Simulating Machining Processes in the Cloud (CAM in the Cloud)

 Simulating and optimizing the manufacturing process before the machines actually start making a new product is one of the key stages in manufacturing engineering. The aim is to minimize manufacturing time, to avoid wasting raw material (resources) and to safeguard the machines from being damaged. One has to compute many possible tool paths, assess them, select an optimum one and finally generate the Numerical Control (NC) code to actually run the machine. All these steps are very time-consuming and minimizing the time to find the best possible solution is crucial concerning the costs for the company.

 

Challenge
The relevant process for Stellba in this study is the computation of the best tool path to machine a Kaplan turbine blade. To find an optimal tool path requires many selections and decisions by the engineer, e.g. material, methodology, etc., and each chosen configuration requires a dedicated simulation run. As these simulations are basically independent from each other, using a parallel computing infrastructure should speedup the iterative process and should allow computing more options to better explore the ‘design space’ and find ‘uncommon’ solutions. The GridWorker software tool by Fraunhofer EAS (a German research institute) is being used to parallelize computations as much as possible to reduce the overall time used. Through GridWorker the available HPC resources are deployed on a number of virtual machines to exploit the power of many computational cores at the same time. 

Benefits
As a result, the HPC resources can enable Stellba to simulate more complex machining tasks more quickly. In fact, the time to compute a best possible toolpath is now only 1/3 of what was necessary before. This provides the opportunity to increase the quality of the machining. Tool paths are now calculated in parallel. The CAM workflow allows the end user to prepare all data sets at once to produce a good machining plan and execute them at once and in parallel in the Cloud instead of having to wait for each individual result in front of his desktop before the next variant can be computed. No high-end number crunchers are needed locally by the end user since the computing power is provided in the Cloud.

Organizations involved
University of Nottingham- United Kingdom
Arctur - Slovenia
Stellba Hydro GmbH & Co KG- Germany