Dynamic Simulation in Engineering: Techniques and Applications
Dynamic simulation is a fascinating field for engineers. It can yield an accurate representation of real-world systems, precisely as the non-expert public would expect expert engineers to produce.
However, accuracy comes at a cost, and dynamic simulation is not produced by pushing a single button!
Before embarking on dynamic simulations, companies should plan material costs such as in-house compute resource investment (proprietary hardware, i.e., compute farms) or cloud computing accessed as a rental or on a pay-per-use basis. Simulation engineers use their expertise (another factor to be considered!) and computational systems to associate inputs composed of 3-D geometrical data, such as CAD 3D shape representations, with other inputs, such as material properties and input forces.
The role of dynamic modelling and simulation software tools is to get valuable outputs to share with designers, directors, C-levels, customers, and suppliers. The software tool that associates outputs with inputs is CAE (Computer-Aided Engineering).
Dynamic Simulation Is Often Not Occurring Online
Dynamic simulation results aren't obtained in real time because of the sophisticated computations associated with CAE.
For example, a dynamic event occurring in ten seconds in the physical world, could be simulated in the virtual space with CAE after waiting ten hours or ten days of computations to be represented with engineering accuracy! This "delay" occurs because of several factors:
1) the lead time between the object shape (CAD) and its physical details (materials, external forces); this is the "preprocessing" phase
2) the computational time (to allow equations to be solved numerically, since there are just very few preset formulas for very simplified shapes of no industrial use); this is the "solver" phase
3) the last and most valuable phase, where most of the time is considered added value, is checking results. This is the "postprocessing" phase.
The first question is how software companies could flatten the preprocessing and solver lead times to zero to let engineers respond in real time. An answer is a data-driven simulation. More about that later on.
Advantages of Dynamic Simulation
Of course, there are tremendous advantages in dynamic simulation with respect to traditional laboratory testing, such as a full capability to explore design variables and the possibility to store and easily reconstruct the process when needed.
Also, dynamic simulation can be the basis of operator training systems. Simulation can be embedded in operator training simulators. Operator training simulators have the advantage, thanks to dynamic modelling, of providing final operators, such as in energy or chemical plants or mission-critical vehicles such as airplanes, with the possibility of testing usage without incurring fatal accidents because of initial lack of expertise.
This justifies the investment in HPC, cloud computing, software licenses, and specialized staff in major automotive or aerospace corporations.
Simulation outputs are precious actionable data that any engineer can understand. Examples of outputs are 3D color plots or global numbers expressing the KPIs of products. In any case, since part of the non-expert public using those actionable data could be made internal decision-makers or final customers, engineers need to set the right expectations before embarking on resource-intensive and time-consuming campaigns.
The Basics - Review
In this article, we'll review the process of creating dynamic simulation models and explore new solutions from Artificial Intelligence. Dynamic simulation involves developing mathematical equations that describe system behavior over time.
These equations consider the interactions of various components within the system and the influence of external factors. However, before delving into the realm of dynamic simulation, let's review the bedrock it stands on — statics and dynamics in physics. These concepts are fundamental to comprehending the behavior of objects in motion and their implications in engineering.
Any system evolving in time i.e., objects in space with time dependence, can be considered a dynamical system. Examples of dynamic systems in nature and technology are all around us. The stream of water in a river, a swinging pendulum, molten glass flowing in a furnace, the car we are driving, and the number of foxes in a forest each year, are forms of dynamical systems.
But in our context, we are particularly interested in the examples coming from fluid mechanics, chemistry, or mechanics.
Statics
Statics is concerned with objects at rest or in a state of equilibrium. Statics is primarily centered around the analysis of forces acting on stationary objects. We often employ statics to scrutinize and design structures that remain stable and apparently motionless, such as bridges, buildings, and other various static components, especially in civil engineering. The foundational equation governing statics is derived from Newton's first law of motion and can be expressed as follows:
Σ F = 0
where: "Σ F" represents the vector sum of all external forces acting on an object. The equation encapsulates Newton's first law of motion i.e. "An object at rest tends to stay at rest unless acted upon by an external force." This equation serves as a pivotal guideline for engineers in assessing the equilibrium of structures and ensuring their stability under the influence of external forces.
Statics and Dynamics
Dynamics, on the other hand, deals with objects in motion. This branch of physics explores the behavior of an object or system subjected to various forces and the resulting changes in its motion. Newton's second law of motion is the cornerstone equation and is expressed as
F = m a
where "F" represents the force applied to an object, "m" its mass, and "a" the resulting acceleration.
This law provides the foundation for understanding how forces affect the movement of objects. It forms the basis for modeling and predicting a dynamic system behavior starting from the knowledge of initial conditions, such as position and velocity, to accurately represent the position evolution over time. This equation relates the force acting on an object to its mass and acceleration.
Now, let's transition from the foundational concepts of physics to the dynamic simulation methods engineers employ today.
A Useful Compromise: Steady-State
Steady-state is a concept within dynamics, but it doesn't encompass all its aspects. Steady-state refers to a specific condition within dynamic systems where certain variables have reached a stable or unchanging state, even though there might still be dynamic processes.
Dynamics typically refers to the study of how systems change over time and includes a wide range of behaviors, including:
- acceleration
- deceleration
- oscillations
- transient responses to external forces or disturbances.
The concept encompasses non-steady-state (changing) and steady-state (unchanging) conditions.
