pSeven User Conference 2023 | Online

Commercial software industry revenues generated by the sale of performance simulation and analysis software (CAE) recently surpassed the amount of revenue generated from mechanical CAD software. This is a watershed moment for use of simulation software tools and methods in the industry. It has happened despite simulation still being largely used later in the lifecycle as a verification and validation tool with modest penetration in the early conceptual design phase of product development. Many studies have documented that simulation can have the greatest transformative impact on businesses if adopted early and used consistently throughout the product life cycle. Use of collaborative engineering platforms that enable a robust design and optimization approach levering best practices capture, process automation and Cloud-deployable, low code vertical applications can unlock the value of the intellectual property held in simulation models developed for verification and validation and make them available to all stakeholders in the product development lifecycle. These platforms also enable a single source of truth with proper management and authentication of simulation data, models and associated test results to be used by various “consumers” of simulation results across the enterprise (design, manufacturing, quality, in service) as well as with global supply chain partners.

System Engineers are more and more used to model-based design. However, optimization techniques remain not so common for them, due to historical reasons. In this project, we show how pSeven Enterprise unlocks the potential of optimization techniques for non-expert users.

A critical step in many model-based activity is the calibration of numerical parameters so that the model fidelity is acceptable. In such tuning process, a set of test data is the reference that the model should match as much as possible. This is typically an optimization task.

As a first step, the M&T team has captured the business rules in a collaborative workflow in pSeven Enterprise. After validation of the business part by the system engineer, we have derived an optimization problem with one objective and several constraints.

We then connected the business workflow to the optimization workflow, so that the non-expert system engineer is able to execute this state-of-the-art optimization process with new input data and make a trade-off analysis between several constraints.

The proposed paper focuses on an innovative solution to integrate machine learning and artificial intelligence capabilities into a digital twin strategy for monitoring a continuous production asset.

This innovative solution is the result of a collaboration between CGI and Datadvance within CGI's Global Innovation Centre (Toulouse, France), which aims to help industrial players around the world realise the benefits of Industry 4.0 and its enabling technologies, including the Internet of Things (IoT), augmented reality and digital twins.

In the concept of Industry 4.0 and cyber-physical systems, digital twins are based on predictive models and the valorization of operational data. They provide the key information to monitor and operate physical assets to maximise efficiency.

However, 'predictive models' can be of very different origin and quality, and there is usually no single model required to build the right digital twin. The challenge is then to build, combine and deploy all these models at scale into an appropriate hybrid Digital Twin.

The proposed process allows real-time interaction with a gas production unit. The aim of monitoring such a unit is to detect anomalies, optimize production, reduce energy consumption, and control maintenance.

The predictive model replaces the simulation model for rapid recommendations, typically for control optimization. It speeds up the generation of results when latency is critical, while maintaining sufficient accuracy.

We will show how the hybrid digital twin is assembled and how it interacts with the physical asset, from data collection and streaming with the IoT, to the feedback loop and operator support with the dashboard.

This project illustrates different scenarios ranging from simple asset monitoring to more complex asset tuning and optimization under changing operating conditions.

The Safran Group’s research & technology center SafranTech is pooling expertise and fostering the development of new technology. One of their activities focuses on gear heat treatment to improve surface resistance. The heat treatment involves some coefficients that are not directly measurable from experiments. In order to quantify those coefficients, pSeven and its optimization algorithms were used to perform a calibration campaign based on the experimental temperature evolution and coupled with Forge (Transvalor) simulations. The entire computation has been automated to give the coefficients in a limited time. The outputs show satisfactory results, and the process is now assessed in various parts.

In the context of the LuFo-MODUL project, the Technical University of Dresden focuses on developing novel materials for aircraft turbine applications. As some material properties are partially accessible to simulation, experimental tests are required. To alleviate the cost and time of the experiments, the pSeven Desktop tool aimed at predicting the optimal properties based on initial data. The actors were facing some limitations due to the high dimensionality of the problem while having a reduced budget and reduced initial information. The methods to cope with those limitations are presented in order to predict the new material properties to be tested.