Work

Statistical and machine learning approach for evaluation of control systems for automatic production lines

Year: 2022

Type: article

Abstract: The manufacturing processes and the control systems for automatic production lines mainly are evaluated through usage of statistical methods as recently machine learning algorithms are also used. The aim of the paper is to present an approach for control measurement systems evaluation, based on a combination of statistical techniques like attribute repeatability and reproducibility analysis, measurement system analysis and supervised machine learning algorithms like random forest and KNN. The proposed method is verified in the production of the G8680x connector, which is used in the automotive industry. The control is performed 100% for all manufactured parts immediately after the “injection molding” process. It is proved that taking advantages of the statistics and machine learning, the manufacturing process and control measurement systems could be evaluated with very high accuracy. The exploration and analysis leads to the formulation of some recommendations in support of process engineers and managers.

Source: Bulletin of Electrical Engineering and Informatics

DOI: 10.11591/eei.v11i5.3664

Authors: Valentin Tsenev, Malinka Ivanova

Institution: Technical University of Sofia

Language: English

Topic: Advanced Statistical Process Monitoring

Subfield: Statistics, Probability and Uncertainty

Field: Decision Sciences

Domain: Social Sciences

SDG: Industry, innovation and infrastructure

Open Access status: diamond

Funder: Bulgarian National Science Fund