By Bobby Carlton
AI driven defect detection is improving the quality of products and increasing efficiency of Siemens’ operations.
Siemens is accelerating their efforts to develop defect detection models using 3D synthetic data generated by NVIDIA Replicator. This latest advancement in manufacturing technology is part of the Siemens goals to advance digital twins in the industrial sector.
It should be noted that FS Studio is a channel partner with Nvidia!
Through NVIDIA’s Replicator tool, Siemens is able to build connections between its digital twins and software-defined AI, and these connections allow engineers to create full-fidelity models that can be accessed and operated by cloud-based systems.
AI-Driven Defect Detection
Due to the complexity of manufacturing processes, AI-driven defect detection could help improve the quality of products and increase the efficiency of Siemens’ operations, or the operation of any manufacturing company.
Unfortunately, developing AI models requires a lot of data and complex training methods. This process can be time-consuming and costly. Although the data may cover various types of defects, it might not be able to identify all of them.
Synthetic Data for Inspecting Motion Control Devices
For over 30,000 customers, Siemens’ Motion Control business unit manufactures motors, drives, and inverters. Its lead electronics plant in Germany, GWE, is currently using AI-based computer vision technology to detect defects.
By using synthetic data, Siemens is able to create sets of realistic images using its digital models and production resources, which can than be used to train AI models. In a recent press release, Nvidia announced that the lead electronics plant, GWE, based in Erlangen, Germany, has been working on AI-enabled computer vision for defect detection using custom methods and different modes of synthetic data generation.
Through the use of this synthetic data, GWE is now able to speed up the time it takes to develop AI models for inspection and can now accelerate the time it takes to develop AI models by up to 90%. This technology has helped the company create models that are more accurate and faster than before.
As a result, Siemens has begun tapping into NVIDIA Omniverse Replicator running on Amazon G5 instances for synthetic data generation, accelerating its AI model development times from taking “months” to “days,” according to the company.
This use of synthetic data has been instrumental in accelerating the development of AI models. It can help companies such as Amazon Robotics train their robots to identify packages, or assist manufacturing companies in many industries improve their efficiencies and synthetic data can be used to perform other tasks such as safety monitoring and robotic bin picking.
By 2024, 60% of the data used for the development of AI and analytics projects will be synthetically generated.
– Gartner Inc.
Synthetic data can help with shortening this workflow and making it more robust by addressing some of the pain points in the data collection and annotation stages:
- Data Collection – Theoretically, an infinite amount of synthetic data can be made available without having to set up the physical environment. This is especially beneficial for data-constrained scenarios, i.e., where the amount of real data that can be collected is limited to non-existing, or that it is very hard to obtain. For instance, if an existing manufacturing line must be stopped to collect the training data it could incur potential production losses. Synthetic data can also provide a much larger variation than the one typically observed when collecting real data. For instance, in a virtual 3D environment it is easy to create varying light or other physical conditions while in the real environment there is generally limited control over these parameters. Thus, utilizing synthetic data can improve the machine learning model’s ability to generalize well when deployed in environments that it has not encountered before.
- Annotation – Manually annotating data is often regarded as a repetitive, mundane task. Or, as it was phrased in a recent article by Google Research: “Everyone wants to do the model work, not the data work“. Often, the human workforce that is annotating the objects lack domain expertise or proper guidance and this leads to non-exact or simply wrong annotations. On the other hand, synthetic data is always accurately annotated, as the annotations (bounding boxes, object contours, etc.) are generated automatically based on complete knowledge of how the synthetic data was formed. This reduces annotation errors that are typical in manual annotation projects.
According to Alex Greenberg, the director of Siemens’ advanced robotics simulation, the company’s AI models can be trained using synthetic data without compromising their accuracy. Unfortunately, the traditional methods of data generation weren’t able to provide the necessary robustness for certain use cases that Greenberg needed. This led to the need for more accurate labeling and data acquisition.
According to Zac Mann, the lead developer of Siemens’ advanced robotic simulation, the rapid evolution of products and materials has created a need for automated manufacturing processes in every industry.
How GWE Used Synthetic Data to Catch Printed Circuit Board Defects
One of the biggest challenges that GWE has to face is the detection of defects during the early stages of production for new lines and products. Defects in printed circuit boards are usually focused on the thermal paste used on certain components. This type of material helps transfer heat from the attached components to the heatsink.
To identify potential issues with printed circuit boards, the software division of Siemens used synthetic data collected by Omniverse Replicator by using AI and synthetic data to create a simulated 3D pipeline and virtual world simulation, which they can use to create realistic and custom scenarios.
The integration of synthetic data into Siemens’ simulation process allowed them to accelerate the development of its products and solutions, and assisted in catching circuit board defects.
Through the use of Siemens SynthAI and Omniverse Replicator, they were able to generate realistic and custom 3D models of its products and manufacturing resources. This process can help improve the efficiency of its inspection models. According to Maximilian Metzner, the global lead for electronics at GWE, the use of synthetic data can speed up the training of AI inspection models by about five times.
Tapping Into Randomization With SynthAI
Through the use of Siemens’ SynthAI software, GWE engineers are able to import 3D CAD models of printed circuit boards into their AI training programs. This approach gave the company powerful randomization capabilities through Omniverse Replicator, and in turn, GWE can now generate a wide variety of 3D models of its products and manufacturing resources.
Tapping into Replicator, SynthAI can access its powerful randomization features to vary the sizes and locations of defects, change lighting, color, texture and more to develop a robust dataset.
Once data is generated with Replicator, it can be run through a defect detection model for initial training. This enables GWE engineers to quickly test and iterate on models, requiring only a small set of data to begin.
“This gives you visibility earlier into the design phase, and it can shorten time to market, which is very important,” said Greenberg.
If you’re interested in using AI and synthetic data for your company, FS Studio is a channel partner with Nvidia. Contact us to schedule a call!
