AI and ML

AI and ML are redefining computational paradigms. AI mimics cognitive functions such as learning, problem-solving, and decision-making, while ML specifically focuses on developing algorithms that are capable of learning patterns from data without explicit programming. The principles of machine learning involve training models on large datasets, validating their accuracy, and continuously refining them to enhance their predictive capabilities.

Geometric Deep Learning

Geometric Deep Learning (GDL) is an emerging domain within AI that extends traditional deep learning methods to handle non-Euclidean data structures such as graphs, manifolds, and point clouds. Unlike conventional deep learning, which operates on grid-like data like images or sequences, GDL incorporates geometric principles to interpret complex and irregular structures. It leverages symmetries, invariances, and geometric priors to develop models capable of analyzing data with rich geometric and topological features. GDL has broad applications across computer vision, natural language processing, drug discovery, and social network analysis. By generalizing neural network architectures to non-Euclidean domains, GDL enables more nuanced and accurate modeling of real-world phenomena.

AI and ML significantly enhance the efficiency and accuracy of Finite Element Analysis (FEA) simulations across various fields. By integrating AI techniques, simulations can be executed faster and more effectively, reducing both computational time and cost. ML models trained on extensive datasets can predict simulation outcomes and optimize parameters, delivering more precise and reliable results. In Explicit dynamics, for example, machine learning algorithms can analyze data from crash, blast, metal forming; etc, traditionally time- and resource-intensive processes, and accelerate them through automated learnin.. Additionally, AI-powered reduced-order modeling (ROM) simplifies complex models, allowing engineers to assess system behavior with minimal computational overhead. This synergy between AI and simulation drives innovation by supporting rapid prototyping, more effective design cycles, and improved engineering decision-making.

Objectives and Scope of the Work

Key Terminologies

• Geometric Deep Learning (GDL): The core technology behind PhysicsAI. Unlike traditional machine learning which often relies on predefined parameters, , GDL directly operates on the geometric data (3D meshes and CAD models) from CAE simulations. It learns relationships between the geometric shape and full contour results.
• Dataset: A collection of historical CAE simulation results used for training or testing a ML models. Typically includes  input geometries and corresponding physics outputs (e.g., stress, strain, temperature, pressure fields).
• Data Source: PhysicsAI supports various CAE file formats  (including .h3d and others compatible with HyperView), enabling organizations to leverage legacy simulation data.
• Inputs: Model features used for predictions, such as cae.coord (spatial coordinates), cae.part_label (part names).
• Outputs: Physical quantities that model is trained to predict, such as pressure, stress, displacement, etc.
  
• Training Loss Curves: Diagnostic tool that chart the value of the loss function over training epochs. They help detect underfitting or overfitting by comparing:
• Training Loss: Error on the training dataset.
• Validation Loss: Error on unseen validation data. Ideally, both curves should converge Persistent gaps may indicate underfitting  or overfitting.
  
• Error Percentage Calculation: Used to quantify prediction accuracy:

Where:

Max_True is the maximum fringe value from the ground truth simulation.

Max_Pred is the maximum fringe value from the AI-predicted result.

Case Study

Automotive Bumper Crash Simulation

In the automotive industry, bumper crash tests are crucial for assessing vehicle safety and durability. These tests simulate real-world impacts to assess how effectively bumpers can absorb energy and protect occupants. 

Methodology

A bumper impact analysis model was developed to study the influence of design parameters on performance. The following parameters were varied:

Fig: Picture showing the Parametric Model Used.

These variations resulted in 45 distinct models, forming the dataset used for ML training through a Design of Experiments (DOE) approach.

Model Development

Altair PhysicsAI, powered by GDL, was used to train machine learning models on the collected dataset. Separate models were developed to predict three mechanical outputs: displacement, stress, and strain. The dataset was split 80/20 for training and testing to ensure robust evaluation.

Loss curves for each output type were monitored during training to assess model performance and convergence. The resulting curves confirmed that the models generalize well to unseen geometries.

Validation and Testing

Model testing:

The trained ML models were then tested using test dataset. This step involved comparing the true results with the predicted results from the AI models to evaluate their accuracy and reliability.

The Error Percentage is 0.163415278%

The Error Percentage is 2.1958%

The Error Percentage is 1.1%

Model Prediction:

For the final step, predict results for a new geometry with the relevant ML models. Using the developed models, successfully predicted displacement, stress, and strain for the new geometry.

The Error Percentage is 1.006%

The Error Percentage is 6.6%

The Error Percentage is 3.76%

The Confidence score signifies that the new design approaches closer to the trained models. This will vary from 0-1.

Effectiveness of AI and ML

The case study illustrates how PhysicsAI drastically reduces simulation costs and time using geometric deep learning. Unlike conventional simulation methods that require extensive computational resources, PhysicsAI learns from previous simulations and directly applies insights to new designs. Predictions are generated within seconds or minutes, even for complex geometries.

