ATCx AI for Engineers

Siloes should be a thing of the past. Teams of designers, engineers, production technicians need to meet multiple objectives and have their products with customers before their competitors. Analyzing processes and product performance does not just require the expertise of data scientists but needs input from the engineering teams. Technology solutions from Altair enable all types of stakeholders to improve their work by utilizing AI. This presentation outlines the challenges industrial and manufacturing firms currently face and how AI based solutions can enable engineers to resolve them.

Participate in our live Q&A session  with Michael Larner, to get answers to all your pressing questions around AI.

The vision of Altair RapidMiner is to enable everyone to leverage machine learning and artificial intelligence to solve industry challenges and use cases and to create value from their data. In order to achieve this, the ease-of-use of the software, not only for data scientist, but also for non-data-scientists like engineers and other domain experts, as well as upskilling of engineers and other domain experts in machine learning and data analysis and their application to industry tasks is critical. This presentation provides a quick overview of how the Altair RapidMiner software platforms makes the use of data analysis and machine learning as easy as possible by using a unifying graphical user interface (GUI), an assistant for automated machine learning (AutoML / AutoModel), a visual data analysis workflow designer, a built-in recommendation system for the next best steps in your data analysis, and use-case-specific application templates. The presentation also describes the RapidMiner Academy and Altair’s Center of Excellence (CoE) approach for upskilling engineers in artificial intelligence and machine learning and their industry applications.

With the emergence of ChatGPT we have reached another pivotal moment with AI. At the very least, it has unlocked a next-level fever of interest in AI and potentially the ability to disrupt whole industries. This has arguably most organizational leaders wondering, how might this impact my business. In this panel discussion, we will explore the potential impact on manufacturers.

As CAE engineers, we generate so much data, most of the time we cannot even store all of them. Despite this, we are not living the successes of other industries such as retail, finance for machine learning and artificial intelligence. There are several reasons for our lagging behind in ML/AI. ​

We work with 3D geometric data that is hard to translate into a format that ML can work with. We most often need to predict large physics fields. Sometimes we need to generate new designs so we can combine with predictive models to find optimum designs. Most importantly we need to have access to all these within the products we are already using to simulate and optimize designs.  If ML processes are not part of our processes, the complexity of the process overrides, and it becomes the biggest of our challenges. ​

This is where Altair shapeAI and physicsAI come to your rescue, allow you; engineers be engineers. shapeAI is Altair’s geometry recognition engine and physicsAI is Altair’s fast physics prediction engine. Both are embedded into HyperWorks and enables you to speed up your CAE modelling process as well as design exploration. In this presentation, you will see what they do in details, how they work in HyperWorks and the value they add to your CAE processes and their outcomes.​

Accelerate your design cycles using state of the art geometric deep learning, available to you directly in the modelling environment. PhysicsAI harvests the power of your CAE data by learning relationships between geometrical shape and full contour results, letting you make lightning-fast design iterations and arrive at better designs, faster. Unlike traditional machine learning, physicsAI learns directly from the geometry without the need for parameterization. This means you can learn from any of your previous simulations without the need for design of experiments. This presentation is an introduction to physicsAI in HyperMesh, showcasing example problems and offers a sneak peak of upcoming features.

Artificial intelligence and machine learning is making its way in all the technology fields, helping reduce product development time and cost, and providing solutions which were not possible with traditional approaches. In the automotive industry, simulations led design has impacted new product development by leveraging the technology in stage gate process. To further reduce development time, integrating machine learning and artificial intelligence to frontload CAE simulations in the concept phase, is the new focus area of simulation groups. Few use cases involving machine learning algorithm have shown promising results underlining the idea that complex physics-based computations can be solved alternatively with data-based models.

Currently the technology is in its nascent stage though having potential to disrupt the simulation field by reducing the number of design iterations as well as the prediction time by a significant margin. This would further reduce pressure on the demand for compute intensive hardware for increased computational requirements. 

