Hero image

FAQs Page

Connect with our team
Data Handling
and Integration
AI Model Development and Management
Platform Features
and Capabilities
Security, Privacy, and Data Ownership
Getting Started with Key Ward

1. What is the maximum size of models that Key Ward can process?

  • Cloud Environment: Key Ward can handle models of any size in the cloud, thanks to its ability to scale resources both vertically and horizontally. This ensures that even the most complex and largest models can be processed without limitations.
  • Local Environment: The capacity is limited by your local hardware. For reference, Key Ward has successfully processed models as large as 50 million cells on local systems.

2. How long does it take to process large datasets?

  • Data Extraction: The time required depends on the size of the file and can range from a few seconds to several minutes.
  • AI Data Preparation: This varies significantly based on the file size and the specific processing steps involved, such as interpolation, mesh decimation, or creating signed distance fields. Times can range from seconds to minutes depending on the complexity.

3. What are the memory and hardware requirements for processing data on Key Ward?

  • Data Extraction: Typically requires 1-2GB of RAM per 1 million cells, similar to the memory requirements of running a CFD simulation.
  • AI Data Preparation: Memory requirements can vary depending on the processing techniques applied, averaging around 1-1.5GB of RAM per 1 million cells.

4. Can Key Ward process models with fine details (e.g., <1mm gaps)?

Yes, Key Ward provides full control over the level of detail in the final model. Users can balance accuracy and performance by adjusting down-sampling parameters. The platform has experience reducing models from 29 million cells down to 1 million, 500k, or even 100k cells while preserving critical details.

5. How customizable is the down-sampling process in Key Ward?

Key Ward offers customizable down-sampling operators based on curvature or field gradients. This allows users to preserve important details in areas with strong curvature or high field gradients, making the down-sampling process highly adaptable to specific project needs.

6. How much data is required for effective AI model training on Key Ward?

The amount of data required depends on the complexity of the physics involved and the goals of the AI model. While some applications may perform well with only a handful of results, others might need dozens or even hundreds of datasets. Generally, it is recommended to start with at least 30-40 simulations before evaluating predictive quality. The key is to focus on the quality of the data and not the quantity.

Image of Mean relative L2 error with confidence band and moving average for pressure dragImage comparing different accuracies of AI predictions on CFD simulations with different number of training datapoints

7. Can Key Ward automatically ingest new data as it is generated?

Yes, Key Ward allows for the automation of data ingestion. Once data extraction settings are configured, the platform can be set to automatically update and process new data as it becomes available in specified locations.

8. What types of data systems can Key Ward integrate with (e.g., Siemens Teamcenter, SQL, Snowflake)?

Key Ward supports integration with a variety of data systems:
  • APIs: Integrates with Siemens Teamcenter/NX via APIs.
  • Databases: Supports SQL-based systems, including PostgreSQL, and can integrate with Snowflake for data storage and querying.

9. Does Key Ward support the processing of complex multi-step data (e.g., different ride heights or operating conditions)?

 Yes, Key Ward can handle multi-step data without impacting the data extraction process. For AI processing, the platform accommodates datasets with high variability, but these require a larger number of training examples to maintain model quality. The platform includes a confidence score to help quantify the similarity between new designs and training data, ensuring accurate predictions.

1. What types of AI models can be created and trained on Key Ward?

Key Ward supports a wide range of AI models, including traditional machine learning models, reduced-order models (ROMs), and deep learning models. The platform is designed to accommodate various data types and complexities, enabling the development of models tailored to specific engineering applications.

2. Does Key Ward offer automatic hyperparameter tuning?

Yes, Key Ward provides automatic hyperparameter tuning for ROMs and machine learning models using techniques like GridSearchCV and RandomizedSearchCV. This feature helps optimize model performance without manual intervention, saving time and improving accuracy.

3. How can I ensure the accuracy of AI models created on Key Ward?

Key Ward incorporates several features to ensure model accuracy:

  • Confidence Scores: These scores indicate how confident the AI model is in its predictions, particularly when the input data deviates from the training data.
Image showing prediction bounds for predicted drag
  • Data Quality: High-quality and representative training data are essential. Key Ward helps manage and prepare data to improve model accuracy.

