Agenda

Application of Machine learning in Aerospace Structure Engineering and Its Future Directions

Presented By
Sharad Anand Shivaprasad (Collins Aerospace)

Authored By
Sharad Anand Shivaprasad (Collins Aerospace)
Sushree Kshirabdhi Tanaya (Collins Aerospace)

This paper explores the application of AI in Aerospace Engineering ,Engineering, focusing on the challenges of analytical derivation, making it cost-effective and less time-consuming, especially during the preliminary design . It covers the novel idea of developing neural network-based algorithms that impact the multi-disciplinary area of aerospace engineering, including structural analysis and fluid dynamics. We review the state of the art, gather the advantages and challenges of AI/ML methods across different aerospace disciplines, and provide our view on future opportunities. The basic concepts and the most relevant strategies for AI/ML are presented together with the most pertinent applications in aerospace engineering, revealing that AI/ML is improving aircraft performance, and these techniques will have a significant impact shortly. As some design parameters cannot be analytically predicted swiftly due to the complex and diverse failure modes involved, which can also be affected by manufacturing processes, it becomes necessary to always depend on tests and simulations for every laminate configuration that may have to be considered for optimization of the performance for a given set of loading and boundary conditions.In order to overcome this inability, a AI-based approach that relies on a Neural Network based Artificial Intelligence(AI) algorithm is proposed for empirically predicting thickness distribution(any design parameter, thickness is considered as example design parameter in this paper) across the structure which will be trained on the database of pressure distribution (Loads) vs thickness distribution (Design Parameter) across defined structure for various prior successful designs. The network can be viewed as a variation of a Radial Basis Function Neural Network with a Gaussian Basis Function and a one-pass learning algorithm, making it highly efficient computationally. The Neural Network algorithm explored in the present study appears more robust in reliably predicting swiftly the approximate values of any design parameter across any component of the aircraft.

HIC Value prediction for Pedestrian Protection using Machine Learning

Presented By
Swapnil Nanaware (Stellantis)

Authored By
Swapnil Nanaware (Stellantis)
Sreenath Mallela (Stellantis) Darshan Pawargi (Stellantis) Ebrahim Saifee (Srellantis)

Predicting Head Injury Criterion (HIC) values during pedestrian protection assessments can significantly streamline CAE analysis. This paper explores various approaches, including machine learning models trained on prior simulation data, automated FEA setups, interpolation techniques for HIC mapping, material optimization, and simplified physics-based models for quick estimates. These methods enhance design efficiency while reducing analysis time. Key parameters affecting HIC values were identified and used to train machine learning models across different vehicle architectures and head form impactor locations. The best-fit algorithms were validated using Odyssee CAE software. The robustness of the ML model was ensured through diverse data set combinations. The model was then applied to predict HIC values using unknown data sets, highlighting effective application areas. The study emphasizes the importance of selecting appropriate training and prediction data for optimal model performance. The generated ML model can further assess parameter sensitivity and optimize design in minutes compared to the traditional CAE approach which take hours.

Chassis Subframe Part Organising using AI

Presented By
Mayank Kapoor (Mobis India Limited)

Authored By
Mayank Kapoor (Mobis India Limited)
Ashok Anumalla (Mobis India Limited)

The organization and naming of components in chassis subframes, often comprising over 20-25 parts, is a time-consuming and error-prone process when handled manually. Each part must adhere to a specific naming format incorporating details such as part name, material type, and thickness. A machine learning (ML)-based approach offers a promising solution to automate this process, leveraging algorithms to classify and organize components efficiently. This proof-of-concept (POC) algorithm for MacPherson-type subframes utilizes Python and TCL scripts for automation. The PythonOCC module is employed to read and interpret the geometry of the components. The extracted geometric data is then broken down into a set of features, which serve as input for ML models to efficiently classify the components. This automated system significantly reduces the time and effort required for organizing subframe components, while eliminating human error. It ensures consistent classification and naming without relying on specialized engineering expertise. With advanced scripts and ML algorithms, the workflow becomes both faster and more accurate, enhancing productivity and reliability. In addition to improving subframe component organization, this approach demonstrates the broader potential of automation in complex assemblies, paving the way for its application across various engineering domains. With its focus on efficiency, precision, and accessibility, this innovation redefines the standards of component management.

Accelerating Head Impact Simulations with AI-Driven Surrogate Modeling for Pedestrian Safety Optimization

Presented By
Tejas Ramesh Chavan (Altair)

Authored By
Tejas Ramesh Chavan (Altair)
B Veda Vamsi Krishna (Altair) Sohan Sundaresh Rao (Altair) Hari Krishna Reddy M (Altair)

Head impact simulations are computationally intensive and represent a major bottleneck in the iterative design and optimization of automotive hoods for pedestrian safety. To address this challenge, a data-driven surrogate modeling approach based on Graph Convolutional Networks (GCNs) was used to significantly accelerate the simulation process. High-fidelity finite element simulation data from multiple head impact locations was used to train the GCN-based surrogate model. The model was designed to predict key outputs such as displacements and accelerations curves at various impact points. From the predicted acceleration curves, the Head Injury Criterion (HIC), a critical metric in pedestrian safety was computed. The trained surrogate model achieved a prediction accuracy of greater than 90%. Once trained, the model was able to generate predictions in seconds, compared to 4 hours typically required by traditional solvers. Structural Optimization was performed with the objective of minimizing HIC values by exploring design variables such as sheet thickness and reinforcement geometry. By replacing the traditional solver with the surrogate model in the optimization loop, a substantial reduction in computational cost and design cycle time was achieved. This AI-augmented workflow facilitates rapid design space exploration and supports the development of safer, pedestrian-friendly vehicle front-end structures, demonstrating the transformative potential of surrogate modeling in simulation-driven design.

