Multiphysics software combines the different strands of engineering in to one simulation.
It might all come under one banner but computer-aided design is a field where separate tools are needed to simulate the many different engineering issues that can go into a single product. They can include issues as diverse as predictions of structural strength, electromagnetic behaviour and fluid flow. But the pressures of the market are bringing these tools together to promote better communication and interaction during the project lifecycle, all under the banner of multiphysics.
“We define multiphysics as anything in the physical world that is mechanical - has moving parts - and thermal properties when temperature changes are involved, as well as magnetics and electromagnetics,” says Subba Somanchi, director of model-based system design at Mentor Graphics. The company’s software is used for various applications in automotive, aerospace and biomedical industries. “But the electronics also interact with the physical world. And the software components that are in control make the model react and respond to the world. That is where it [multiphysics simulation] has the most benefit.”
Multiphysics is a field that needs integration, says Bjorn Sjodin, vice president of product management at Comsol: “People used to try and stitch these packages together using file imports but this is very cumbersome. It is still happening today and different software vendors are trying to make it smoother, but others are trying to combine these tools under one umbrella.”
The general rule is that the more multiphysics phenomena which can be included in any one simulation, the more accurate the final model will be, with the software allowing designers and engineers to test and verify virtual prototypes before moving to the physical construction stage which can help to minimise the number of errors in the final build.
The simplest multiphysics case could include thermal structure interaction where there is thermal stress on another material, but at the other end of the spectrum are in-fluid chemical flows, where heat transfer affects those chemical reactions, for example where three of four different phenomena are involved with multiple different equations required to simulate each.
“In a car you might have very complicated electronic and chemical reactions going on, where you have heat transfers that run the risk of the battery blowing up because people have not done the multiphysics analysis and did not understand the thermal properties involved,” Sjodin says. “It can be the same when working with laundry detergent where a truckload of chemicals are left in the sun to the point where they self-combust which partially destroys the detergent so that when it reaches the customer it isn’t any good.”
A specific example of multiphysics simulation for automotive use comes from Schrader Electronics, which employed the structural mechanics and CAD Import modules of Comsol Multiphysics to test a tyre-pressure monitoring sensor. It mounts inside the wheel to measure tyre pressure even when the car is in motion. Schrader engineers gradually introduced natural parameters into their structural simulation to see how the device would predict in different conditions - everything from the forces of rotating wheels to environmental stresses such as temperature change, shock and vibration, allowing them to test different geometries, materials and load scenarios.
For much larger systems, a project at the Technical University of Braunschweig is investigating new concepts in aircraft design to get better lift and focuses not just on aerodynamics but the elasticity of the wing structures and their acoustic properties.
Robotics provides another rich field for multiphysics as these so-called cyber-?physical systems combine electronic control with physical problems. Connections between robots and other systems over the network provide a further increase in complexity. Understanding the multi-domain problems that these robots present to designers and engineers is challenging, particularly when it comes to humanoid robots equipped with multiple systems to make them move combined with more complex electrical and mechanical elements that control the precise trajectory and position of the robotic arms.
Engineers and researchers at the Robotics and Mechatronics Center (RMC) of the German aerospace research centre DLR are currently developing robots capable of learning about and interacting with their environment. One is Agile Justin, a two-?armed humanoid specimen developed using Mathworks’ Matlab and Simulink which senses its surroundings using stereo cameras and RGB-D, torque and tactile sensors in its head, joints and fingers.
DLR RMC engineers found it was too complex to write separate control algorithms for the robot’s hands and arms manually in C/C++, especially when they started creating single control loops for multiple degrees of freedom from the device’s fingertip to shoulder. Rather they needed tools that would enable them to automatically generate code from design models and do the hardware in the loop (HIL) testing required for algorithms that compensate for sensor noise and precision, and verify the elasticities in the joints gears and non-linearities in the motor torque.
“You have to be able to represent the environment, so there is physical spatial data, and you have to have a good way to express that, capture and represent the environment in a convenient way for engineers, then decide what environmental features can be identified and interacted with,” says Mark Walker, senior engineer at Mathworks. “That is a very different type of problem compared to a big field of pixels that scans through very quickly.”
The biomedical industry too is an innovative user of internet connected devices for patient monitoring purposes. Arguably, the original wearable multiphysics biomedical device was the simple hearing aid but technology has moved on to create a diverse array of devices - everything from fitness tracking devices such as Fitbit and Nike’s Fuelband to insulin pumps and injection trackers for diabetes and even a sensor-based onesie, called Mimo, that tracks a baby’s health, sending updates to a smartphone.
“The biomedical industry is closely related [to wearable consumer electronics] particularly when you consider a pacemaker for example, which could also cook the body and damage tissue because of the heat generated by it [wireless transmission],” said Chris Wolfe, lead product manager for multiphysics at Ansys.
Comsol Multiphysics has been used in the design of a pacemaker electrode for modelling of ionic distribution of electrolytes in human tissue, for example, as well as systems that monitor the electrical activity in the heart. In other work, engineer Joel Gibbard’s Open Hand project formulated the idea of how an amputee can create their own prosthetic hand using a 3D printer. This was designed using National Instruments LabView software and data acquisition hardware. The system let Gibbard prototype different electromyogram (EMG) signal filtering techniques that measure the electrical activity of muscles at rest and during contraction to analyse how well and how fast the nerves can send electrical signals.
A simulation that combines both robotics and medical equipment comes from the French Atomic Energy and Alternative Energies Commission (CEA LIST), which used Comsol Multiphysics to build a predictive model of a phase change actuator for use in miniature surgical devices. The design relies on a microactuator based around a paraffin and a wax hydrocarbon that expands 10 to 20 per cent by volume when heated from a solid to a liquid, combined with carbon black particles that create a conductive composite to support Joule heating when an electric current is passed through it. As such, the simulation incorporated geometric, thermal mechanical and electrical parameters, explains CEA LIST researcher Panagiotis Lazarou.
“This is a multiphysics problem with nonlinear electrical connectivity, density, and specific heat capacity, and a changing viscosity, all of which affect the deflection [of the simulated membrane]. Moreover, resistivity increases as temperature rises, since the carbon particles spread apart when the paraffin expands,” Lazarou says. “We use Comsol as a prediction tool … to easily parameterise and change the actuator height, the membrane thickness and wax composite model.”
Lazarou is currently applying his simulation to the design and optimisation of an integrated miniature actuator which should be able to produce the high loads, range of movement and low electrical consumption required to meet medical requirements and keep production costs down, with a prototype due for completion by 2015 to be thoroughly tested before being integrated into a robotic surgical tool.
Accessible computing power
Increasing compute power has helped bring multiphysics to a wider range of engineers. And software vendors have become increasingly flexible in the way they package and sell their simulation tools, with individual modules designed for moderately powerful dual- or quad-core computers.
“You cannot buy a computer now that is not suited for simulation because the CPUs are so strong and the RAM is so cheap,” Sjodin says. “It is just a matter of how large and realistic you want the simulation to be.”
“More and more we are getting complicated setups where we need a big server farm to run a large number of very complex simulations,” Somanchi adds. “Some companies are already set up for that but in some cases we do help people run things in our [Mentor Graphics] infrastructure. And now that you can buy computer power from Amazon Web Services and others, our software is evolving to make use of that [public cloud] technology.”
Jobs can be initiated from HTML web-based portals for example, which offer some degree of customisation in terms of the how the application looks and feels, what elements it includes and how it can be integrated with existing web sites.
“A typical scenario would be where you have a research and development department with maybe a few people in that organisation serving a larger engineering and product manufacturing department who may not know our software or anybody else’s,” says Sjodin “We offer [the R&D experts] the opportunity to create an app that is so easy [for others] to use they just have to type in the numbers.”
These efforts will help extend the reach of multiphysics applications to a much wider engineering user base and allow more designers to come together.