Augmenting CAE in Automotive design with AI tools
Data Science, Engineering, Machine Learning, Optimization

Augmenting CAE in Automotive design with AI tools

In the past few decades, integrating power of computing into mechanical engineering practices was primarily a matter of generating input data from the computational study. Moreover, they prove useful in iterative improvement of the design, and its eventual realization. Computers augment humans in design and manufacturing and present the scope of automation wherever needed.

Finite Element Analysis(FEA) for Optimization

The development of finite element solutions for the explanation of functional engineering began with the advent of digital computers. The essence is that a set of governing algebraic equations is solved using the abilities of a computer. Hence, FEA is developing to a high degree of practical computation and application, which has significant importance in the engineering study.

In most engineering simulations, the aim is to arrive at optimal parameters to obtain the desired results – meaning, for example, to optimize weight, material, shape, and thickness of each component in the system is under consideration. Unlike most research papers that centres on limited dimensions (design variables), the reality is that the variables associated are in large numbers and hence:

  • Each simulation needs several hours and is expensive.
  • Manual intervention for updating design variables are labor intensive.
  • User need to learn to setup the optimization process and requires special training.

Optimization principles and techniques have been implemented in many fields to manage various practical problems. In light of progress in computing systems, optimization procedures have become more valuable and adopted in various engineering applications. FEA Optimization helps users to improve the design and the performance of the component while satisfying all the functional attributes concerning constraints and objective function.

Disadvantages of Conventional CAE Optimization

  • Can handle limited response variables and setting up dynamic load cases are challenging.
  • Qualitative analysis pertaining to the FE data are not captured.
  • Mistakes by users can remain undetected and debug takes more time.
  • The optimization process setup is manual and requires special training for users.
  • Manual intervention for updating design variables are labor intensive.

Intuceo-Ex CAE Meta Modelling (CMM) Optimization

CAE Meta-Modelling(CMM) is a workflow-based application that decreases the samples (Design of Experiments), develops ‘digital twins’ for simulation adopting surrogate models and uses various algorithms for generating the optimal solutions. Intuceo-Ex serves in understanding multidisciplinary attribute roles, finding correlations, performing impact analysis by executing the automatically-trained meta-models or carrying out full-scale design space exploration.

Advantages of Intuceo-Ex

  • Can handle multidisciplinary optimization (multiple response variables) and can be easily configured to manage dynamic load cases.
  • Qualitative analysis shows the correlations as well as the top sensitive parameters affecting the model.
  • Fine-tuned and accurate surrogate models with advanced machine learning algorithms.
  • The optimization process setup is automated and saves training time and cost.
  • Design variables are automatically updated by the Intuceo-Ex and requires no external intervention.

Business Benefits of Intuceo-Ex

  • Accepts direct FE model and needs no additional efforts.
  • Reduces cycle time by ~35-40% and cost by ~30%. For Eg: Concept development in Body in White(BIW) design phase cut down time by 30% employing Intuceo-Ex.
  • Reduced computational costs when compared with traditional FE/ optimization solvers.
  • The large quantity of simulation data is captured as a numerical format for future studies which is often lost due to the size.

Intuceo-Ex CAE Meta-Modeller with FE Solver

The Intuceo-Ex is a complete software package for structural optimization toward the CAE domain.

The structural optimization is extensively carried out during the early development phases of the component design. The application of structural optimization methodology in the preliminary design phases enables designers to combine various component properties and to achieve the optimal design configuration to meet the project requirements.

Using advanced automation, required variables from a FE model are extracted and using the inbuilt FE solver computed. The corresponding responses are collected, and the Response Surface Model (RSM) are built.

Response Surface Models (RSM) and optimization tools have continued to gain value. The improvements in the data science and the processing capabilities of the cloud have presented the opportunity to expand the design process by surrogate models for responses and then to optimize the designs utilizing these models. Wherever a substantial quantity of simulations are involved, using the first few sample design combinations and their responses, the surrogate models can be formulated. Moreover, these models help users by reducing the number of simulations, thus saving time.

Design of experiments (DoE) – is a division of applied statistics that deals with outlining, examining, and interpreting controlled experiments to assess the factors that regulate the value of a parameter or group of parameters. The generated samples are computed through conventional solvers along with their geometric parameters to get the responses for each design combination. Individual simulation in solvers is computationally intense and expensive. Advanced algorithms are used to identify accurate samples and also reduce the sample size.

Modelling – Several advanced algorithms and machine learning tools are presently available to formulate reasonably accurate surrogate models for linear or non-linear behavior of responses concerning design variables. Dimensionally reduced models some time present best results by decreasing the noise. Meta-model concepts eliminate designer bias, and model parameter tuning requires proficient skills in data science.

Optimization – It is a CPU intensive quick comparison of combinations of values that provide the nearest design parameters, meeting the objectives and constraints set up for study. Selection of algorithms is vital to assure that best designs are uncovered by traversing through all potential zones in the design space that converges faster to meet the objectives. Identifying sensitive parameters concerning the objective functions is an added advantage.

Unlike the output of solvers in traditional computer-aided engineering (CAE), which is crucial data, this data science-based procedure has an opportunity to collect the data of all the current and past analysis. Statistical analysis on this data in addition to insights from earlier models, will contribute useful information to designers. Augmented analytics now allows merging of human intelligence with machine learning to get most of the data generated in the process.

At Intuceo©, we have developed a meta-model based optimization application employing the right balance of all influencing factors, i.e., minimal DoE, fine-tuned surrogate modelling engine and advanced optimization logic.

Interested to see more engineering case studies? Click here.

You can contact us at or Request a Meeting here.