4. Can AI training be automated on Key Ward?

Yes, once an initial training pipeline is established, Key Ward allows for full automation of AI model training and retraining. For example, you can configure the platform to retrain a model after a specific number of new data points have been processed, ensuring continuous model improvement.
Image comparing different accuracies of AI predictions on CFD simulations with different number of training datapoints

5. How does Key Ward handle AI model updates and retraining?

 Key Ward supports automated model retraining by linking it to the data preparation pipeline. As new data becomes available and is processed, the platform can automatically update the model, ensuring that it remains accurate and relevant over time.

6. What is the confidence score in Key Ward, and how is it used?

The confidence score in Key Ward is a metric that quantifies the uncertainty of the AI model's predictions. It is particularly useful when input data is significantly different from the data the model was trained on. A lower confidence score suggests that the model's prediction might be less reliable, guiding users to be cautious or to gather additional data.

7. Can models trained on Key Ward be exported for use in other platforms (e.g., FMU, Python)?

Yes, models trained on Key Ward can be exported in several formats, including Functional Mock-up Units (FMU), hdf5, and Python pickle (pkl) files. This flexibility allows you to integrate the models into other workflows or platforms seamlessly.

1. How can Key Ward reduce FEA and CFD workload?

Key Ward automates the entire workflow from data extraction of CFD and FEA simulations to model predictions using machine learning, ROMs, and 3D deep learning models. This automation eliminates the need for manual coding, significantly reducing the time and effort required to manage and analyze complex simulation data. Additionally, Key Ward supports the merging of test data with CFD and FEA data, enhancing AI model training and data analysis capabilities.

2. Can Key Ward automate the entire data processing pipeline?

Yes, Key Ward allows full automation of the data processing pipeline. Once set up, the platform can automatically process new data, prepare it for AI modeling, and even trigger retraining of AI models based on new data points. This ensures a seamless and efficient workflow, minimizing the need for manual intervention.

3. What boundary zones and volume regions are supported by Key Ward?

Key Ward supports an unlimited number of boundary zones and volume regions, regardless of complexity. It can process boundary zones of any type and handle multiple volume regions, including those with different material properties. This flexibility makes Key Ward suitable for a wide range of engineering applications.

4. Can Key Ward's features be customized for specific use cases?

Yes, Key Ward offers a high degree of customization to meet specific project requirements. Users can tailor down-sampling, data processing, and AI modeling settings to preserve critical details or optimize performance for their unique applications. The platform also supports the creation of templates for recurring tasks, streamlining the workflow for specific use cases.

5. Does Key Ward offer a no-code environment for AI and data processing?

Yes, Key Ward is designed as a no-code platform, enabling users to manage, process, and analyze data, as well as train AI models without needing to write code. This feature is particularly beneficial for engineers who may not have extensive programming experience, as it simplifies the process of leveraging advanced AI and data analytics tools.

6. How does Key Ward handle scripting and automation for complex workflows?

While Key Ward provides a no-code environment, it also supports scripting and automation for users who need to implement complex workflows. The platform allows users to script custom processes and automate entire pipelines, ensuring that even the most intricate tasks can be efficiently managed within the Key Ward environment.

1. Who owns the data uploaded to Key Ward?

Clients retain full ownership of all data uploaded to Key Ward. The platform ensures that your data remains your property, and it will not be accessed or used by anyone else unless explicitly permitted by you.

2. How does Key Ward ensure data privacy and security?

Key Ward is built on Amazon Web Services (AWS) infrastructure, utilizing the Frankfurt (europe-west3) region for data processing. The platform is designed with strict security protocols to protect business data, including encryption, access controls, and regular security audits. Additionally, Key Ward offers a single-tenant deployment option for enhanced data isolation.

3. Does Key Ward offer single-tenant deployment for enhanced security?

Yes, Key Ward offers a single-tenant deployment model, where each client has a completely isolated instance of the application and database. This ensures that your data and processes are securely separated from those of other clients, providing maximum security and customization flexibility.

