Research on a multivariate measurement system for polyethylene gas pipelines utilizing a variable weight UMAP model

Area of Science:

• Industrial Engineering
• Quality Control
• Data Science

Background:

• Evaluating measurement system performance for gas-polyethylene pipelines is complex due to interacting quality indicators.
• Traditional methods struggle with the high-dimensional and non-linear nature of quality data.

Purpose of the Study:

• To develop an advanced multivariate measurement system analysis model.
• To enhance the evaluation of measurement system performance in gas-polyethylene pipeline production.

Main Methods:

• Utilized Variable-Weight Uniform Manifold Approximation and Projection (VUMAP) for dimensionality reduction.
• Implemented a dynamic weighting method using K-Means++ clustering and Particle Swarm Optimization (PSO).
• Refined VUMAP to better preserve non-linear relationships in high-dimensional data.

Main Results:

• Developed an integrated multivariate model for comprehensive measurement system assessment.
• The dynamic weighting method improves accuracy over traditional uniform weighting.
• The refined VUMAP enhances data structure exploration and reliability.

Conclusions:

• The proposed model effectively overcomes limitations of traditional methods in measurement system analysis.
• The model provides a more accurate and reliable evaluation of quality indicators in pipeline production.
• Validated effectiveness using a gas-polyethylene pipeline production dataset.