In theory, this should give more or less the same results, except possibly on local phenomena. On the other hand, the beam model requires much less power and therefore a much lower calculation time.
In your case, there may be a check to be done because the delta is strangely close to a factor of 10. Wouldn't you have mixed kg and newtons for example. In any case, if it is normal not to find exactly the same result, the difference should not exceed a few percent. The difference therefore comes from elsewhere.
It is often thought that the volume model is more precise and therefore more accurate than a beam model. But this idea is not systematically correct depending on the models and solutions sought. If it is to calculate a set of members, a beam model is more than sufficient.
Where you have to be careful is if you mix the 2 types of modeling (e.g. volumetric fitting + beam bar). For example, the nodes of the volume elements do not take into account rotation (a face with at least 3 points, managing translations is enough). A beam connected to a volume node would therefore only be ball jointed and not recessed. Travel is then more important.
I noticed during the beam simulation that I also had solid parts, and as you said, mixed is not really suitable for it, and my differences probably come from that.
To be clear, I can stay all in volume (for the precision of the results), even if it is a mechanically welded element for example; provided I have sufficient computing power, am I right?
No, you have to mesh the beam profiles and the sheets in volume or even better in shell, but BE CAREFUL you have to create a relationship between the 2 types of mesh,
--> Contact between set between beams and contact faces
It seems to me that when you model a beam your software takes into account all the deformations that your stresses cause. Your deformations and resistances are therefore calculated inside your beam. The stones are therefore more numerous than for volume. Indeed, the volume uses a surface mesh. This explains why your results are so different. Basically, we only use the volume if we believe that the results will be faithful to reality... and about the same as the calculation with a beam. The calculations are therefore lightened and your computer will go faster. So you shouldn't make a mistake in your estimate.
From memory and if I don't make mistakes, it seems to me that the difference between a beam and a volume can be explained in this way.
So if I understood correctly, I have to keep the beams as beams (upn style, IPE, etc...)
The bars (square tube, rectangle, etc....) as a bar, and the sheet metal as a volume, in the simulations I can do, while putting connectors where I know there is contact between beam, bar and volume.
A priori, if the calculation time or the size of the model are not a problem, nothing prevents you from staying on volume. Generally speaking, the volume mesh could be the solution to all problems since it does not require any simplification, and therefore potentially no approximation. However, this does not mean that there are no rules to follow for a volume mesh. For example, to mesh a plate, you need to have at least 3 knots in the thickness. The problem is that obviously, it's very heavy. As soon as the models become a little more complex, it quickly becomes unmanageable for our machines.
In order to solve these size problems, simplifications in beam and shell have been added. This allows calculations to be made much faster, but they cannot be used in all cases. For a beam, the cross-section must be constant and its dimensions must be small in relation to the length. For a plate, the same kind of thing, the size of each element must be at least 4 times larger than the thickness. For complex shapes, there is no choice, you have to stay in volume.
From there, it's up to each person to choose what seems to suit them best.
After, for the details of connectors, etc not having SW, I can't help you much. On the other hand, maybe with views of the results it will be possible to find a way.
So, we made the sheet metal rack and we loaded it according to the solidworks simulation, we obtained a result close to the study all in volume (volume study displacement of 12.07 mm, in real 11 mm approximately, measurement with a rangefinder).
The conclusion seems to me to be ready-made. The volume simulation was the right one (as I said, it's the one with the least risk of error). It is normal not to find exactly the same result because it should not be forgotten that it is only a model. In reality, there may be slight differences in the properties of the material or the loading + the measurement uncertainties.
Obviously, this does not explain the difference with the beam model. Without seeing the model, it is difficult to say much more. I have already mentioned a number of factors to look out for. However, given the difference close to a factor of 10, this looks more like an input error (load in kg instead of Newtons, missing a 0 in Young's modulus, ...), which would give a displacement of 12.3 mm. Has the model been verified by someone else?