Development of an AI model of Airbag Simulation Using Image Processing

Authored & Presented By
Arjun Dev (MOBIS INDIA LIMITED)

This paper presents an AI-driven approach for airbag deployment using image processing along with algorithm. Collecting airbag images and then image data is processed to extract key features, which are then fed into a machine learning model trained. Rapid training and validation of models using MLP(multilayer perceptron) with radial basis functions , Kriging, and neural networks, allowing for quick and accurate prediction. Simulation output data from LS-DYNA is used to generate training datasets. The integration of image-based feature extraction method and integration of AI/ML algorithm with LS-DYNA simulation results in a hybrid system that significantly reduces computational time while maintaining accuracy.

Accelerating Battery Pack Thermal Simulations using Physics-Informed Neural Networks

Presented By
Prashant Kumar Srivastava (oorja)

Authored By
Prashant Kumar Srivastava (oorja)
Vineet Dravid (Oorja Energy)

Modeling the thermal behavior of a battery pack is essential for assessing its health and longevity. In electric vehicles, the battery pack is the core energy source. As current flows through the pack, ohmic heating occurs due to the internal resistance of the cells and the contact resistance between the cells and connectors. This heat raises the temperature of both the cells and the connectors. Additionally, the battery pack contains highly conductive thermal pads that transfer the heat generated by the cells to the pack’s enclosure. Heat is then dissipated through either forced or natural convection to maintain the battery pack’s optimal operating temperature. Traditional methods like finite element and finite volume analysis have been effective for modeling heat transfer in battery packs. However, these systems involve numerous parameters that influence the governing equation, such as pack current, cell thermal conductivity, internal resistance, connector thermal conductivity, and the convection heat transfer coefficient on the enclosure. Simulating different combinations of these parameters using conventional methods can be computationally expensive and time-consuming. Additionally, meshing a battery pack is a labor-intensive and tedious process, especially due to the presence of varying length scales, which significantly adds to the poor quality of the mesh. In this work, we employ Physics-Informed Neural Networks (PINNs) to model heat transfer in battery packs, effectively incorporating the key parameters that govern the system. Specifically, we use a Fourier Neural Network (FNN) architecture to capture the high-frequency solutions of the heat diffusion equation for materials with properties that differ by up to two orders of magnitude. This approach enables efficient modeling across varying material properties.

Machine Learning-Enhanced Reduced Order Modeling for BIW Design Optimization: A Study in Frequency Response Reduction

Presented By
Vidit Sharma (ESTECO India Software PVT. LTD)

Authored By
Vidit Sharma (ESTECO India Software PVT. LTD)
Anuj Shukla (ESTECO India Software PVT. LTD) Eshan Amalnerkar (ESTECO India Software PVT. LTD)

This study introduces a computationally efficient Reduced Order Modeling (ROM) approach for optimizing the BIW of a car, with the objective of reducing maximum steering wheel velocities in frequency response simulations. High-fidelity finite element simulations are often computationally expensive, particularly when exploring multiple design variables. To mitigate this, Proper Orthogonal Decomposition with Interpolation (PODI) was utilized to reduce the dimensionality of simulation data obtained from NASTRAN solver, where frequency response analysis was performed for varying shell thicknesses of 16 key BIW components. A ROM was constructed enabling accurate predictions of maximum steering wheel velocities across different design configurations. ROM allowed for the efficient handling of reduced data, ensuring that the model retained critical physical attributes while significantly lowering computational costs. Furthermore, machine learning techniques, particularly Radial Basis Function (RBF) networks, were integrated with the ROM to accelerate performance predictions across the design space. In addition, a Multi-Objective Robust Design Optimization (MORDO) framework was implemented to address uncertainties in design variables and operating conditions, ensuring that the optimized designs were both high-performing and resilient under real-world conditions. The ROM, enhanced with RBF networks, was coupled with HYBRID, a multi-strategy optimization algorithm, to efficiently minimize maximum steering velocities within the design space. The integration of ROM technology, RBF networks, and optimization strategies in modeFRONTIER resulted in significant reductions in peak steering wheel velocities. This approach highlights the effectiveness of combining ROM technology with machine learning and optimization for industrial design applications where computational efficiency and accuracy are critical.

GenAI Solutions for Synthesis of Body Structures

Authored & Presented By
Shubham Verma (Detroit Engineered Products (DEP))

This work explores the use of Conditional Generative Adversarial Networks (cGANs) for structural design generation, focusing on automotive Body-in-White (BIW) stick models. The goal is to generate feasible geometry representations conditioned on target mechanical properties—specifically bending and torsional stiffness. The BIW geometry is voxelized to transform the structural data into a grid-based format suitable for deep learning. The generator within the cGAN is trained to synthesize voxelized BIW geometries based on the provided stiffness parameters, while the discriminator evaluates their structural plausibility. This method enables rapid exploration of design alternatives that meet specified performance criteria, potentially accelerating the early stages of vehicle architecture development. The approach demonstrates the capability of cGANs to capture complex structural dependencies and encode functional relationships between geometry and performance. Future enhancements could include incorporating additional structural attributes, increasing voxel resolution for finer detail, or integrating performance validation feedback into the training loop for more accurate generation.

Reimagining Crash Simulations with AI: Transforming Legacy Data into Scalable Surrogate Models

Presented By
Celal Karadogan (DYNAmore – an ANSYS Company)

Authored By
Celal Karadogan (DYNAmore – an ANSYS Company)
Shyam Vasu (Ansys)

Artificial Intelligence (AI) and Machine Learning (ML) are increasingly transforming Computer-Aided Engineering (CAE) workflows by enabling faster design iterations and reducing computational costs. Ansys SimAI is a deep learning-based surrogate modelling platform designed to replicate complex physics-based simulations with significantly reduced runtimes. This paper presents the application of SimAI in modelling an automotive side impact scenario, using physics-based high-fidelity data generated from LS-DYNA simulations. The study begins with the development of an AI model trained on a dataset where the pole impact position and door beam configurations are systematically varied. The trained model demonstrates strong predictive capabilities, accurately capturing vehicle deformation patterns and curves of pole force progress across a range of input configurations. In a second case, a separate model is trained using simulations with fixed pole and door beam settings, but varying rocker panel reinforcements. Again, the AI model shows excellent generalization, accurately predicting results for unseen reinforcement geometries. To expand the model’s applicability, both datasets are combined to train a comprehensive surrogate capable of simultaneously accounting for variations in pole position, door beam configuration, and rocker reinforcement design. This integrated model successfully predicts the deformed shape and force-time histories for new design combinations within the training space, showcasing SimAI’s ability to generalize across a multi-dimensional design space. The results demonstrate that expensive and time-consuming LS-DYNA simulations can be leveraged effectively to build AI models that dramatically reduce the time required for design exploration. In addition, SimAI models are valuable tools for continuous engineering development because they can be trained further with new data, allowing for iterative improvement and adaptability. This approach not only accelerates product development cycles but also opens up opportunities for more comprehensive design optimization using legacy and newly generated CAE data.