4. What is the process for handling sensitive data on Key Ward?

Key Ward is intended for processing business data, not personal data under GDPR. If you need to handle sensitive data, you must notify Key Ward in advance. The platform processes and stores metadata and machine learning models derived from your content in an anonymized form for quality control and product improvement purposes. If you require enhanced security, the single-tenant model offers strict isolation of customer databases and processes, ensuring sensitive data is handled securely.
Data extraction can happen locally or in the cloud. If locally, all simulations and CAD files stay on premise.

1. What are the installation and setup requirements for Key Ward?

Key Ward is a cloud-based platform, meaning there is no complex installation process. Users can access the platform directly through a web browser. For on-premises deployments, specific hardware and software requirements will be provided, tailored to your local environment's capabilities. Key Ward's team offers support to ensure a smooth setup and integration process.

2. How can my team start using Key Ward?

To get started with Key Ward, your team can sign up for an account and access the platform via the cloud. Key Ward provides detailed onboarding resources, including tutorials, documentation, and customer support, to help your team quickly learn how to use the platform's features effectively. For enterprise clients, Key Ward offers tailored training sessions to ensure your team can fully leverage the platform’s capabilities.

3. What support is available for onboarding and using Key Ward?

Key Ward offers comprehensive support during the onboarding process, including access to tutorials, user guides, and customer service. Additionally, for enterprise clients, Key Ward provides personalized training sessions and dedicated account managers to assist with any questions or issues that arise. Continuous support is available to ensure users can maximize the platform's benefits.

4. Can I use my existing data and models with Key Ward?

Yes, Key Ward is designed to integrate seamlessly with your existing data and models. The platform supports a wide range of data formats and can import models from various engineering tools. This compatibility ensures that you can leverage your current data assets within Key Ward’s powerful AI and data processing environment, enhancing your workflows without starting from scratch.

5. How can my team get started with AI using Key Ward?

Key Ward is specifically designed to lower the barrier to entry for AI adoption. With its no-code environment, even teams without extensive AI expertise can start building and training models. The platform provides guided workflows, pre-built templates, and AI tools that simplify the process of setting up AI-driven projects. For teams looking to explore more advanced AI techniques, Key Ward offers resources and support to scale up your AI initiatives.
Data Handling and Integration

1. What is the maximum size of models that Key Ward can process?

  • Cloud Environment: Key Ward can handle models of any size in the cloud, thanks to its ability to scale resources both vertically and horizontally. This ensures that even the most complex and largest models can be processed without limitations.
  • Local Environment: The capacity is limited by your local hardware. For reference, Key Ward has successfully processed models as large as 50 million cells on local systems.

2. How long does it take to process large datasets?

  • Data Extraction: The time required depends on the size of the file and can range from a few seconds to several minutes.
  • AI Data Preparation: This varies significantly based on the file size and the specific processing steps involved, such as interpolation, mesh decimation, or creating signed distance fields. Times can range from seconds to minutes depending on the complexity.

3. What are the memory and hardware requirements for processing data on Key Ward?

  • Data Extraction: Typically requires 1-2GB of RAM per 1 million cells, similar to the memory requirements of running a CFD simulation.
  • AI Data Preparation: Memory requirements can vary depending on the processing techniques applied, averaging around 1-1.5GB of RAM per 1 million cells.

4. Can Key Ward process models with fine details (e.g., <1mm gaps)?

Yes, Key Ward provides full control over the level of detail in the final model. Users can balance accuracy and performance by adjusting down-sampling parameters. The platform has experience reducing models from 29 million cells down to 1 million, 500k, or even 100k cells while preserving critical details.

5. How customizable is the down-sampling process in Key Ward?

Key Ward offers customizable down-sampling operators based on curvature or field gradients. This allows users to preserve important details in areas with strong curvature or high field gradients, making the down-sampling process highly adaptable to specific project needs.

6. How much data is required for effective AI model training on Key Ward?

The amount of data required depends on the complexity of the physics involved and the goals of the AI model. While some applications may perform well with only a handful of results, others might need dozens or even hundreds of datasets. Generally, it is recommended to start with at least 30-40 simulations before evaluating predictive quality. The key is to focus on the quality of the data and not the quantity.

