Picture Exp_14 Picture Exp_15 Picture Exp_16 Picture Exp_17 Picture Exp_18 Picture Exp_19 Picture Exp_20

Optimizing moulds and stamping processes for bipolar plates used in fuel cells

Executive Summary

Optimizing tools for stamping processes, finding the right process parameters and ramp-up forming production of sheet metals is a big topic in manufacturing industry, especially for those segments heavily handling sheet metals. This experiment is specifically looking into bipolar plates for fuel cells, an emerging market contributing to renewable greener energy supply, e.g. for Fuel Cell Electric Vehicles (FCEV). However, to gain market share, fuel cell production costs need to decrease.

The goal of this Application Experiment is to reduce development cost and time-to-market for new bipolar plates for fuel cells by accelerating engineering and manufacturing processes through cloud-based optimization technology. Technically, it is expected to lower barriers for SMEs to access engineering design workflows and computational resources. Economically, Borit – the end user in this experiment - expects 5 to 10 percent cost reduction. This cost reduction will mainly be achieved in the design and test phase for new products and can be translated in a reduction of total design time for the forming tool and the number of iterations required on the press (= less test tooling and press time) before a product which is up-to its specification is achieved. On average this would reduce the design time by several weeks and the testing time ‘on press’ by several days. This equals to approx. € 30,000 to 40,000 per new product (designer, less tools, les press depreciation and operator costs). If five new products are introduced each year, this could save up to € 200,000 per year and faster time-to-market; in a best-case scenario up to 3 months. This might also increase the employment rate by 1 or 2 operators in 2 to 3 years. The ISV Noesis will gain access to new markets and market segments by offering cloud-based services on an affordable pay-per-use model.

The process at Borit currently implemented does not involve simulation software for the forming process. Borit starts with creating the geometry of the bipolar plates in CAD. The corresponding tool geometry is also created using CAD software. The tool is then manufactured based on the CAD design. The mould is tested ‘on press’ and the design is iteratively altered until the plates have the intended nominal shape. This iterative process can take up to 6 to 8 weeks for just one plate depending on the plate complexity.

The manufacturing challenges are related to the quality of the resulting bipolar plates, e.g. not fully formed plate features and/or local rupture of the plate material. The simulation challenges are related to the accuracy of the model. A high fidelity simulation tool that can cut down the number of iterations needs to be found that responds in reasonable time using affordable compute resources. In addition, a simulation model needs to be defined that captures the plate behaviour with high enough resolution representing features like small radii and narrow channels.

The approach in this experiment is to introduce a combination of forming simulation software and the optimization tool Optimus by Noesis using HPC resources. Optimus defines the simulation strategy and launches the HPC service. Optimus acquires the results and builds a surrogate model from these simulation results using machine learning. The surrogate model is returned to the user to be implemented as process control.


The experiment result is expected to close the loop between the Design Engineering Process and Manufacturing Life Cycle by the development of surrogate models based on high-fidelity simulation with approximately 30 percent of time reduction for the trial-and-error process (2 to 3 weeks faster than the current 6 to 8 weeks) and significant improvements to the manufacturing quality. Moreover, it is expected to reduce the scrap rate from 5 to 2 percent; potentially € 10,000 per year of avoided scrap. With the introduction of simulation and optimization tools more complex and challenging designs can be handled thus motivating Borit to come up with innovative shapes for their bipolar plates that further improve product performance.

Technical impact for Noesis as ISV comprises lowering access barrier to computational resource and advanced engineering workflows, supplying a general platform for providing engineering and computational services in different contexts without sharing confidential IP or methodologies and providing a tool allowing workflow developers to offer customers tailored solutions on the cloud.


Faster engineering based on better simulations will allow the end user Borit to reduce the design time (value 10 k€/iteration). Based on the product complexity, up to 8 iterations can are required for a new product. Reducing this by 25 percent, adds up to € 70,000 to € 100,000 per year for five new products, to reduce the amount of non-productive (test-)time on the press (100 €/h). Virtual testing can reduce the test hours on press (8 hours/test) also by 25 percent - a total reduction of 7 to 10 days can be achieved which equals € 6,000 to €8,000 per year, to reduce the number of tools to be produced (value up to € 20,000 per tool). For five new products (average complexity) Borit expects a reduction of 5 to 7 test tools and to reduce time to market (value for the fuel cell manufacturer). Moreover, quality-improved forming capabilities and increased customer satisfaction will attract additional customers. This may create additional revenue of several hundreds of thousands of € per year. The reduction in design time and total time to market will allow the existing employees to handle more projects per year.

For Noesis, the ISV, the CloudFlow-powered engineering workflows are expected to lead to 2-3 users to join the platform during the first year: this quantifies to about to about € 120,000 per year initially and increase with time to € 400,000 per year as the user-base grows. These revenues accounts for all the items in the ManuCloud service subscription, that includes the Optimus optimizer deployed on the CloudFlow infrastructure as well as the tools for the workflow preparation and post-processing that are made available to the user.

On top of these estimates, there is the possibility to add engineering services offered to meet specific customer needs in terms of workflow complexity, specific optimization routines to be implemented, ad-hoc deployments on their production machines. These services could produce about € 30,000 to € 40,000 in the first year to € 200,000 in the 3rd year after the end of the application experiment. Since these are not recurring revenues, the sustainability of the engineering services business is supported by the continuous innovation of the end-user products and the increasing expertise of the end users with new and more sophisticated analyses done with Optimus. Eventually this expected volume of business could translate to 4 new FTEs in the first 3 years (1 in the first year and the others as soon as the business takes off) as well as in the consolidation of the current jobs against adverse market conditions.