A Novel Cost-Effective Brake Monitoring System Using Minimal Sensing and Data-Driven Modeling

Presented By
B Veda Vamsi Krishna (Altair)

Authored By
B Veda Vamsi Krishna (Altair)
Kamikkiya P (Altair) Hari Krishna Reddy M (Altair) Akash Yella (IIT Tirupati) Sriram Sundar (IIT Tirupati)

Real-time brake monitoring system is one of the important aspects in the context of vehicle safety and performance. Conventional brake monitoring systems depend on a multitude of sensors leading to high integration costs, increased system complexity and maintenance challenges especially in real-world operating conditions. This work addresses these challenges by taking a practical approach to brake monitoring, minimizing sensor dependency and utilizing power of machine learning. A controlled disc brake setup was developed to demonstrate the concept. Instead of relying on numerous sensors, the system recorded only the angular velocity of the brake in real time. The acquired real-time data was further processed to extract segments of velocity signals specifically associated with braking events automatically. From these segments, statistical features were computed, which exhibited a strong correlation with the target performance metrics. These features served as inputs to an artificial neural network (ANN) model, which was trained to estimate critical parameters such as contact forces and brake torque quantities that were typically difficult to measure under real-world operating conditions. A validated multibody dynamics (MBD) model was used for ANN training, achieving an accuracy of greater than 90%. The approach enables real-time brake monitoring that is scalable, cost-effective, and suitable for manufacturers, operators, and service providers, reducing the need for extensive hardware and maintenance.

Advanced Hybrid Algorithms for Predictive Maintenance and Remaining Useful Life Estimation of the Batteries

Presented By
Prashant Kumar Srivastava (oorja)

Authored By
Prashant Kumar Srivastava (oorja)
Vineet Dravid (Oorja Energy)

Battery health monitoring and prognostics have become essential in ensuring the safety, reliability, and longevity of energy storage systems, particularly in electric vehicles and renewable energy applications. This study presents a comprehensive framework for real-time battery health assessment and remaining useful life (RUL) prediction using a combination of data-driven and physics-based approaches. Key parameters such as power loss, state of health (SOH), internal resistance, and temperature are continuously monitored through Controller Area Network (CAN) data and diagnostic algorithms. Advanced machine learning models, including federated machine learning and regression methods, are integrated with electrochemical modeling to improve accuracy in predicting degradation trends and failure modes. The models have capability to auto-calibrate themselves with the availability of the new data. The proposed system not only enhances early fault detection but also supports predictive maintenance, reducing downtime and lifecycle costs. Experimental validation on lithium-ion battery datasets demonstrates high accuracy and robustness of the prognostic models, showcasing their potential for deployment in real-world applications.

ML-Driven Optimization of Battery Pack Performance during Concept Design

Presented By
Anit Jain (Dassault Systemes)

Authored By
Anit Jain (Dassault Systemes)
Jayant Pawar (Dassault Systemes) Girish Patil (Dassault Systemes) Swapnil Shete (Dassault Systemes)

Modern vehicles face a growing array of requirements driven by regulatory standards, consumer demands, and competitive pressures. Concept Structure Engineering (CSE), enabled by the 3DEXPERIENCE platform, facilitates the creation of numerous design alternatives for components and larger automotive systems, such as such as body and battery structures, which are subject to critical load cases, including safety tests. These load cases interact with objectives like cost and mass, complicating the sequential resolution of interconnected design challenges and hindering OEMs' accelerated time-to-market strategies. A promising approach to overcome these challenges involves transforming concept modeling into a comprehensive, vehicle-wide optimization task, which requires an evolving suite of technological advancements and iterative plan-do-study-act cycles. Key technologies for this vision include parametric CAD modeling, process automation, advanced finite element meshing, scenario-based simulation, high-performance computing integration, automated post-processing, and data analytics, seamlessly supported by the 3DEXPERIENCE platform. The integration of machine learning (ML) and artificial intelligence (AI) enhances these capabilities, enabling rapid simulation and predictive insights. A shared data model and embedded communication tools streamline process automation and improve team-based issue tracking. This paper presents initial minimum viable products (MVPs) using CSE on the 3DEXPERIENCE platform with implicit associative modeling, allowing for consistent assembly generation of complex systems like Battery Pack. Leveraging visual scripting and parametric modeling, specific cases such as optimizing Battery Pack under side pole impact are highlighted. ML models play a critical role in accelerating the assessment of design alternatives, identifying structural reinforcements that achieve an optimal balance between mass, cost, and regulatory compliance. Future developments will aim to extend these methods to multi-physics, multi-load case scenarios, offsetting the increased simulation demands with predictive efficiencies. By leveraging historical finite element models and machine learning models, OEMs can enhance adaptability, swiftly address evolving safety standards, and drive data-driven, efficient development cycles.