Steady-state is a specific subset of dynamics where certain variables within a system have reached a stable, unchanging condition, often after a transient period of change. Also, there could be a specific time when rates of change become zero or periodic, but the system may still have ongoing dynamic processes.
So steady-state is a subset within a broader range of behaviors and conditions that include changing and unchanging states. Steady-state is an important concept within dynamics because it represents a condition where a system's behavior stabilizes, and scientists or product designers can focus on other topics such as spatial distribution.
While in unsteady situations a variable ϕ has a dependence on time ϕ=ϕ(t), in the steady state simulation models such a dependence does not exist. However, this does not mean that ϕ is a constant! The space distribution ϕ=ϕ(x,y,z) is still to be solved. The role of analyses and simulations is modeling via equations a system to get in response the pointwise value ϕ=ϕ(x,y,z), helping to determine the system performance. Thus, steady-state simulation models are a valuable modelling approach and can often yield faster responses than dynamic models.
A very popular steady-state example within the CFD (computational fluid dynamics) realm is the simulation of turbulent fluids via the so-called Reynolds Averaged Navier Stokes (RANS) approach. Whilst turbulence is per se a non-steady state, intermittent phenomenon, it can be captured via steady-state simulation that focuses on average values of quantities like velocity i.e. solving velocity u by not computing explicitly its fluctuation u' but only its average contribution: u = U + u'(t)
Thus, steady-state simulation can bypass excruciatingly complex issues in computational modelling. Steady-state simulation can be implemented in F1 car aerodynamics, stress analysis, and many other topics where it is preferred because while giving the same non-steady-state simulation, the computational time is less demanding and researchers and designers can focus on complex model issues related to the spatial distribution of temperature, pressure, or stress.
An example where the same object can be considered steady or unsteady state simulation is an automotive HVAC system for air conditioning. For instance, the air blower is rotating. However, for most purposes, its motion can be reframed within a reference system that is rotating. Therefore, the analyses are static. Similarly, there are opening and closing gates to mix air. However, most studies are concerned with fixed positions.
AI models (more specifically, the deep learning surrogate of CAE) focus on data rather than physics. AI exploits the data to create a data-driven prediction model that can operate in real time. Independently, if the simulation model is a steady-state simulation or a dynamic simulation, AI technology will yield engineering predictions in a matter of seconds instead of hours. This approach can help companies in product development and accelerate the decision time on more robust and eco-friendly products. The AI validation or testing phase occurs when a fraction of AI output predictions are compared to the CAE known outputs.
Example: Crash Simulation Case Study. CAE and AI
Crash simulations are vital in the automotive industry for evaluating vehicle safety. By modeling vehicle collisions engineers can assess how different car parts and safety features perform during accidents. This information is critical for designing vehicles that protect occupants and meet safety standards. Moreover, it helps reduce the need for physical crash tests, saving time and resources.
Chemical Engineering Dynamic Process Simulation (1D/3D)
In chemical engineering, dynamic process simulation plays a pivotal role in the design and optimization of chemical processes described with a dynamic model, i.e. variable in time. Using a dynamic model of pipes and vessels allows engineers to study how fluids and chemicals flow and react as they traverse a network of interconnected components in a plant. These simulations can be used to understand how changes in operating conditions, such as temperature, pressure, and flow rates, affect the overall process.
Dynamic simulations are invaluable in chemical plants, where the production of various chemicals, pharmaceuticals, and petrochemicals involves intricate reactions, separations, and heat exchanges. By simulating these processes, engineers can identify bottlenecks, optimize energy consumption, and enhance overall plant efficiency.
Stirred reactors are commonly used in the chemical industry for mixing and reacting various components. Dynamic simulations in 3D provide a detailed understanding of the fluid dynamics, heat transfer, and chemical reactions occurring within the reactor. This level of modeling is essential for optimizing reactor design, process control, and product quality.
Dynamic simulations of stirred reactors can be used in pharmaceutical manufacturing, where precise control over reaction conditions is critical to achieving high product purity and yield. By simulating the behavior of reactants, heat transfer, and mixing within a 3D space, engineers can make informed decisions to maximize production efficiency and minimize waste.
An example of a result that would be conceptually impossible with steady-state simulation is computing the time it takes for a dynamic mixer to achieve the desired components initially separated within a vessel. Once a steady state is reached either the simulation is stopped or more detailed 3D inquiries on the spatial distribution are carried out. To improve safety protocols and optimize semi-batch processes in chemical reactors, dynamic simulation is a crucial tool.
A Plot Is Worth 1,000 Words
The figure below is a graphic simplification with an XY plot that shows one of the many ways to differentiate between transient (time-dependent) models and steady-state simulation models. A scalar (such as pressure or temperature) or vector (such as velocity) variable φ is in theory a function of space and time φ=φ(x,y,z;t). However, right of the dashed line (assuming no new input intervenes to perturb the system), simulation experts focus on the spatial distribution (i.e. φ=φ(x,y,z)). This is per se a daunting task (e.g., finding point-by-point distribution of air velocity or pressure around a car) and caution should be used to understand what assumptions led to the steady state simulation, e.g., in the case of car aerodynamics:
- Is a RANS turbulent model enough to represent turbulence?
- Are physical phenomena unsteady, is flow oscillating, and can it be averaged to a steady solution?
- Are any inputs unsteady, are specific parts like a sunroof oscillating?
- Is there any delayed behavior that would introduce perturbations later on?
