Cloud Based Optimisation Of An Extrusion Die
Many products surrounding us in our everyday life are created by polymer extrusion: foils, sheets, pipes, window profiles, etc. The market volume for polymer extrusion processing industry is € 900 Mio. (foils and sheet will be approx. € 250 Mio.).
The aim of the experiment is to optimize the production process with optimized extrusion dies. A particular extrusion die – a flat slit die – is being optimized in this application experiment. Flat slit dies are particularly complex due to structural expansion. The simulation of the interaction of melted plastics with the geometry of the flat slit extrusion die is computationally expensive and technically demanding as fluid-structure interaction appears and need to be coupled in the simulation – actually two simulators need to be coupled to solve for these physical phenomena.
The expected technical impact is savings in energy and raw material consumption, reductions in production down time, improvements on the final product quality and avoidance of recall actions for imperfect products. Economically, it is expected that EMO, the manufacturing SME, doubles their market share for flat slit dies from 5 percent to 10 percent in a € 100 Mio. market, INO enlarges their customer base by 50 SMEs and DHCAE extends their HPC portfolio with Infiniband-based cluster and online post-processing services.
After the design and manufacturing of an extrusion die, a tool must be put into operation, which is a complex process in itself. During the commissioning, iterative tests with the tool are conducted to detect necessary changes of the flow channel design. This very time consuming and cost-intensive process has not changed for decades. The manufacturing and commissioning of a flat slit die can take more than three months. Depending on the die size, costs of more than € 400,000 can occur, e.g. for quality steel, hours of work, commissioning material, etc.
The main challenge in plastic mold extrusion is to generate a uniform flow field across the exit plane of an extrusion die. To simulate this process fluid-structure interaction needs to be put in place accounting also for thermal effects. Coupled simulations are especially demanding as for each time step the result of the fluid solver has to be exchanged with the result of the structure solver, so that the bi-directional influences (forces) can be taken into account for the next solver step. This application experiment approaches this challenge by coupling two cost-efficient, scalable Open Source solvers with a highly customized simulation model for non-Newtonian polymer flow. This will be integrated into an extremely user-friendly workflow within the Cloud- Flow environment. Due to the HPC cluster an outstanding improvement of the process is expected both in terms of runtime and in process quality.
DHCAE can provide low-latency HPC-based workflows to customers (significantly reduce computational times for large problems: before CloudFlow – max 72 cores, with CloudFlow – several 100 cores). Large computational problems can be calculated in 10 percent of the time that typical CFD workstations need. Customers will be able to use FSI coupling for steady state simulations without the necessity to struggle with installing all the required libraries, setting up the HPC back-end, etc. The experiment provides a template so that the other solvers from DHCAE can be offered as a Cloud service in the future as well.
INO can provide new simulation and optimization services based on Fluid-Structure-Interaction simulations and significantly increase computational power available due to the Arctur HPC Cluster to solve customer problems more rapidly or approach problem complexity which was not possible to be solved efficiently before – without HPC resources.
EMO will use the optimized extrusion die flow channel designs to produce better and cheaper products. The advanced CFD and CSM technology is planned to be introduced for in-house usage without the need of buying hardware resources and the corresponding administrative effort to maintain them.
The experiment can show that the simulation times for large problems can be speed up by a factor of 10 compared to in-house hardware and by a factor of 3 compared to existing Amazon-based HPC architecture. The CloudFlow platform gives new sales opportunities to DHCAE, in particular selling own CFD solvers with a payper- use approach, and with no installation trouble on the customers side (new services, rare users with temporary demand). The existing software, in particular the pre-processing system CastNet, will get more attractiveness and this will result in 10+ additional soft licenses to be sold. DHCAE expects three new SME customers within 1 year and 10 new SME customers within three years. The revenue is expected to be increase by 20 percent, which is € 70,000, and one additional engineer will be employed.
Within the next three years INO is expecting that two new full-time equivalents (FTEs) can be hired to join the company. With the increased possibilities of workflows and HPC power, INO will address 500 new customers in the next three years and expects to triple their sales of € 80,000.
EMO can avoid recall actions for each manufactured extrusion die. One recall actions implies costs of approx. € 100,000 for transportation (17 tons weight), modification and commissioning. The recall actions represent up to 30 percent of the production costs. Due to a long-time average of 12 recalls per year this will save € 1.2 Mio. per year. Commissioning time will be reduced down to 70 percent and equals a time-to-market reduction of several weeks. General production costs can be reduced by 30 percent, particularly due to savings of energy, raw materials and personal costs. Right now EMO has too less experience with this tool, but we expect to have a cost reduction of about 20-25 percent.
Arctur being an infrastructure provider expects that the economic impact form this experiment will be reflected through the increased sale of Cloud and HPC resources. The increase of the sales follows the increase of new users using the services at offer. The secondary beneficial impact for Arctur is the widening of the service portfolio and references. The service will be on offer in collaboration with the specialist knowledge of the field specialist of this experiment.