Predicting Washer Button Reliability using Finite Element Simulation

Presented By
Pramod Kanawade (Whirlpool)

Authored By
Pramod Kanawade (Whirlpool)
Sachin Nilawar (Whirlpool)

Whirlpool Corporation is a global leader in home appliances, with over 110 years of innovation. Known for its iconic brands like Whirlpool, KitchenAid, and Maytag, the company offers a wide range of products, including washing machines, refrigerators, dishwashers, and ovens. The reliability of user interface buttons on washing machines is paramount for user satisfaction and product longevity. Physical buttons are subjected to repeated mechanical stress, moisture exposure, and varying user interactions, making their reliability a critical design consideration. This abstract explores the application of simulation techniques to assess and enhance the reliability of washer buttons. By employing computational models, we can simulate various operational scenarios, including repeated button presses with varying force and frequency, exposure to humid environments, and potential impact events. Finite element analysis (FEA) can be utilized to analyze the mechanical stress and strain distribution within the button structure and its underlying mechanisms, predicting potential failure points under different loading conditions. Furthermore, agent-based modeling can simulate user interaction patterns and identify high-stress usage scenarios. Material properties, button geometry, and internal component design are key parameters incorporated into the simulation models. The outcomes of these simulations provide valuable insights into the expected lifespan of the buttons, identify design weaknesses, and enable engineers to optimize button design for improved durability and reliability, ultimately leading to enhanced product quality and reduced warranty claims. This simulation-driven approach offers a cost-effective and efficient method for ensuring the robustness of washer button interfaces.

Simulation Driven Tool for Predicting the Structural Integrity of Spin-on Filters

Authored & Presented By
Dinesh Lonkar (Atmus Filtration Technologies Inc.)

Atmus filtration Technologies is a global leader in filtering solutions for commercial vehicles, off-road vehicles, and equipment. Spin-on filters are one of the most extensively utilized products. Spin on filters are made up of mainly 5 to 6 parts and their design can vary depending on customer requirement and applications. Based on the need, different configurations of these filters are produced using appropriate part sizes and geometries. For each configuration, a laboratory test is required to determine its structural integrity. This task is time-consuming and creates a bottleneck in the time to market principle. To address this issue and provide quick solutions simulation-based tool was developed. Detailed simulations were performed for various configurations, designs, and test specification variables. Response curve optimization was developed and validated to ensure a good correlation between this tool and lab tests. This tool allows the product team to understand filter frequency, stresses in critical parts, ,and the design margin. The product team can easily determine the best configuration for spin on filters to meet the customer applications requirement. This tool is leveraged to reduce the time and cost involved with lab testing.

Simulation of Assembly stress in Tube

Authored & Presented By
Anjali Ramachandran (Caterpillar Inc)

Tube cracks & leaks caused by assembly stress pose significant challenges in industries and cause detrimental consequences in compromising the structural integrity, reduced performance and reduced quality. These stresses are often identified after the failure has occurred. This abstract focus on how simulation techniques can be leveraged to accurately analyze assembly stress, potential critical areas and suggest alternate assembly sequence. Simulating assembly stress can also aid the engineers in making informed decisions regarding necessary design modifications and procedural changes. Tubes and similar complex components with multiple non-planar mounting points can be deformed during assembly, causing residual stresses reducing fatigue life. If the assembly stress is above the yield strength of the material, it is unacceptable as it often relates to an uncontrolled assembly process, which does not allow for proper assembly. Manufacturing review should ensure that the part is being produced correctly and that there are no issues with the assembly process. To make this a pre corrective action, Finite element analysis methodology is adapter to simulate the stress and deformation occurring in tubes during the fitment process. First a verification is done to realize how changing assembly sequences and tolerance variation due to assembly affects the stress generated in tubes. Then a simulation study is conducted to analyze the stress values in tubes for an assembly sequence and identify critical points with high stresses. Further a study is conducted to optimize the assembly sequence which has reduced stress values compared to the baseline study. Not considering stresses generated during assembly can often cause failure and warranty claims to the enterprise thus affecting quality. Hence simulating these stresses in the design concept stage is necessary to ensure quality. Identifying an optimal assembly sequence with keeping assembly stresses in mind is important to arrest these issues beforehand.

FEA study on the effect of prying using contact elements and validation with classical hand calculation in the typical structural brackets

Presented By
Kuralanban Ramu (BIPL)

Authored By
Kuralanban Ramu (BIPL)
Ramkumar Selvarajan (BIPL) Mohanraju Sundarraju (BIPL)

Typical brackets are used in the structural members which are subjected to tension loading. An additional load is developed in the attachment due to the prying effect, which is above and beyond the applied tensile load in the attachments (tension clips, fittings, brackets, etc.), since eccentric load in the fitting produces heel-toe effect. Classical methods are available to calculate the effects of prying for simple geometries (such as L, T, Z shaped brackets / clips) subjected to tension loading. In this study, the 2-dimesional FEA model of typical sheet metal bracket subjected to tension loading are modelled with contact elements (linear and non-linear). These contact elements are modelled to simulate the prying effect in the bracket contacts. Different contact elements are studied to simulate the prying effects using simple bracket geometry and loading conditions. The obtained FEA results are compared with classical hand calculation to validate the simplified model. This simple model analysis approach can

Thermo-Mechanical Reliability Study Of BGA Packages Using Finite Element Analysis

Presented By
Manish Kumar (VALEO)

Authored By
Manish Kumar (VALEO)
Surendharan Nesan (Valeo)

Ball Grid Array (BGA) packages are integral to modern electronic assemblies, offering high-density interconnects and superior electrical performance. However, they are prone to failure under thermal stress, particularly during thermal shock events. Thermal shock failures in BGAs occur due to rapid temperature fluctuations, causing differential expansion between the BGA package, solder joints, and the PCB. This abstract explores the key failure parameters associated with BGAs during thermal shock, emphasizing factors such as thermal gradients, material properties, and mechanical stresses. This paper examines how temperature cycling induces thermal expansion mismatches between the BGA's materials, including the solder balls, PCB, and the die, leading to stresses that can cause solder joint fatigue, crack formation, and eventual failure. The study incorporates critical parameters such as the coefficient of thermal expansion (CTE), solder joint geometry, and die and package size on the lifecycle of BGA during the thermal shock event. This paper highlights the importance of understanding these parameters to improve BGA reliability and mitigate the risks of failure in mission-critical electronic systems. A printed circuit board having ball grid array packages soldered is used in the study. A analytical model is built in ANSYS Sherlock, which is later coupled with ANSYS Mechanical for further study. The findings provide insights for optimizing BGA designs, soldering processes, and packaging materials to enhance reliability and minimize failures due to thermal shock.