• Manual effort is significantly reduced. Traditional workflows often require extensive preprocessing, geometry cleanup, meshing, and setup. With PhysicsAI, much of this is automated, enabling engineers to focus on high-value tasks such as analysis and design optimization.
• Rapid design exploration becomes possible with the platform’s cloud-connected burst compute capabilities, which allow hundreds of simulations to run simultaneously across GPUs. This is particularly impactful in crash scenarios, where multiple variants must be evaluated for compliance and safety. Engineers can quickly iterate and optimize designs, resulting in safer, higher-performing components.
  

Limitations of AI and ML

Despite their advantages, AI and ML present several limitations:

• Training Data Quality: The accuracy of ML models heavily depends on the quality and diversity of training data. Biased or insufficient datasets can lead to unreliable predictions.
• Generalization to Unseen Geometries: Models may not perform well when applied to geometries outside the scope of the training data, especially if there are significant topological changes.
• Initial Investment: Building comprehensive datasets and developing robust models requires significant time, expertise, and resources, which may pose a challenge for smaller organizations.
• Tool Limitations: In the case of PhysicsAI, predictions are reliable only within the dimensional variations of the training dataset. Major changes to topology may fall outside the model’s predictive capability.

These constraints highlight the importance of ongoing research, data quality management, and thoughtful model development to fully leverage AI in simulation-driven design.

Conclusion

A machine learning model was successfully developed and validated using Altair PhysicsAI, leading to significant reduction in the time and resource requirements for explicit dynamic bumper impact analysis. This case study demonstrates the transformative potential of AI and ML in engineering simulations, enabling faster design exploration, reducing computational costs. And improving workflow efficiency in traditionally simulation environments.

Future Work

Future efforts will focus on expanding the dataset to encompass a broader range of impact scenarios and bumper geometries, thereby enhancing the model's generalization capabilities. The applicability of this  methodology will also explored across other explicit dynamic analysis domain. Additional work will aim to improve model accuracy and robustness, ensuring reliable performance under increasingly complex and varied conditions.

References

About the Authors

Ayyappa Gadhi
Cyient

Ayyappa Gadhi received his Bachelor's degree in Mechanical Engineering from Amrita Vishwa Vidyapeetham and has hands-on experience in the Finite Element Analysis (FEA) domain, in both static and dynamic analysis. He is proficient in leveraging AI tools in Computer Aided Engineering (CAE), notably utilizing PhysicsAI to simulate FEA analyses from the early stages of design. Currently working at Cyient, he supports Pratt & Whitney in advanced engineering initiatives. Prior to this, he played a pivotal role in developing machine learning models using PhysicsAI frameworks, contributing to their successful presentation at prominent industry events such as the GCC (Nov 2024) and Altair ATC (Apr 2025). His interests lie in integrating AI & ML with traditional engineering workflows to enhance simulation accuracy and efficiency across CAE applications.

Raviraj Shrivastava
Cyient

Raviraj Shrivastava is a Mechanical Engineer with a Master’s degree from BITS Pilani and has hands-on experience in the FEA domain, specializing in structural and modal analysis. He is currently working as a CAE Engineer at Cyient, where he supports John Deere and previously contributed to SME Automation for Pratt & Whitney engines. His expertise includes parametric modeling, DOE setup, fatigue analysis, and design automation using tools like ANSYS, HyperMesh. Raviraj is passionate about applying simulation-driven design approaches to enhance structural reliability in the automotive and aerospace sectors.

Sathish Kumar Garala
Cyient

Sathish Kumar Garala received Master of Technology in Design Engineering from IIT Hyderabad and has over 10+ years of experience across IT & Engineering industries. He predominantly works in Crash/Impact Analysis using CAE tools to support the customers of Aero and Automotive domains. He carried out and led projects on functional evaluations of aero engine structures like bird strikes on Spinner cones, Fan blade-off events, blade out containments in compressor/turbine sections etc. On Automotive discipline, he extensively worked on sub-system and full vehicle crash evaluations for frontal and side impact load cases for passenger cars and vans. He also takes responsibility as one of the SMEs in LS-Dyna for crash studies at Cyient. His interest lies in exploring new designs, materials and technologies that can help in building much safer aero structures and automobiles.

Leela Kishore Haresamudra
Cyient

Leela Kishore Haresamudra holds an M.Tech degree from Bangalore University and brings over 19 years of experience in the Aerospace and Defense sector. He has extensive expertise in the design, analysis, and simulation of complex engineering systems. His core areas of specialization include Stress Analysis and Engineering Vibrations, with advanced proficiency in Low-Cycle Fatigue (LCF), High-Cycle Fatigue (HCF), and multidisciplinary optimization techniques. Leela Kishore began his career at ADA-DRDO, where he contributed to the LCA Tejas project, supporting both military and naval variants. At Cyient, he has played a key role in supporting major aero engine programs for clients such as PWA, PWC, and IHI. As a Subject Matter Expert (SME) within the Service Area, he collaborates across teams at Cyient to explore automation opportunities and develop innovative digital platforms. Currently, he performs multiple roles including Delivery Management, SME, and Innovation Champion for both the Service Line and Business Units.

Lakshman Kasina
Cyient

Lakshman Kasina, an M.Tech graduate from IIT Madras, brings over 22 years of extensive experience in design, analysis, and simulation across various engineering domains. His expertise spans Stress Analysis, Impact Dynamics, Engineering Vibrations, and Composite Modeling. He is proficient in Low-Cycle Fatigue (LCF) and High-Cycle Fatigue (HCF), as well as advanced simulation techniques like multidisciplinary optimization. Lakshman has published over 10 papers in national and international conferences, demonstrating his significant contributions to the field.

About Cyient

Cyient (Estd: 1991, NSE: CYIENT) delivers intelligent engineering solutions across products, plants, and networks for over 300 global customers, including 30% of the top 100 global innovators. As a company, Cyient is committed to designing a culturally inclusive, socially responsible, and environmentally sustainable tomorrow together with our stakeholders.

For more information, please visit www.cyient.com

EU AI Act Implementation Timeline

The implementation of the EU AI Act follows a phased timeline to ensure stakeholders have sufficient time to adapt and comply. Below is a general overview based on the regulation’s staged rollout:

Use of AI in Healthcare

AI is rapidly transforming the healthcare industry by enhancing diagnostic accuracy, optimizing treatment plans, streamlining workflows, and improving patient outcomes. Here are some of the major use cases of AI in healthcare, categorized by domain:

Possible Challenges that Deployers/Manufacturers/Providers of AI for Medical Purpose Might Face

Integrating AI into medical devices presents complex compliance challenges, particularly due to the dual regulatory landscape governed by the EU AI Act-2024/1689, and the EU MDR-2017/745 or EU IVDR-2017/746. The intersection of AI functionality and medical safety introduces both technical and procedural hurdles.

Conformity Assessment Strategy for AI-enabled Medical Device (AIMD)

When an AI system is integrated into a medical device or constitutes a standalone AI-enabled medical device, the conformity assessment process is not handled separately under the EU AI Act. Instead, it is embedded within the existing regulatory pathway defined by the EU Medical Device Regulation (MDR 2017/745) or In Vitro Diagnostic Regulation (IVDR 2017/746).

Compliance Checklist for Deployers, Manufacturers, and Providers of General-Purpose, and High-Risk AI systems

About the Authors

Sathish Kumar Thiagarajan is a seasoned Controls & Automation Engineer with over 18 years of global experience in managing large-scale industrial automation projects involving PLCs, SCADA, and Drives. He specializes in optimizing technical workflows, ensuring regulatory compliance, and leading cross-functional teams to deliver seamless IT/OT integration solutions. Known for enhancing operational efficiency and driving cost-effective innovations, his expertise helps shape transformative strategies in industrial automation.

Srinivasu Parupalli is an experienced Systems Engineer with expertise in program management and delivery across multiple domains, including Industry 4.0, Manufacturing, Embedded Systems, IoT, Software Applications Development, and Cloud Integrations. He has extensive experience in end-to-end product development and has been instrumental in building and training teams on emerging technologies such as Ignition, Solumina, Aveva, and SCADA systems for deployment in diverse customer projects. With a strong background in industrial automation, he has worked across various industries, including Manufacturing, Energy, Utilities, Healthcare, and Process Automation, developing MES, SCADA, and HMI solutions integrated with other applications. His expertise lies in customer engagement, requirements analysis, and risk management, ensuring the successful execution of complex automation projects.

About the Author

Abhishek Kumar
Subject Matter Expert in Medical Device Regulatory and Quality Assurance

Abhishek Kumar is a Subject Matter Expert in Medical Device Regulatory and Quality Assurance with over 14 years of experience. He has led the EU MDR 2017/745 sustenance program, managed multiple global engagements for top medical device companies, and supported the gap assessment, remediation, and submission of 70+ technical documents across EU MDR, ASEAN MDD, NMPA (China), Taiwan, and 10+ 510(k) submissions. He has authored 40+ Clinical Evaluation Reports (CERs) for Class I–III devices in line with MEDDEV 2.7.1 Rev-4 and developed proposals for market access in the U.S., Europe, and APAC (including ASEAN, China, Taiwan, and Japan). He also prepared and implemented regulatory plans for new product development across 90+ countries through feasibility analysis and cross-functional coordination.

About Cyient

Cyient (Estd: 1991, NSE: CYIENT) delivers intelligent engineering solutions across products, plants, and networks for over 300 global customers, including 30% of the top 100 global innovators. As a company, Cyient is committed to designing a culturally inclusive, socially responsible, and environmentally sustainable tomorrow together with our stakeholders.

For more information, please visit www.cyient.com