Case Study
Request a quote
Back to Blog
13 September 2023

Custom Datasets, Synthetic Data, and Simulation Will Drive Factories Towards Success

By Bobby Carlton

FS Studio Case Study on Custom Datasets Used for Machine Learning Algorithms

Our client saw the value in machine learning and needed custom high-fidelity virtual scenes that would be used to create important datasets for machine learning algorithms. FS Studio took unusable pre-made virtual assets and turned them into a library of datasets that the client was able to use with machine learning algorithms to give them valuable information that was data rich.

Problem: The customer owned a library of pre-made virtual assets that they wanted to use to create custom datasets for machine learning algorithms. Unfortunately, the pre-made virtual assets were either broken or they needed additional work done to them for them to work in their virtual environment. The pre-made assets owned by the client needed to be modified. Without modification, the client would not be able to achieve their goals.

Solution: FS Studio worked closely with the client by taking the pre-made library assets they purchased and converted them into usable virtual assets that allowed the client to place avatars and high-fidelity 3D models of their own hardware into the scenes. FS Studio disassembled the pre-made files using Houdini to put them back together to leverage python scripts. This was a high touch process for FS Studio that showcased the incredible work done by our art team.

Result: Through the work at FS Studio, the client was able to take their library of scenes and have more control and flexibility over how they used them and could implement them for their own data set and was able to achieve their goals with machine learning giving them a robust dataset to train their algorithms.

Lets Dig Deeper into Why Custom Datasets, Synthetic Data, and Simulation are Important

Custom datasets, synthetic data, and simulations are crucial tools for the future of factories as they provide a cost-effective, risk-reducing, and innovative means to optimize operations, train employees, reduce environmental impact, and ensure compliance with regulations. By leveraging these technologies, manufacturers can adapt more rapidly to evolving market demands and maintain their competitiveness.

These technologies are indispensable tools in helping enterprises thrive and innovate within a virtual framework. They enable organizations to replicate real-world scenarios with precision and flexibility, fostering growth in various ways.

  • Custom datasets tailored to specific business needs enable more accurate analysis, decision-making, and predictions, thus optimizing resource allocation and operational efficiency.
  • Synthetic data generation allows companies to harness the power of artificial intelligence and machine learning without compromising sensitive or limited datasets. This not only enhances the development of cutting-edge algorithms but also ensures data privacy and security compliance.
  • Simulations provide a risk-free environment for testing new strategies, products, or services, helping enterprises anticipate and mitigate potential challenges while accelerating innovation.

By leveraging these tools, businesses can reduce costs, minimize risks, and drive continuous improvement, ultimately propelling their growth in an ever-evolving digital landscape.

Custom Datasets to Train Machine Learning Models

Custom datasets can significantly enhance the efficiency and productivity of factories across various industries. By creating and utilizing tailored datasets, factories can fine-tune their operations, optimize processes, and make data-driven decisions. One key benefit is the ability to gather real-time data from sensors, machinery, and production lines, enabling predictive maintenance. This proactive approach to equipment upkeep can prevent costly breakdowns and downtime, ensuring smoother and uninterrupted production. Moreover, custom datasets can be used to monitor the quality of products throughout the manufacturing process, enabling early detection of defects and reducing waste.

Additionally, custom datasets can facilitate the implementation of advanced automation and robotics in factories. These datasets can train machine learning models to recognize patterns and anomalies in production, allowing robots to perform tasks with greater precision and efficiency. This not only increases the overall output but also improves worker safety by relegating dangerous or repetitive tasks to automated systems. Furthermore, custom datasets can aid in supply chain management by providing insights into demand forecasting and inventory optimization.

Factories can align their production schedules with market demands, reduce excess inventory, and minimize storage costs, ultimately streamlining their operations and improving their bottom line. In essence, custom datasets are a vital tool for modernizing and revolutionizing factory operations, making them more efficient, cost-effective, and adaptive to evolving market dynamics.

Custom Datasets

Enhancing Real-World Decision-Making with Synthetic Data

Boeing serves as a prime example of embracing the industrial concept. Their aspiration is to develop the next-generation aircraft within this virtual realm. A pivotal aspect of Boeing’s vision involves constructing digital replicas, known as digital twins, of actual objects or systems. By generating digital twins for both the aircraft and the manufacturing process, Boeing can simulate intricate operations before implementing them in real-world manufacturing settings. Essentially, in this industrial paradigm, insights from simulated data inform real-world actions.

Central to this approach is the idea of synthetic data, data produced through simulations and algorithms, as opposed to relying solely on data from the physical world. For Boeing, this proved immensely beneficial, as creating an aircraft inspection system based solely on real-world data, like thousands of plane photographs, proved exceedingly challenging. To overcome this data scarcity, they crafted a digital twin of the aircraft, generating over 100,000 simulated images—a feat unattainable and cost-prohibitive in the physical world. Armed with this abundant simulated data, combined with real-world imagery, their system effectively compared the present state of an aircraft with its historical condition using augmented reality.