Thermal Buckling Prediction For Natural Fiber Reinforced Composites Using AIML

Presented By
Guru Anandan (Altair)

Authored By
Guru Anandan (Altair)
Vinoth Dhanavel (Altair)

Natural fibre-reinforced composites, such as jute fibre composites, are increasingly used in the automotive sector due to their lightweight nature, sustainability, and contribution to reduced carbon emissions. However, accurate evaluation of their structural stability under thermal loading conditions is crucial to ensure reliable service performance. Specifically, predicting the critical buckling temperature — the temperature at which the laminate becomes unstable — is vital for design and safety considerations. Thermal buckling behaviour of laminated composite is influenced by several factors, including ply orientations, stacking sequences, and the presence of cutouts (such as circular, rectangular, or slotted holes) at various positions within the plate. Traditional analysis methods are computationally intensive and time-consuming, especially for complex, non-standard configurations. This study aims to develop and deploy an Artificial Intelligence (AI) / Machine Learning (ML) model capable of predicting the critical buckling temperature of jute fiber-reinforced composite rectangular plates with arbitrary, non-parametric designs. The model will be trained on data generated from thermal buckling simulations of various laminate configurations and cutout geometries, enabling rapid, accurate prediction for novel designs without the need for extensive finite element simulations.

Study Of Thermo-Mechanical Behavior Of Overmolded Parts Under Sudden Temperature Fluctuations

Presented By
Manish Kumar (Valeo)

Authored By
Manish Kumar (Valeo)
Arunprasad Janakiraman (Valeo)

Overmolded polymer parts are widely used in engineering applications based on the application needs and multiple benefits. However, Certain environment conditions can lead to localized stress concentrations, delamination, and premature failure, particularly in regions influenced by fiber orientation and weld line formation during molding. This work presents a coupled simulation methodology to evaluate the thermo-mechanical response of overmolded components under exposure to thermal shock—sudden temperature variations. Fiber orientation tensors are extracted from mold flow simulations and incorporated into the finite element model using an orientation-dependent constitutive formulation based on the Mori–Tanaka micromechanics approach. Weld line regions are identified and major weld line regions are selected for the material degradation due to weld lines in the simulation. The simulation is performed for one thermal shock cycle. The stress results are exported to perform the fatigue calculation for the required number of cycles. The effect of boundary conditions , fiber orientation and weld line are studied .This approach offers a predictive tool for design validation and performance optimization of overmolded structures operating in thermally aggressive environments.

Thermal structural coupled analysis for two manufacturing processes used in semiconductor industry

Authored & Presented By
Soumik Ghatak (ANSYS Inc.)

Many manufacturing processes in semiconductor industry require coupled structural-thermal behavior. These kinds of processes need structural solver as well as conjugate heat transfer solver. In this presentation, simulation methodology for two manufacturing processes have been demonstrated using coupled structural thermal solver. One is laser cutting and another one is ultrasonic welding. Methodology for both analyses has been demonstrated for proof-of-concept model and suitable post processing items have been shown. Laser cutting is one of the major manufacturing processes used for wafer dicing. For laser cutting analysis, specific card for heat flux trajectory has been implemented and strongly coupled thermal structural problem has been solved with temperature dependent erosion of the material. Based on the simulation result, users can estimate how much heat flux is needed to cut the material and what would be profile of the surface after the process. Ultrasonic welding is another important welding method in semiconductor manufacturing industry. For this analysis, suitable thermal property of material has been used, and temperature dependent tying of contact has been explored. Based on the simulation result, users can estimate how much heat is generated and how much power is needed to join the material.

The Role of Virtual Testing in Passive Automotive Safety - Past, Present and Future

Presented By
Kirankumar Pedamallu (Arup India Pvt Ltd)

Authored By
Kirankumar Pedamallu (Arup India Pvt Ltd)
Maruthi Kotti (Arup India Pvt Ltd)

Virtual testing has emerged as a transformative tool in passive automotive safety, addressing the limitations of traditional physical crash tests while enhancing cost efficiency, scalability, and real-world scenario coverage. By leveraging computational models and finite element analysis, virtual testing enables the simulation of diverse crash conditions, occupant postures, and vehicle configurations without requiring extensive physical prototypes. The current roadmap emphasizes integrating virtual testing with physical validation through hybrid methodologies, particularly for regulatory compliance and NCAP assessments. Initiatives like the EU-funded OSCCAR project aim to standardize Human Body Models (HBMs) for evaluating diverse occupant demographics (e.g., varying sizes, genders, and postures) and emerging autonomous vehicle layouts. Euro NCAP’s 2025 protocol formalizes a five-step virtual testing process combining OEM simulations with physical verification tests, using ISO scores and injury risk KPIs for validation. HBMs now play a central role in addressing the biomechanical limitations of Anthropomorphic Test Devices (ATDs), offering detailed injury prediction for organs and skeletal structures. However, current HBMs face challenges in muscle reflex accuracy, neural delay variability, and injury risk function standardization. NCAP frameworks also lack universally accepted HBM assessment criteria, relying instead on ATD-derived metrics that may not fully capture human physiological responses Future advancements will focus on possible digital twin integration, improved HBM fidelity for pre-crash kinematics, and AI-driven scenario generation in combination with Active safety scenarios. The industry is moving toward "full virtual testing" paradigms that combine HBMs with validated vehicle environment models, particularly for evaluating advanced driver assistance systems and unconventional seating configurations. These developments aim to extend passive safety protections to vulnerable populations while maintaining regulatory rigor through hybrid physical-virtual validation frameworks. We will explore the current simulation tools and their ecosystem, underscoring the importance of evolving virtual assessment practices to align with global safety protocols and their seamless integration.

