We Need a Simulation
Trends in design and manufacturing are driving engineers of all disciplines to a more integrated environment. We talk about this trend using terms like “Digital Twin” and “Virtual Factory”. The reality is that solutions in product design, factory design, and manufacturing execution are more tightly coupled than ever before. It is not a great surprise that this leads to confusion when engineers from different disciplines use the same word to mean different things. The term “simulation” is a great example of this issue.
To avoid confusion, engineers need to focus on the question that needs to be answered and be specific about the simulation required. If we do not know why we are building a simulation, we will never know when we are done.
The more we realize that requests can be misinterpreted, the less likely it will happen.
So, what types of questions need to be answered? Let’s look at a few different categories of simulation and the types of questions each helps to answer so we can understand the variety better.
When discussing the design of a product, we may want to know is it strong enough? Can it be lighter? Will a different material improve performance? Can the product be more aerodynamic? To answer these types of questions, we need tools to perform “finite element analysis” and “computational fluid dynamics”. In each case, engineers will use shorthand and say, “we need a simulation."
Finite element analysis involves modeling the surfaces, materials and forces that act on a specific part or assembly to understand how they will deform. This highlights the weak points of a part and helps engineers understand and improve their design. It can even suggest where material should be added for strength or removed to reduce weight.
Example Part – Finite Element Analysis
Computational fluid dynamics requires the same detailed look at the surfaces of parts and assemblies but focuses on the flow of a fluid (usually air or water) over the surface to measure drag, turbulence, and even the noise generated by the flow of the fluids. These analyses are becoming more important in electric vehicle designs that need to reduce drag to improve battery life and control frequency and pitch from other parts of the car that are much more apparent without a combustion engine to drown them out.
Example – Computational Fluid Dynamics
Equipment Design and Execution
Once engineers have a product, they need to figure out how to make it. This introduces a new class of simulations and a completely different set of engineers who address these challenges. If the product is a fender for a new automobile design, the engineers will need to design a process that creates these parts efficiently. Questions that arise include, What robot and weld gun should I use for welding this part? Where should the robots be placed? Can I meet the cycle time expectations? Does my robot program work correctly?
Equipment simulations like these rely on predefined commercial libraries that know the speed and limitations of the equipment. When working with robotics, the software provides tools to check for collisions as a robot moves and they even generate the programs that will run the robots in the real world. These models become the focal point of system design reviews. Engineers can see how the robot approaches each weld and make changes to work in concert with custom tooling that is clamping the parts or even change the position of the robot to make it safer for workers that are loading parts into the welding fixture.
Equipment Design “Simulation” Example
For both cases discussed so far, the simulations are deterministic. This means that each time the simulation runs with the same inputs, the exact same results are generated. These models avoid the introduction of randomness and focus on the detailed interactions to make complex numerical calculations to arrive at a specific solution.
Discrete Event (Factory) Simulations
There is a whole other class of simulations that look at stochastic events. These models are used to look at larger complex systems to understand their behavior when the concept of randomness is incorporated into the problem. On a generic level, these types of models can be used to assess traffic patterns, weather, or even stock prices based on historical data. In terms of a digital twin in manufacturing, typical questions include, do I have enough inventory? How many forklifts do I need in the shipping department? Am I going to make my customer’s required delivery date? Am I spending my capital effectively?
These models will rely on historical data to estimate when random events like a machine breakdown will stop a manufacturing system or they can be used to model the variation in the time required by a manual task such as assembling two parts. When considering an entire manufacturing system, there are hundreds of random (stochastic) processes that are all interacting to represent the overall system.
When machines break, buffers either fill or empty. When an operator takes too long with an assembly, the next operator in the process may need to wait for colleague to complete their task. These tools take an industrial engineering approach to use probability and confidence intervals to estimate throughput and draw conclusions about the most efficient ways to address the randomness we see in everyday manufacturing systems. These models will be run for long periods of time with different streams of random numbers to make sure that there is a statistical confidence that differences between the system design ideas are real and not due to random chance.
Discrete Event - Automated Warehouse “Simulation”
In addition to these broad categories, people will sometimes refer to physical tests (such as a wind tunnel) as a simulation. They will also “simulate” situations using Excel spreadsheets at varying levels of complexity. Bottom line, “Simulation” can mean so many different things. Context is required. As our solutions become more integrated, it is important that we take the time to define what we need and understand the questions so we can provide effective answers.