Simulating Various Entities, from Entire Factories to Individual Products

Digital twins hold substantial value in factory settings. Siemens Digital Native Factory in Nanjing, China, serves as a testament to this approach, where they simulated the entire factory during the planning phase. This enabled Siemens to identify and rectify planning errors, optimizing the construction process. Siemens reports a 200 percent increase in capacity and a 20 percent boost in productivity due to this digital counterpart.

Beyond planning, digital twins continuously prove their worth in enhancing existing factory operations. Sensors scattered across the plant feed data to the digital twin, which can then be analyzed to gain insights into the system’s performance. Based on these insights, manufacturers can fine-tune workflows and experiment with changes within the digital twin before implementing them in reality. In essence, digital twins serve as a valuable tool for informed decision-making in the manufacturing realm.


Predicting Outcomes

The concept of digital twins extends to replicating individual machines with precision. Siemens, in collaboration with NVIDIA, is pioneering the creation of digital twins that closely resemble physical machines and facilitate real-time interaction. This technology empowers engineers to manipulate parameters, such as temperature, within the digital twin, providing precise insights into real-world machine behavior. Moreover, it aids in predicting machine failures and optimizing maintenance schedules.

Beyond machinery, products themselves can be subjected to simulations for predictive purposes. This not only holds the potential to revolutionize product design but also provides valuable insights into the performance of existing products in real-world scenarios. For instance, Kaeser, a manufacturer of compressed air and vacuum products, employs digital twins to replicate customer-deployed compressed air systems. This enables the monitoring of product conditions, remote fault detection, and timely maintenance.

Immersive and Collaborative Product Design with VR

Digital twins play a pivotal role in streamlining the product design process by enabling designers to simulate, identify flaws, and refine designs before physical production. This approach addresses quality issues effectively, with Boeing citing that over 70 percent of its quality concerns stem from design-related factors.

Another technology closely related to the industrial concept is virtual reality (VR). Designers can immerse themselves in an interactive VR environment, enabling them to envision and modify products in a fully immersive space. Picture, for instance, designing an automobile by stepping into a VR rendition of its interior. The future holds the promise of designers increasingly adopting VR headsets for an enriched design experience. When integrated with digital twins, changes made in the virtual metaverse space reflect instantly in the digital twin, offering real-time design adjustments.

Immersive VR design spaces also facilitate remote collaboration among design teams, bridging geographical gaps. Tools like NVIDIA’s Omniverse enable real-time collaborative design, allowing teams to work together seamlessly, as if they were physically present, with immediate visual feedback on changes made.

Key takeaways from adopting simulation, custom datasets, and synthetic data:

  • Cost-Efficiency: Setting up and maintaining real-world manufacturing processes and equipment can be expensive. Synthetic data and simulations offer a cost-effective alternative for testing and refining factory processes without the need for physical resources. This can lead to significant cost savings in terms of materials, labor, and energy consumption.
  • Risk Reduction: Simulations allow manufacturers to identify and mitigate potential risks and bottlenecks in a controlled environment. This reduces the likelihood of costly errors, accidents, and production delays when implementing new processes or technologies in real factories.
  • Faster Innovation: The rapid pace of technological advancement means that factories need to adapt quickly to remain competitive. Simulations enable manufacturers to experiment with new ideas and technologies more rapidly, speeding up the innovation cycle.
  • Optimization: Synthetic data and simulations enable the optimization of factory operations. Manufacturers can fine-tune processes to maximize efficiency, minimize waste, and improve product quality, all without disrupting actual production.
  • Training and Skill Development: Simulations provide a safe and controlled environment for training factory workers and engineers. This can be especially valuable for onboarding new employees and upskilling existing ones. Workers can gain hands-on experience with complex machinery and processes without the risk of accidents.
  • Environmental Impact: Factories have a significant environmental footprint. Simulations can help reduce this impact by optimizing resource utilization and energy consumption, ultimately contributing to sustainability goals.
  • Scaling and Flexibility: Manufacturers can use simulations to test different scenarios for scaling production up or down as demand fluctuates. This flexibility is essential in a rapidly changing market.
  • Remote Monitoring and Control: With the integration of Internet of Things (IoT) devices and sensors, simulations can be used for remote monitoring and control of factory processes. This allows for real-time adjustments and troubleshooting, even from a distance.
  • Quality Control: Simulations can be used to model and predict product quality, helping manufacturers maintain consistent quality standards and reduce defects.
  • Regulatory Compliance: Meeting regulatory requirements in manufacturing can be complex. Simulations can assist in ensuring compliance by testing processes and designs against regulatory criteria before implementing them in the real world.

To learn more about our work in custom datasets, synthetic data and simulation, reach out to us for a discovery call on how our team can help you.