Improving paper machine clothing supplier's industrial internet offering with artificial intelligence
Janhunen, Lasse (2020)
2020
Konetekniikan DI-tutkinto-ohjelma
Tekniikan ja luonnontieteiden tiedekunta - Faculty of Engineering and Natural Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2020-01-07
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-201912106723
https://urn.fi/URN:NBN:fi:tuni-201912106723
Tiivistelmä
Overall amount of data has grown exponentially during the last few years. The increase in the availability of data has driven companies and countries towards digitalization with growing pace. Therefore, the industrial internet applications have become more successful than ever. These applications provide companies more tools to utilize data-driven decisions. In paper industry, paper machine original equipment manufacturers have started to utilize the industrial internet capabilities with increasing pace. The increasing competition has led to the fact that today, utilization of the possibilities offered by industrial internet is part of target organization’s (Valmet) main strategies. Thus, the paper machine clothing (PMC) unit of Valmet has commissioned this thesis work.
The goal of this research was to improve Valmet’s PMC unit’s industrial internet offering. Improvement actions taken were to enhance the existing offering through customer feedback and to provide additional value with artificial intelligence. The approach towards the subject was to find out the existing theory behind the operational context of the fabrics, discover possible developmental actions through prototyping and by creating value-adding AI models to support the offering.
During this research process it came evidently clear that the initial industrial internet applications would have good applicability in pilot customer’s daily routines. Though good developmental points were discovered from the prototyping phase, the functionality issues of the initial industrial internet applications during the timeframe of this thesis limited the quality of the feedback. More thorough study for customer feedback should be conducted after the applications have been in daily use for solid amount of time.
This research provided two value-adding models for industrial internet applications. The idea for the models sprung from the hopes of the target company. Initially, fabric delivery cycles have been defined more or less by hand. Thus, the Monte Carlo simulation to optimize delivery cycles and to manage risk governing possible shortages was illustrated as the first model. The second model aimed to enhance the first model by conducting estimations of remaining fabric lifetime from customer’s mill’s process data. Neural network was chosen as the machine learning method for this model. Both models were tested with actual process data and the results of the case study were polarized. The simulation model provided valid results and first indications showed that it would bring true added value to the target organization. However, the results of the second model indicated that with available data valid results were not acquired. The results of this study indicate that the artificial intelligence models can be utilized to fabrics industrial internet but more emphasis should be pointed on the comparison of different machine learning methods and to enhance the quality and quantity of the available data.
The goal of this research was to improve Valmet’s PMC unit’s industrial internet offering. Improvement actions taken were to enhance the existing offering through customer feedback and to provide additional value with artificial intelligence. The approach towards the subject was to find out the existing theory behind the operational context of the fabrics, discover possible developmental actions through prototyping and by creating value-adding AI models to support the offering.
During this research process it came evidently clear that the initial industrial internet applications would have good applicability in pilot customer’s daily routines. Though good developmental points were discovered from the prototyping phase, the functionality issues of the initial industrial internet applications during the timeframe of this thesis limited the quality of the feedback. More thorough study for customer feedback should be conducted after the applications have been in daily use for solid amount of time.
This research provided two value-adding models for industrial internet applications. The idea for the models sprung from the hopes of the target company. Initially, fabric delivery cycles have been defined more or less by hand. Thus, the Monte Carlo simulation to optimize delivery cycles and to manage risk governing possible shortages was illustrated as the first model. The second model aimed to enhance the first model by conducting estimations of remaining fabric lifetime from customer’s mill’s process data. Neural network was chosen as the machine learning method for this model. Both models were tested with actual process data and the results of the case study were polarized. The simulation model provided valid results and first indications showed that it would bring true added value to the target organization. However, the results of the second model indicated that with available data valid results were not acquired. The results of this study indicate that the artificial intelligence models can be utilized to fabrics industrial internet but more emphasis should be pointed on the comparison of different machine learning methods and to enhance the quality and quantity of the available data.