Transforming Surgical Stapler Development: From Physical Testing to Virtual Validation

Presented By
Sunil Sutar (Dassault Systemes Solutions Lab, Pune)

Authored By
Sunil Sutar (Dassault Systemes Solutions Lab
Pune)
Sachin Ural (Dassault Systemes Solutions Lab) Anand Pathak (Dassault Systemes Solutions Lab) Himanshu Gururani (Dassault Systemes Solutions Lab)

Minimally Invasive Surgery (MIS) has gained widespread adoption due to its advantages of smaller incisions, faster recovery times, and improved surgical precision. Surgical staplers, essential for wound closure in nearly 80% of procedures, are widely utilized but increasingly associated with malfunctions, and adverse patient outcomes. The rising incidence of stapler failures highlights the urgent need for improved testing – we demonstrate how virtual validation methods can be used to augment conventional physical tests while simultaneously providing greater reliability and efficiency. This study presents a simulation-driven framework to optimize and validate surgical stapler designs, specifically targeting common failure modes such as malformed staples and blood leakage. A unified modeling and simulation approach is implemented, featuring parametric exploration of stapler and anvil geometries. Tissue behavior is modeled using a Hyperfoam material model, while stapling dynamics, including complex tissue-stapler interactions, were captured through the Coupled Eulerian-Lagrangian (CEL) method. Eulerian Volume Fraction tracking and minimum principal stress analysis enabled detailed assessment of staple formation and identification of potential leakage pathways. The framework developed for stapler validation has been used in real-world settings to design staplers with substantially reduced risk profiles, and can easily be extended or customized for any surgical stapling application. This virtual validation methodology offers a scalable, robust solution for advancing surgical stapler development, enhancing device reliability, and accelerating innovation in medical device manufacturing.

A Digital Twin Framework with Machine Learning for Real-Time Life Prediction of PCBs Under Vibration Loading

Presented By
Sohan Rao (Altair Engineering India Pvt. Ltd.)

Authored By
Sohan Rao (Altair Engineering India Pvt. Ltd.)
Kamikkiya P (Altair Engineering India Pvt. Ltd.) Harikrishna Reddy (Altair Engineering India Pvt. Ltd.)

Printed Circuit Boards (PCBs) are typically designed with conservative assumptions to withstand worst-case mechanical loads. While this design strategy provides a safety margin, it often does not represent the range of dynamic conditions PCBs experience in real-world applications, particularly in the automotive, aerospace, and industrial sectors. In these environments, PCBs are regularly subjected to complex, time-varying vibration profiles that can induce early-life failures, especially due to solder joint fatigue and progressive structural degradation. Traditional simulation methods and design validation techniques are limited in their ability to predict such failures under actual operating conditions. This study introduces a digital twin framework aimed at improving failure prediction and life estimation for PCBs exposed to realistic mechanical environments. The framework integrates a high-fidelity finite element model, real-time sensor feedback, and supervised machine learning to create a robust, adaptive model of the PCB’s mechanical response. A detailed finite element model was developed to capture the layer-specific material properties and orthotropic behavior of the PCB. This model was calibrated and validated using experimental modal analysis to ensure accurate representation of the board’s dynamic characteristics. Simulations were conducted across a range of vibration scenarios to generate data on stress distribution, modal displacement, and fatigue-relevant features. These datasets were used to train a machine learning model capable of identifying key failure modes and estimating the remaining useful life of the board, with particular emphasis on solder fatigue. To validate the digital twin under operational conditions, PCBs were tested on a shaker table operating in the 100–1000 Hz range. Accelerometers placed at critical locations recorded vibration data in real time, which was streamed back into the digital twin environment for continuous model refinement and validation. The results demonstrate that combining physics-based simulation, real-time monitoring, and data-driven prediction enables more accurate, real-time health assessment of PCBs. This approach offers a practical path toward predictive maintenance, reliability testing, and lifecycle management in vibration-critical applications.

Leveraging Finite Element Simulation and Machine Learning Algorithm for Reaction Force Prediction of Solenoid Bracket Plate

Presented By
Mayur Sabale (Knorr-Bremse Technology Center India Pvt. Ltd.)

Authored By
Mayur Sabale (Knorr-Bremse Technology Center India Pvt. Ltd.)
Pradip Sontakke (Knorr-Bremse Technology Center India Pvt. Ltd.) Shashank Borde (Knorr-Bremse Technology Center India Pvt. Ltd.)

Over the years Finite Element Analysis (FEA) has become the most important part of design and development process of mechanical components. While FEA provide quick insights into the behavior of a component under given operating conditions, designer needs to do iterative refinement to achieve the expected outcome. This process becomes time consuming for complex components. To reduce the time required for development, we have used an approach where the data from FEA is used to train Machine Learning (ML) model and predict the output. This study examines the use case of solenoid bracket plate, which is used to secure solenoids in an assembly at a certain location using bolts. In the conventional method, FEA is used to determine the reaction force that the bracket plate claws exert on the solenoids as a result of bolt tightening. It is of utmost importance to achieve the specific value of a reaction force to avoid any leakages from the assembly. The new approach, identifies key bracket plate design parameters and generates a parametric FEA model. Important design parameters are used in a Design of Experiments (DOE), and a database of FEA results is created to train ML model. A trained ML model is used to predict reaction force at the claws. Lastly, predicted reaction forces are validated using FEA simulation for the set of design parameters.

Air Channel Warpage Optimization with Injection Molding Simulation

Presented By
Sreevin Satheesan (Whirlpool Corporation)

Authored By
Sreevin Satheesan (Whirlpool Corporation)
Harashal Sawant (Whirlpool Corporation)

This paper investigates the challenges associated with the injection molding of plastic components in home appliance manufacturing, with a particular focus on addressing energy inefficiencies in washer units. Conducted within the context of Whirlpool Corporation's ongoing efforts to enhance product performance and manufacturing efficiency, the study leverages advanced Computer-Aided Engineering (CAE) techniques—specifically Finite Element Method (FEM)-based injection molding simulations—to identify and resolve critical production issues. Air channel warpage, resulting from dimensional deviations during the molding process, is identified as a primary cause of increased energy consumption due to air leakage. The objectives of this work are to (i) identify the root causes of air channel warpage, (ii) develop and implement strategies to mitigate the warpage, and (iii) restore dimensional accuracy within specified tolerances. The study utilizes methodologies such as sample data analysis, correlation, and Design of Experiments (DOE) to optimize process parameters. The findings highlight that mitigating air channel warpage significantly improves energy efficiency, reduces scrap rates, and enhances resource utilization in production. This paper underscores the vital role of simulation-driven engineering solutions in optimizing manufacturing processes, advancing product quality, and promoting sustainable production practices in the home appliance industry.

