Predicting Tensile Properties of Technical textiles with Artificial Neural Networks in the Product Development Process A case study on the requirements for an ANN implementation, the process improvements, and resulting competitive advantages
Ladda ner
Publicerad
Författare
Typ
Examensarbete för masterexamen
Master's Thesis
Master's Thesis
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
The end-user environments of technical textiles are often characterized by extreme
weather conditions, leading to aging which implies the need for testing in the product
development process (PDP). This thesis aims to suggest an Artificial Neural
networks (ANN) model predicting the aging of technical textiles and an AI strategy
for companies in the technical textile industry. The resulting improvements
and competitive advantages of such an implementation will also be studied. A
case study of an aging measurement analysis in the testing and validation phase of
the PDP in the technical textile industry was conducted. Through semi-structured
interviews, focus group and workshop, the PDP, aging measurement analysis, requirements
for an ANN implementation, and the resulting improvements and competitive
advantages were mapped. This was analyzed using theory on ANNs, the
stage-gate process, requirements for AI implementation, improvements from AI, and
competitive advantages. It was found that the aging measurement depends on an
in-depth understanding of the complex relationships between material components.
As a result, several iterations in the PDP to ensure that the prototype meets the
aging requirements are often required and lead to a time-consuming and costly process.
Another finding was that the industry is characterized by low digital maturity.
It was concluded that two single hidden layer backpropagation feedforward ANN
models were suitable for predicting the tensile properties. It is recommended to
initially use eight categories of input parameters and then employ regularization.
Sigmoid and ReLU are suggested as activation functions and Levenberg-Marquardt
as optimization functions. A combination of error measures are suggested. This
thesis concluded that an AI strategy for the technical textile industry to implement
ANN successfully should include plans to fulfill requirements regarding, internal and
external data, data anonymization, automated data collection, computing power infrastructure,
knowledge, and employee AI trust. Further, it was concluded that the
implementation of ANN could improve the knowledge in the PDP, leading to improved
speed, quality, and cost. In the technical textile industry, the corresponding
competitive advantages are, time to market, superior products, innovation capacity
and profitability.
Beskrivning
Ämne/nyckelord
Technical textile industry, Tensile properties, Sun-simulation, Product development, Process improvements, Competitive advantages, Artificial Neural Networks