry thickening process in [39], where observational data contained undetected and uninterpretable multicollinearity, given the application of typical observational-based regression analysis and the need for careful handling ( [40]). This opinion was also strengthened by the authors of [19], who wrote a systematic literature review of classic RSM development and showed that orthogonality between factors should be reached to perform individual analyses of each factor.…”

Section: Synthesis and Discussionsupporting

confidence: 55%

Response Surface Methodology Using Observational Data: A Systematic Literature Review

2022

View full text Add to dashboard Cite

In the response surface methodology (RSM), the designed experiment helps create interfactor orthogonality and interpretable response models for the purpose of process and design optimization. However, along with the development of data-recording technology, observational data have emerged as an alternative to experimental data, and they contain potential information on design/process parameters (as factors) and product characteristics that are useful for RSM analysis. Recent studies in various fields have proposed modifications to the standard RSM procedures to adopt observational data and attain considerable results despite some limitations. This paper aims to explore various methods to incorporate observational data in the RSM through a systematic literature review. More than 400 papers were retrieved from the Scopus database, and 83 were selected and carefully reviewed. To adopt observational data, modifications to the procedures of RSM analysis include the design of the experiment (DoE), response modeling, and design/process optimization. The proposed approaches were then mapped to capture the sequence of the modified RSM analysis. The findings highlight the novelty of observational-data-based RSM (RSM-OD) for generating reproducible results involving the discussion of the treatments for observational data as an alternative to the DoE, the refinement of the RSM model to fit the data, and the adaptation of the optimization technique. Future potential research, such as the improvement of factor orthogonality and RSM model modifications, is also discussed.

show abstract