Optimizing Elastomer Sealing for IP-Rated Enclosures: A Nonlinear Simulation Approach with Preload Bolted Systems

Presented By
Dore Prasad R (Tyco Electronics Corporation India (P) Ltd.)

Authored By
Dore Prasad R (Tyco Electronics Corporation India (P) Ltd.)
Sandeep Santhanagopal (Tyco Electronics Corporation India (P) Ltd.)

In Aerospace, Defense, and Marine (ADM) applications, achieving robust sealing performance is critical for maintaining Ingress Protection (IP) ratings against environmental stressors like dust and water. Traditional simulation methods rely on linear approximations and simplified contact mechanics, which fail to accurately capture the nonlinear, hyperelastic behavior of elastomeric gaskets under screw preload. While conventional FEA can predict general deformation trends, it often lacks the fidelity to model real-world sealing performance due to insufficient material characterization and oversimplified boundary conditions. To address these limitations, this paper presents an advanced simulation-driven methodology combining high-fidelity nonlinear FEA with hyperelastic material modeling. A key innovation is the development of a physics-based transfer function that correlates IP ratings with sealing pressure, friction coefficients, compression ratios, material properties, and preload forces. Experimental material characterization was performed to extract precise stress-strain data, which was curve-fitted to a Neo-Hookean model for accurate hyperelastic behavior. The FEA model incorporates large deformations, contact nonlinearities, and screw preload effects in ANSYS, enabling a system-level assessment of sealing integrity. A sensitivity analysis was conducted to identify critical design variables, while IP test conditions were translated into pressure boundary conditions to predict minimum sealing pressures for target IP ratings. The model was validated against physical tests, demonstrating >90% accuracy in predicting gasket compression and contact pressure distribution. Deviations (<10%) were attributed to surface imperfections and manufacturing tolerances, highlighting the model’s robustness. By adopting this methodology, design cycles can be reduced by ~40%, and physical prototyping costs can be cut by ~30%, while enabling rapid design iteration. This approach advances simulation technology by providing a predictive, physics-based framework for sealing performance optimization, moving beyond empirical trial-and-error methods. The results underscore the value of hyperelastic modeling and system-level FEA in achieving reliable IP-rated enclosures for demanding ADM environments.

Surrogate Based Multi-Objective Process Optimization of Laminated Composite Panels

Presented By
Soban Babu Beemaraj (Tata Consultancy Services Limited)

Authored By
Soban Babu Beemaraj (Tata Consultancy Services Limited)
Yagnik Kalaria (Tata Consultancy Services Limited) Amit Salvi (Tata Consultancy Services Limited)

The present study aims to minimize the process induced residual stress and the time at which degree of cure reaches 75% of flat laminated composite panel for a given batch time of 5400 and by changing process parameters. This is a multi-objective optimization problem. Increasing cure temperature causes faster degree of cure, more process induced residual stress, and inferior product quality. At lower cure temperature slower degree of cure, lower process induced residual stress, improved product quality, but increased process time and hence cost. So, need a compromise agreement between product quality and process cost. Classical genetic algorithm is interfaced with data driven surrogate model of the panel. The surrogate model is built from design of experiments with 500 latin hypercube samples. Global sensitivity analysis is performed on DOE data to identify the cure process parameters which influence the output response more. Then most influencing parameters are given with higher mutation percentage in Genetic Algorithm (GA). During optimization, developed GA does an intelligent process parameter search based on their fitness values and achieves faster convergence. Fitness values are linear combinations of objective functions and corresponding weight vectors. Pareto front is found by varying weight vectors. Obtained optimum values of reduced residual stress and degree of cure are reported. 64% reduction in residual stress is found while optimizing cure process. Robustness of the final performance is also quantified using local sensitivity analysis.

Robust Design Optimization of Turbine Discs

Authored & Presented By
Abhiram D R (Infosys)

This study presents the robust design optimization of the 2D cross-section of two-stage turbine discs for large civil gas turbine aero engines. Turbine discs, classified as critical parts, must satisfy various safety criteria defined by certification agencies like EASA and FAA due to the hazardous consequences of their failure. These discs endure high centrifugal and thermal loads, which are cyclical and lead to fatigue. In addition to integrity and fatigue requirements, turbine discs must meet functional requirements such as disc lean and rim-roll, limiting allowable deflection at the disc rim. System-level requirements, such as weight limits, must also be adhered to, making turbine disc design suitable for multi-objective optimization. This work conducts multi-objective optimization of the 2D cross-section of two-stage turbine discs. A systematic approach based on design-of-experiments (DoEs) is followed to identify critical design parameters affecting disc characteristics. This approach helps manage the number of design parameters in the final DoEs. DoEs also aid in generating surrogate models to predict disc characteristics, including integrity margins, stress (and fatigue life), functional behaviors (disc lean and rim roll), and weight, based on design parameters. These surrogate models, verified with a separate validation dataset, range from simple polynomials to complex models like radial and elliptical basis functions. The surrogate models facilitate multi-objective optimization to achieve an optimal 2D cross-section for the turbine discs, satisfying all criteria. The full paper will detail the optimal solution, benefits compared to a baseline design, selected design parameters, DoE construction, surrogate models, optimization objective function, and robustness assessments.

Optimization Automation for Subframe Design Improvement

Presented By
Mayank Kapoor (Mobis India Limited)

Authored By
Mayank Kapoor (Mobis India Limited)
Ashok Anumalla (Mobis India Limited)

Design improvement for chassis subframes is critical during the development process, particularly when failures occur in CAE simulations or actual fatigue testing. Traditional methods for enhancing structural performance relied on a hit-and-trial approach, which was time-intensive and often ineffective in achieving optimal results. To address this issue, a streamlined solution using Altair Optimization techniques was introduced. This approach involved targeting the localized regions experiencing fatigue failures, and analyzing them using static stresses obtained at the instance of damage. By improving the static strength of these local regions, the overall fatigue life of the subframe was significantly improved. However, the manual nature of this process, involving multiple sub-steps, led to lengthy optimization cycles of 2-3 weeks per design and was prone to errors from minor data inaccuracies. Automation was subsequently developed to minimize manual intervention and streamline the workflow. Using TCL scripting, the optimization setup was automated, enabling rapid improvement of subframe fatigue performance. This innovation reduced the optimization cycle to just 1-2 days, significantly enhancing efficiency while maintaining accuracy and reliability. This automated approach not only accelerates the design improvement process but also ensures consistent and robust results, offering a transformative advancement in subframe optimization practices. It demonstrates the potential of automation in addressing complex engineering challenges with precision and speed.

