Wirtschaftliche und technische Modellierung und Selbstoptimierung von Vliesstoffkrempeln

  • Economic and technical modelling and self-optimization of nonwoven cards

Cloppenburg, Frederik; Gries, Thomas (Thesis advisor); Schlichter, Stefan (Thesis advisor)

Düren : Shaker (2019)
Book, Dissertation / PhD Thesis

In: Textiltechnik/Textile Technology
Page(s)/Article-Nr.: VI,380 Seiten : Illustrationen, Diagramme

Dissertation, RWTH Aachen University, 2019

Abstract

The carding process is the most relevant process for the production of nonwovens from fibers of limited length (staple fibers). The quality parameters of nonwovens are complexly influenced by a large number of properties of the raw material, the production settings and the environmental conditions. In particular, the setting of the card has an enormous influence on the quality of the nonwoven. The aim of this work is to optimize the cards under technical and economic aspects using machine learning methods. For this purpose, the influences on the resulting nonwoven quality and the production costs are determined and a measuring system is developed which records all variables during the production process. The effect of the influencing variables on the target variables of product quality and production costs is modelled with the help of machine learning methods. Based on the developed models a simulation is developed, which simulates the expected product quality and the expected energy consumption for different set-tings. For each setting point the specific production costs are calculated which are expected at the given settings. An optimization algorithm based on the ε-constraint method is developed, which systematically browses through the simulation results and selects the setting where the required quality is achieved and the production costs are the lowest. The developments will be validated on two production lines. The investment is evaluated monetarily using a dynamic payback period calculation. Depending on the degree of previously realized production optimization, the payback period lies between 3 and 14 months. The measurement system is then further developed to classify the subjectively perceived optical uniformity of a carded web using cost-effective camera technology and neural networks. It is shown that the detection of the aesthetical nonwoven quality is successful with low-cost camera technology and an artificial neural network.

Identifier