Simulation Machine Learning Automates Complex Verification Processes

Related Vendors

CADFEM Germany GmbH

GEMÜ Gebrüder Müller Apparatebau GmbH & Co. KG

FAULHABER Drive Systems

Gemü has over 2,000 metallic valve bodies in its portfolio, resulting in a large data pool of CAD geometries. For each individual part, proof of load capacity according to the Pressure Equipment Directive must be provided through simulation—a case for automation and AI.

Gemü Gebr. Müller Apparatebau GmbH & Co. KG is a leading manufacturer of valves as well as measurement and control systems for liquids, steam, and gases. The family-owned company from Ingelfingen in Hohenlohe operates globally and is known for innovation and digitalization, with the aim of delivering exceptional added value to its customers.

Simulations have had a firmly established role in Gemü's development processes for many years. Simulations are conducted using Ansys; the partner for everything related to simulation is Cadfem. Since its introduction, the range of applications has been continuously expanding—both in terms of physical domains and the ways in which simulation is utilized.

Automation of Simulation in Focus

A few years ago, Gemü began exploring the topic of simulation automation with PyAnsys in collaboration with Cadfem. The goal: to make processes more efficient, faster, and better. This is particularly suited for repetitive and standardizable tasks, which simulation engineers view as a necessary burden because it leaves them with less time for their demanding creative development tasks. Therefore, at Gemü, automation is also understood as a contribution to greater employee satisfaction.

Automate Monotonous Tasks

Metallic valve bodies are pressure-bearing components that require proof of load-bearing capacity. Gemü offers customized solutions and has well over 2,000 variants in its constantly growing program. The required proof is performed using simulation with Ansys based on each individual CAD geometry, which manually is a tedious and monotonous process. On the other hand, these are ideal conditions for automation.

Script-Based Automation Reduces Effort

The development of an automated workflow was tackled in collaboration with Cadfem. Beyond the realization of this specific project, Gemü's goal for the collaboration was the transfer of knowledge for future applications. One of the challenges with the valve bodies was that surface components and other metadata important for automation were not natively embedded in the CAD data. Instead, they had to be derived based on valve body type, nominal diameter, and connection type. Their variability and lack of standardization created a particularly complex situation, leading to the fact that the purely script-based automation approach using PyAnsys ultimately delivered “only” good results, which required significantly less effort than before but still had to be manually reworked.

Machine Learning Recognizes CAD Geometries

With the AI solution Stochos from Cadfem's partner PI Probaligence, it has subsequently been possible to fully automate the entire workflow—geometry preparation, modeling, standard-compliant evaluation according to the Pressure Equipment Directive, and reporting—while significantly improving the key factor: result quality.

Using Stochos, a machine learning model was trained with geometry-based metadata, which reliably processes the diverse properties of the CAD geometries and almost fully automates the recognition of pressure-loaded surfaces as well as inlet and outlet surfaces. These surfaces then directly serve as the basis for the further automated model setup of the non-parametric, "dead" geometries with PyAnsys.

I follow the principle: whatever is feasibly automatable should also be automated. This allows experts to focus on the exciting cases and explore new areas of application.

Marco Wissinger, Team Leader Simulation Engineering at Gemü

The automation solution is already being used productively at Gemü and relieves the simulation engineers. However, development is far from complete. Rather, this is just the beginning: in addition to the newly acquired expertise, individual workflow components can also be used in other automation applications.

Machine Authentication

Machine Operator Authentication - Why knowing "who" matters