Machine learning is actively applied not only in computer engineering but also in the CAD/CAE field. The determination of design direction in the early stage of development relied on the experience of mechanical engineers, but these days, when data-driven decision-making is required, a highly reliable performance prediction system is required from the early stage of development. Therefore, this presentation describes the process of developing a performance verification system using machine learning based on CAE data. The verification algorithm was developed by collecting and time-series data by dividing the product into structural performance prediction and development stage performance prediction. Machine learning requires large amounts of data. However, since CAE requires a lot of resources to produce data, it is difficult to secure a large amount of data. To overcome this, CAE result data of existing projects were used for learning.

Improve your design by exploring the design space smartly and efficiently. Learn patterns, quickly predict performance, and prescribe optimum designs. Altair HyperStudy© multi-disciplinary software enables designers and engineers with automatic processes combining state-of-the-art mathematical methods and data-mining to make better decisions and increase productivity. This presentation is showcasing various application examples.

The knowledge gained from the evaluation of simulation results often results more from the understanding of a component’s behavior than from a scalar target value to be achieved.
So how can we optimize a component to a desired behavioral pattern? How do you evaluate thousands of simulations in the context of an optimization study?  This presentation will show an approach how to evaluate and optimize even complex behavior like a vehicle crash with the help of pattern recognition.

The Age of Digital Twins has arrived and is here to stay. Even though digital twin technology isn’t necessarily new, its adoption is sweeping regions and industries at astonishing rates. Organizations are rushing to adopt digital twin, learning how they can use it for different applications and purposes, and foresee even more growth in the coming few years. ​

From optimizing production operations, to monitoring supply chain, Digital Twins are playing a pivotal role across the Enterprise. Data, AI, Physics, IoT & Infrastructure are the key ingredients that go into the making of a Digital Twin. But making all these key ingredients work harmoniously with each other and exchanging insights still remains one of the greatest challenges. Another key aspect is the collaboration between people and the departments involved in the making and operation of a Digital Twin. ​

This track will focus on how to bring people, technology and processes together to create a “frictionless” environment for creating and leveraging Digital Twins. We will explore various technology components and listen to industry leaders who have successfully laid down foundations for scaling up Digital Twins in their organizations. ​

When we talk about digital evolution, smart manufacturing, industry 4.0 or Internet-of Things (IoT) the common factor is data. Whether coming from simulations or sensors, today the challenge is not about having data available; it’s more about efficiently and effectively using it. This is a well-known problem among companies trying to distill the needed information from their large data pools. For this purpose, in the heart of any application around data, there are three main steps to be tackled: data preparation, data processing and Result visualization. Data processing is often the phase where artificial intelligence together with the traditional equation-based approach, integrated in a digital twin, help to make the smart use of the data. In this phase, the prepared data are processed with different algorithms to fulfill various tasks: descriptive, diagnostic, predictive or prescriptive. Altair on the one hand is democratizing the needed technology, on the other hand acts as true partner helping and supporting the digital transformation guiding companies in their digital journey.

Efficient mixing of reclaimed asphalt product (RAP) is a crucial part of the process of making hot mix asphalt more sustainable but the traditional optimization approach, which is heavily reliant on physical trial-and-error, is prohibitively time consuming and expensive. Optimization using simulation provides a significant advantage. This work demonstrates an efficient methodology for virtual design optimization that combines high-fidelity physics-based simulation, High Performance Computing (HPC), machine learning and optimization to rapidly identify the globally optimal equipment design.

The methodology consists of parametrizing the equipment geometry, automatically generating and running Discrete Element Method (DEM) simulations for a well distributed quasi-random sample of the geometric parameter space on the cloud and fitting a response surface to the results using machine learning. A Multi-Objective Genetic Algorithm (MOGA) is then utilized to rapidly estimate the globally optimal parameter set from the fitted response surface.  Optima from the MOGA were verified and a new mixer configuration was identified which provides similar mixing with fewer paddles, lower torque, and greater mass flow representing a superior design.

This presentation explores the emergent field of Digital Engineering and its potential applications in the Aerospace industry, particularly through the concept of Digital Twins. An extensive overview of three types of Digital Twins – Data Driven, Physics Based Data Driven, and Hybrid Data Driven – is provided, with an emphasis on their respective data types, sources, machine learning techniques, applications, and implementation times.