Image of Mean relative L2 error with confidence band and moving average for pressure dragImage comparing different accuracies of AI predictions on CFD simulations with different number of training datapoints

7. Can Key Ward automatically ingest new data as it is generated?

Yes, Key Ward allows for the automation of data ingestion. Once data extraction settings are configured, the platform can be set to automatically update and process new data as it becomes available in specified locations.

8. What types of data systems can Key Ward integrate with (e.g., Siemens Teamcenter, SQL, Snowflake)?

Key Ward supports integration with a variety of data systems:
  • APIs: Integrates with Siemens Teamcenter/NX via APIs.
  • Databases: Supports SQL-based systems, including PostgreSQL, and can integrate with Snowflake for data storage and querying.

9. Does Key Ward support the processing of complex multi-step data (e.g., different ride heights or operating conditions)?

 Yes, Key Ward can handle multi-step data without impacting the data extraction process. For AI processing, the platform accommodates datasets with high variability, but these require a larger number of training examples to maintain model quality. The platform includes a confidence score to help quantify the similarity between new designs and training data, ensuring accurate predictions.
AI Model Development and Management

1. What types of AI models can be created and trained on Key Ward?

Key Ward supports a wide range of AI models, including traditional machine learning models, reduced-order models (ROMs), and deep learning models. The platform is designed to accommodate various data types and complexities, enabling the development of models tailored to specific engineering applications.

2. Does Key Ward offer automatic hyperparameter tuning?

Yes, Key Ward provides automatic hyperparameter tuning for ROMs and machine learning models using techniques like GridSearchCV and RandomizedSearchCV. This feature helps optimize model performance without manual intervention, saving time and improving accuracy.

3. How can I ensure the accuracy of AI models created on Key Ward?

Key Ward incorporates several features to ensure model accuracy:

  • Confidence Scores: These scores indicate how confident the AI model is in its predictions, particularly when the input data deviates from the training data.
Image showing prediction bounds for predicted drag
  • Data Quality: High-quality and representative training data are essential. Key Ward helps manage and prepare data to improve model accuracy.

4. Can AI training be automated on Key Ward?

Yes, once an initial training pipeline is established, Key Ward allows for full automation of AI model training and retraining. For example, you can configure the platform to retrain a model after a specific number of new data points have been processed, ensuring continuous model improvement.
Image showing continuous update on datasets using HUB

5. How does Key Ward handle AI model updates and retraining?

 Key Ward supports automated model retraining by linking it to the data preparation pipeline. As new data becomes available and is processed, the platform can automatically update the model, ensuring that it remains accurate and relevant over time.

6. What is the confidence score in Key Ward, and how is it used?

The confidence score in Key Ward is a metric that quantifies the uncertainty of the AI model's predictions. It is particularly useful when input data is significantly different from the data the model was trained on. A lower confidence score suggests that the model's prediction might be less reliable, guiding users to be cautious or to gather additional data.

7. Can models trained on Key Ward be exported for use in other platforms (e.g., FMU, Python)?

Yes, models trained on Key Ward can be exported in several formats, including Functional Mock-up Units (FMU), hdf5, and Python pickle (pkl) files. This flexibility allows you to integrate the models into other workflows or platforms seamlessly.
Platform Features and Capabilities

1. How can Key Ward reduce FEA and CFD workload?

Key Ward automates the entire workflow from data extraction of CFD and FEA simulations to model predictions using machine learning, ROMs, and 3D deep learning models. This automation eliminates the need for manual coding, significantly reducing the time and effort required to manage and analyze complex simulation data. Additionally, Key Ward supports the merging of test data with CFD and FEA data, enhancing AI model training and data analysis capabilities.

2. Can Key Ward automate the entire data processing pipeline?

Yes, Key Ward allows full automation of the data processing pipeline. Once set up, the platform can automatically process new data, prepare it for AI modeling, and even trigger retraining of AI models based on new data points. This ensures a seamless and efficient workflow, minimizing the need for manual intervention.

3. What boundary zones and volume regions are supported by Key Ward?

Key Ward supports an unlimited number of boundary zones and volume regions, regardless of complexity. It can process boundary zones of any type and handle multiple volume regions, including those with different material properties. This flexibility makes Key Ward suitable for a wide range of engineering applications.