Process Development for reduced variation in suspension level performance using MORDO method

Presented By
Jugal Pathak (Mahindra &amp;amp; Mahindra)

Authored By
Jugal Pathak (Mahindra &
amp
amp
Mahindra)
Ganesh Lingadalu (Mahindra & Mahindra)

Competition in automotive industry is fierce and manufacturers must meet the needs of their costumers to maintain or increase their market. It is important for Original Equipment Manufacturers (OEMs) to create affordable vehicles with better driving experiences, comfort for occupants and safety for costumers. Suspension architecture plays critical role in vehicle ride, handling and stability behavior and is directly connected with driving experience. Incorrectly positioned hardpoints can significantly affect the performance of a vehicle. Consequently, recommended tolerances for suspension hardpoints positioning are made to avoid deterioration of Kinematics and Compliance (K&amp;C) parameters. Therefore, part level tolerance design is critical in vehicle development to minimize production cost and reduce vehicle performance deviation. The purpose of this paper is to develop a process for identifying robust design solution in terms of architectural hardpoints for MacPherson type front suspension without compromising K&amp;C parameters. Multi-objective Robust Design Optimization (MORDO) method has been adopted to assess the variation in performance parameters for given tolerances in suspension hardpoints. K&amp;C analysis is carried out using ADAMS/CAR software and Multi-Objective Robust Design optimization is executed using Mode Frontier. Keywords: Suspension part Tolerance, Kinematics and Compliance (K&amp;C), Multi-objective Robust Design Optimization (MORDO)

Multidisciplinary Design Optimization of a Typical Civil Aircraft Wing

Presented By
Bharath Kanaparthi (Dassault Systemes India Private Limited)

Authored By
Bharath Kanaparthi (Dassault Systemes India Private Limited)
Geetha Avula (Dassault Systemes India Private Limited) Dayanidhi Panda (Dassault Systemes India Private Limited) Emlin Vathsalan (Dassault Systemes India Private Limited)

Multidisciplinary Design Optimization (MDO) of an aircraft wing is very critical in the design stage of a civil aircraft to obtain enhanced performance with minimum possible structural weight. It involves interaction of aerodynamics and structures, wrapped around a process integration and optimization tool for exploring the optimal parameters and performance in a multidisciplinary environment. In practice, it is often observed that integration of multiple tools poses challenges that restricts engineers to explore the design space. This makes the process iterative, requiring manual intervention and therefore time consuming. The parametrized design variables considered in this study for the aerodynamic optimization of the wing planform are wing area, aspect ratio and taper ratio. The design variables for the structures are the thickness of spar, rib, stringers and skin. The specified constraints are the fuel volume dominated by the space between front and rear spar of the wing, the permissible stress and buckling factor. The location of the ribs and spars have also been parametrized. The aerodynamic loads for the cruise angle of attack are computed using CFD on the CAD models generated in the design space using CATIA Apps and transferred to the structural mesh model for further stress and buckling analyses on using SIMULIA Apps on the 3DEXPERIENCE platform. The optimization by design of experiments is carried out in the form of a nested approach i.e.: the structural optimization within an Aerodynamic optimization loop. The set objective has been to obtain optimal performance of a wing with a maximum Lift to Drag ratio with minimum structural weight subject to the fuel volume, stress and buckling constraints. The present abstract explains in detail the developed MDO system based on the seamless integration of CAD-CFD-FEA tools to obtain the optimal design parameters meeting all the necessary constraints within the 3DEXPERIENCE Platform.

Design And Analysis Workflows For Topologically Optimized Lattice Structures Generation Conforming Duty Cycle

Presented By
Dheeraj Kapse (CNH Industrial Technology Services (India) Private Limited)

Authored By
Dheeraj Kapse (CNH Industrial Technology Services (India) Private Limited)
Stefano Largo (CNH Industrial Italia SPA) Alessio Serracca (CNH Industrial Italia SPA)

With the growing demand for lightweight, high-performance components in industries such as aerospace, automotive, and biomedical, lattice structures have emerged as a viable solution, offering a superior strength-to-weight ratio. This paper presents 2 state-of-the-art end-to-end workflows for the generation of topologically optimized lattice structures that not only meet mechanical performance targets across duty cycles but also adhere to Additive Manufacturing (AM) constraints. First workflow is suitable in the scenario when part design is unavailable. It starts with defining a raw design space to generate an optimised lattice design confirming duty cycle and printing feasibility. This process has flexibility of utilizing NURBS (Non-Uniform Rational Basis Spline) curve for generating smooth geometry encompassing Lattice. Second workflow is suitable for converting an actual production part into a lattice part using Field Driven Design analogy. This process is suitable for parts where original production part is available and the user has lot of design constraints. The proposed workflows integrate advanced topology optimization algorithms with nature inspired lattice generation techniques. By coupling finite element analysis (FEA) with optimization routines, the structure is first optimized for stiffness, strength, and fatigue over the defined duty cycle. This optimized topology with added known Factor of Safety is then translated into manufacturable lattice geometries. Further, the workflow incorporates build orientation optimization and print simulation to validate manufacturability and minimize print-induced distortions. The results are validated through case study on polymer-based additive manufacturing (3D Printing) platforms. This contribution showcases a comprehensive and robust design-to-manufacturing pipeline that can be readily adapted across industries seeking to harness the full potential of topology-optimized lattice structures, ensuring performance fidelity and print reliability. Keywords: Lattice, Additive Manufacturing Simulation, 3D Printing, NURBS