The first type, Data Driven, is grounded in real-world data collected via onboard sensors and is ideal for repetitive operations, such as commercial aircraft predictive maintenance. The second type, Physics Based Data Driven, leverages synthetic physics-based simulation data, offering promising prospects for both current and future aircraft operations. The third type, Hybrid Data Driven, combines both real-world and synthetic data, reducing the need for physical sensors and enabling optimized design definition. The presentation concludes by recommending the use of Digital Engineering and Virtual Environments for generating Physics-Based synthetic training datasets for machine learning and AI applications in aerospace. The benefits of such an approach are explored, including the ability to predict future strategies for aircraft sustainment, autonomous flight, future mission planning, and enhanced aircraft designs.

The definition of digital twin, since it was first coined in 2010 originated from NASA to improve physical-model simulation of spacecraft, has been subject to many interpretations at various levels. At Altair, we have first-hand experience implementing Digital Twin across the system life cycle. In this presentation, we will discuss how Altair uniquely enables an open architecture based, vendor agnostic Digital Twin practice integrating multiple data streams across the enterprise from requirements to in-service data connecting physics-based models with data assimilated from inspection and sensors.

“Democratization” and “AI” are two words that are heard more often nowadays in the context of making use of AI as a commodity rather than a specialty. The true power of technology can only be leveraged when it sees widespread adoption across the enterprise. In other words, democratizing technology is the most significant factor for its success and its ability to deliver ROI. ​

AI democratization is largely dependent on transformation of “people” and “processes”, rather than just technology. The true transformation of manufacturing can be achieved through empowering engineers and technicians with AI and making them part of the transformation story. In the last few years, we have seen organizations upskilling their engineers with Data Science & AI techniques, helping them deliver remarkable solutions for improving yield, optimizing operations and lead time reduction. ​

This track will focus on the best practices for achieving AI democratization and the key aspects of people upskilling. We will hear stories of democratization of AI from organizations who have managed to transform a large segment of their workforce to create “AI Powered” Engineers.

With the rise of Artificial Intelligence (AI), every industry is experiencing a transformation. Manufacturers are struggling with vast amounts of complex data that require specialized domain expertise to extract valuable insights. Data Scientists, in high demand yet rare, typically come from a Computer Science background, lacking the deep mechanical expertise necessary for solving manufacturing challenges. As a result, significant time is often spent on bridging the gap between business requirements and data understanding. Conversely, engineers and manufacturing experts lack the necessary Data Science knowledge to effectively handle the ever-increasing volumes of data.

This presentation provides strategies for fostering a data-driven decision-making culture across the enterprise, focusing on innovative Data Science platform, modern enablement techniques, and tailored value-based use-case selection methodologies. By democratizing data science through enabling engineers, organizations can empower their workforce to harness the power of data and drive meaningful insights. Through this discussion, we will explore various approaches and best practices to enable enterprises to fully leverage their data assets and make informed decisions by utilizing their engineers and fostering a of collaborative culture with existing Data Science experts.

Join Karan Bedi, Digital Transformation and Technology Leader at Avery Dennison as he shares how the global company, with $8.4B in global sales, addressed their talent shortages in data science by focusing on targeted upskilling of existing employees and detail how this led to a more energized workforce. He will also detail how they facilitated the knowledge transfer from process experts, which is typically hard to distill information, using the Centre of Excellence Framework and how this meant they could drive new uses into production and increase revenue.

Aerospace companies collect endless amounts of time series data from sensors, systems, and simulations. Engineers can find practical insights in this data with machine learning and AI projects when they know where to look and what to do with that data. Predicting failure in engines, turbo fans, motors, and any moving parts or systems is impossible if there is no starting point in the data.

This presentation will discuss the challenges around sifting through available data and transforming it for use in machine learning projects, as well as demonstrate how engineers can complete a motor failure prediction project with raw data in Altair RapidMiner, a full-capability data platform in a no-code environment.