4. Can Key Ward's features be customized for specific use cases?

Yes, Key Ward offers a high degree of customization to meet specific project requirements. Users can tailor down-sampling, data processing, and AI modeling settings to preserve critical details or optimize performance for their unique applications. The platform also supports the creation of templates for recurring tasks, streamlining the workflow for specific use cases.

5. Does Key Ward offer a no-code environment for AI and data processing?

Yes, Key Ward is designed as a no-code platform, enabling users to manage, process, and analyze data, as well as train AI models without needing to write code. This feature is particularly beneficial for engineers who may not have extensive programming experience, as it simplifies the process of leveraging advanced AI and data analytics tools.

6. How does Key Ward handle scripting and automation for complex workflows?

While Key Ward provides a no-code environment, it also supports scripting and automation for users who need to implement complex workflows. The platform allows users to script custom processes and automate entire pipelines, ensuring that even the most intricate tasks can be efficiently managed within the Key Ward environment.
Security, Privacy, and Data Ownership

1. Who owns the data uploaded to Key Ward?

Clients retain full ownership of all data uploaded to Key Ward. The platform ensures that your data remains your property, and it will not be accessed or used by anyone else unless explicitly permitted by you.

2. How does Key Ward ensure data privacy and security?

Key Ward is built on Amazon Web Services (AWS) infrastructure, utilizing the Frankfurt (europe-west3) region for data processing. The platform is designed with strict security protocols to protect business data, including encryption, access controls, and regular security audits. Additionally, Key Ward offers a single-tenant deployment option for enhanced data isolation.

3. Does Key Ward offer single-tenant deployment for enhanced security?

Yes, Key Ward offers a single-tenant deployment model, where each client has a completely isolated instance of the application and database. This ensures that your data and processes are securely separated from those of other clients, providing maximum security and customization flexibility.

4. What is the process for handling sensitive data on Key Ward?

Key Ward is intended for processing business data, not personal data under GDPR. If you need to handle sensitive data, you must notify Key Ward in advance. The platform processes and stores metadata and machine learning models derived from your content in an anonymized form for quality control and product improvement purposes. If you require enhanced security, the single-tenant model offers strict isolation of customer databases and processes, ensuring sensitive data is handled securely.
Getting Started with Key Ward

1. What are the installation and setup requirements for Key Ward?

Key Ward is a cloud-based platform, meaning there is no complex installation process. Users can access the platform directly through a web browser. For on-premises deployments, specific hardware and software requirements will be provided, tailored to your local environment's capabilities. Key Ward's team offers support to ensure a smooth setup and integration process.

2. How can my team start using Key Ward?

To get started with Key Ward, your team can sign up for an account and access the platform via the cloud. Key Ward provides detailed onboarding resources, including tutorials, documentation, and customer support, to help your team quickly learn how to use the platform's features effectively. For enterprise clients, Key Ward offers tailored training sessions to ensure your team can fully leverage the platform’s capabilities.

3. What support is available for onboarding and using Key Ward?

Key Ward offers comprehensive support during the onboarding process, including access to tutorials, user guides, and customer service. Additionally, for enterprise clients, Key Ward provides personalized training sessions and dedicated account managers to assist with any questions or issues that arise. Continuous support is available to ensure users can maximize the platform's benefits.

4. Can I use my existing data and models with Key Ward?

Yes, Key Ward is designed to integrate seamlessly with your existing data and models. The platform supports a wide range of data formats and can import models from various engineering tools. This compatibility ensures that you can leverage your current data assets within Key Ward’s powerful AI and data processing environment, enhancing your workflows without starting from scratch.

5. How can my team get started with AI using Key Ward?

Key Ward is specifically designed to lower the barrier to entry for AI adoption. With its no-code environment, even teams without extensive AI expertise can start building and training models. The platform provides guided workflows, pre-built templates, and AI tools that simplify the process of setting up AI-driven projects. For teams looking to explore more advanced AI techniques, Key Ward offers resources and support to scale up your AI initiatives.