How to automate tests on a physical model without real data

Click For Summary

Discussion Overview

The discussion centers on the challenges of automating tests for a physical model of an aircraft or car without access to real-world data. Participants explore methods for validating the model's outputs and ensuring they are reasonable and realistic.

Discussion Character

  • Exploratory, Technical explanation, Debate/contested

Main Points Raised

  • One participant questions the feasibility of automating checks on the model's results without real-life data, suggesting that without such data, the outputs cannot be assured to relate to actual physical behavior.
  • Another participant suggests performing a rough sanity check by calculating rates, accelerations, and "jerk" to see if they remain within reasonable limits, noting that specific limits depend on the type of aircraft and maneuvers.
  • A different participant proposes that if one models their own car, they should be able to validate it effectively, implying that self-created models may allow for better internal consistency checks.
  • There is a suggestion regarding programming languages, with one participant recommending object-oriented programming (OOP) languages for modeling, while also mentioning Python and MATLAB for handling complex mathematical formulas and numerical analysis.

Areas of Agreement / Disagreement

Participants express differing views on the possibility of automated testing without real data. Some argue it is not feasible, while others suggest alternative validation methods. The discussion remains unresolved regarding the best approach to ensure model accuracy.

Contextual Notes

Participants highlight the limitations of relying solely on input permutations for validation, as this may not provide assurance of the model's accuracy in relation to real-world behavior. There is also an acknowledgment that more detailed knowledge of specific parameters is necessary for accurate modeling.

DML
Messages
1
Reaction score
0
I want to code up a (physical) model of an aircraft or car, and I want to create an automated integrated test to check that the code produces reasonable and realistic results. For example, I want to check that the path taken by the aircraft or car is physical and reasonable.

However, I don't have real, physical data of an aircraft or car to test my model against.

Is it possible to check the results from my code in an automated way without having real-life data to compare it against? If not, is the only viable alternative to test hundreds of permutations of the inputs to the code and manually check the output to see if the results are reasonable? If so, what is the best way to minimise time spent on checking the errors?

If it's not possible to check the model in an automated way then how to people/companies check that their physical models are as correct as possible? What methods do they use to ensure the greatest accuracy possible?
 
Last edited:
Technology news on Phys.org
DML said:
Is it possible to check the results from my code in an automated way without having real-life data to compare it against?
No.

DML said:
If not, is the only viable alternative to test hundreds of permutations of the inputs to the code and manually check the output to see if the results are reasonable?
I don't see that that will give you any assurance at all: all it will tell you is that your model produces outputs which you think are reasonable, not that it produces outputs that bear any relation to a real aircraft or car.
 
Last edited:
For a very rough sanity check, you can calculate the rates, accelerations, and "jerk"s (the third derivative of position) and see if they stay within reasonable limits. Airplanes of certain types will have limits, depending on the types of maneuvers that they are performing. Also, remember that gravity and its direction play important roles in the limits. Anything more accurate would require more knowledge of the airplane parameters.
 
DML said:
I want to code up a (physical) model of an aircraft or car
Well if you model your own car, you should be able to validate it pretty well, no?
 
Depends what language to use, in general I would advice you to use OOP(object oriented) languages (since you will mostly work with object like models), or if you rather have complex or/and various mathematical formulas and this kind of stuff then you might try Python with some numerical analysis libraries... Maybe even matlab
 

Similar threads

What is the purpose of the OSI model and how does it partition the flow of data?
  • · Replies 1 ·
Replies
1
Views
2K
Misc. Digitizing a Boat Hull: Mac26 Modeling Data
  • · Replies 33 ·
2
Replies
33
Views
4K
What is the Best 2D Physics Engine for a Real-Time Simulation Program?
  • · Replies 1 ·
Replies
1
Views
2K
Undergrad Does the New Muon g-2 Measurement Indicate Physics Beyond the Standard Model?
  • · Replies 2 ·
Replies
2
Views
3K
Undergrad Are Real Numbers Essential in Scientific Measurements and Models?
  • · Replies 85 ·
3
Replies
85
Views
9K
AlphaEvolve - The Future of Math?
  • · Replies 5 ·
Replies
5
Views
1K
"Expected Result" Bias in Research
  • · Replies 5 ·
Replies
5
Views
2K
Graduate Modeling the damping of a physical rigid pendulum
  • · Replies 1 ·
Replies
1
Views
2K
FSK Demodulation: Testing Code Without Receiver/Radio
  • · Replies 3 ·
Replies
3
Views
2K
MATLAB MATLAB Code For Analyzing Experimental Data
  • · Replies 12 ·
Replies
12
Views
4K