1. Conventionally the certification of a Unit involves testing of the UNIT, where all inputs and expected results for those inputs are specified in tests and results are used as part of cert evidence.
Correct. Performing HIL can give extra evidence for certification, but does not fully replace tests with 100% real setups.
2. In a HIL setting where plant is in Closed loop with the Unit under test, how do you verify the controller (unit under test performance), for instance for a PID, where control unit performance is also a function of the plant model? Are there measures required to ensure the plant model is representative of real world?
Validating plant models can be quite an issue. However, most commercial off-the-shelf simulation software solutions provide with proven models whose behaviour has been compared with real components. It’s true that when wrong results are observed, they could in theory come either from the controller under test (which is anyway what we need to track down) or from the test bench itself (hardware and/or software). The HIL test bench is supposed to have been validated prior to its interaction with the real controller, buy anyway when some undesired behaviour is observed in the overall closed-loop system, further investigation allows to discriminate the origin.
3. How do you validate your sensor models?
Usually commercial off-the-shelf simulation software solutions provide with proven models whose behaviour has been compared with real components
4. Validation would need to be performed during a physical test in my opinion. That may be during a physical test of the sensor during environmental testing and/or vehicle level testing involving additional instrumentation that is independent of the sensor. Hope that helps :-)
Right, but all tests can’t be done with physical prototypes, either because of the cost, delay or simply technical feasibility. HIL is not supposed to replace physical tests, but to complement them, in order to ensure a better test coverage and more flexibility in the validation process.
5. Hello sir, I'm curious about how can we implement HIL for take-off simulation of a hybrid UAV (VTOL + Fixed wing), could you please share some insight on this?
HIL makes sense when it comes to test a real controller, before the first flight. As long as the tests/simulations remain 100% virtual (no real hardware such a controller is involved), then offline simulation tools are sufficient. Take-off simulation involves multiple control systems (flight control, avionics, motor control, etc), so it depends on which subsystem to need to focus.
6. What can we learn from Silicon Valley software companies who are able to move in a very agile manner, and what are the limitations of this for aerospace applications?
Well, Agile methods start to be considered in aerospace industry, even though they are still marginal. The advantage of Agile methods is a higher flexibility in the design and implementation stages, with the possibility to adapt the final product almost on-the-fly (no joke here). This is required in some industries where development cycles must be very short, with many further software updates (such as Apps, Games, etc). Aerospace industry is less impacted by these short development delays (you can easily count on 10 to 15-year long development cycles), but safety is key. This is why we spend a lot of time defining in detail all modules/components and their interactions in order to reduce the risk of facing a catastrophic event.
7. Do you have to certify your test bench/simulator if you are using the evidence as part of your certification process?
It’s not necessary. What needs to be certified is whatever component that has to fly.
8. What kind of automation techniques are used in generation of Testcases in HIL Testing?
There are multiple techniques, depending on the complexity of the system (and especially the amount of I/Os interfaces). There more I/Os, the more combinations to test. I’ve already seen projects where all combinations of inputs where tested, even though some of them were either redundant or not meaningful. Besides that, there are specialized tools such as MaTeLo (https://matelo.all4tec.com/) which automatically generate test cases based on a use-case approach and an objective of test coverage.
9. What advice would you give to someone trying to set-up HIL testing from scratch? What are the main barriers you experienced when developing these capabilities? are you finding HIL is as beneficial as it's advertised?
I’d say : take your time. There are a lot of expectations around HIL approach, but behind every well-performing system, there is a significant dose of effort. Some time and energy are required to tame such complex systems, but they remain tools. The effect you get depends on how well you master them.
HIL is definitely beneficial and it’s proven that it has significantly improved product’s quality while reducing delays, risks and costs. But there is no magic : hard work and expertise are still required.
10. What do you suggest to the students in Aerospace regarding HIL simulation? What are the things we need to know before learning HIL simulation? and What are the Career opportunities?
You need to understand the system you are studying and define what you expect. An HIL test bench is nothing more than a computer with I/O and some wiring. It will be effective if it’s properly prepared/configured/operated based on the system under test’s specifications. Most industries involving control development (aerospace, automotive, railway, naval, energy conversion systems, power systems, electro-mechanical systems, robotics, …) seek engineers and technicians to work with simulation-based test benches at positions such as modelling specialist or test engineer, among others.
11. Aside from the complexity that Chris brought up, are the costs drastically different using HIL for non-mechanical systems such as battery modules versus mechanical?
Sure, as long as it involves control systems. Battery Management Systems is a trendy application, where batteries are emulated (down to the cells). HIL makes sense in this case because faulty conditions can be easily emulated (cell balancing, wrong cell wiring, overheating,…)
12. Will autonomous or remotely piloted aircraft testing change existing testing methods? How will this technology affect the use of HIL simulation?
In principle I can’t see a significant change, except the complexity of the control systems, which increases drastically, as well as the volume of data to be taken into account (radars, lidars, video streams…). Simulation is being used more and more to bring additional evidence for the certification of autonomous vehicles.
13. How helpful it is to understand the basics of HIL?
It helps to understand alternative approaches to design and validate complex control systems.
14. Dear Soheil and Speakers, thank you very much for this very interesting talk and exchange. A short question regarding the point you mentioned about building a resilient system and taking into consideration the "unexpected behaviours". In terms of control, it is the FTC (Fault Tolerant Control) strategy; Did you ever consider the ATC (Attack Tolerant Control) or cyber-security into the HIL process? thanks in advance for your answers.
I can observe an progression in the development of Cyber-security applications, where the resilience and recovery of critical systems must be assessed. Because it relies a lot on the control strategies, HIL is an approach that makes sense.
15. There are few faults in Aircraft which surprises even the manufacturers because the aircraft has flow for quite long time, can HIL be useful to detect those faults for systems in Use?
Definitely. The essence of HIL is to perform tests whose conditions are hard to reproduce in physical prototypes. But you still need to know what you are looking for.
16. Does Amesim software work with Opal-RT via co-simulation environment?
Yes. AMESim is capable to generate C-code that can be embedded, via Simulink, inside OPAL-RT Simulators.
17. What is different technology in HIL platform simulator? Which is preferred system with single or different technology depending on applications.
Most commercial test benches are based on 2 technologies : DSP/CPU and FPGA. DSP/CPU are used for massive / parallel computation with higher time steps (in the range of ~10 µs to 1 ms), to compute complex systems including plant models and controls. FPGAs are used for I/O management and faster simulations of smaller systems, such as electrical converters with high switching frequencies, which usually require time steps below 1 µs.
18. How long do you do MIL and SIL before doing HIL?
It depends on the project and how much the Model-based design approach is anchored in the Organisation. I’ve seen companies that don’t do MIL & SIL, but only HIL. Some others spend a lot of time on MIL (design) and SIL (pre-implementation), and less on HIL because they end up doing more physical tests.
19. How good is HIL for rocket design simulations?
Unfortunately I don’t have much experience on this particular application, but here are some relevant papers: