Computer Modelling in 2D - Part II

 

Manuals

Back in 2002 I ordered the full set of silux manuals as shown below:—



The silux business model was to provide the program itself as a free download, together with some free documentation, and then to sell these manuals to serious users.

The User Guide has about 460 pages, the Script Language Guide about 160 pages, and the Upgrade Manuals for Versions 1.2 and 2.0. have 40 and 31 pages respectively. The Tutorial/Cookbook/Samples is just a copy of the three free pdf documents.

The whole lot cost me 180 Euros, with free delivery.

I now consider myself a proficient silux user, and these manuals were, and still are, very useful. Unfortunately, as far as I know, they are no longer available from silux ag in Switzerland. However, the most important documents of all, particularly for beginners, are the three pdf documents which are still available from the silux website. These are certainly enough to get started on some useful modelling.

Tips

Since I have learned most of these things "the hard way," the following list of workarounds etc for silux's various bugs and quirks may be of some interest to other users:—

Exploding models (rare but can happen): Try making all links and/or interaction constants harder. Or, make a small object of mass 1 gram or less (which I call a "timer"); place it where it won't hit anything else; make it a non-member for simulation (uncheck that in the edit object box) but still interacting. The aim is to reduce the time between iteration cycles, dt.

Also, if high-energy impacts take place over very short times, e.g. with "custom" interaction constants harder than "very hard", or with extremely strong compression springs, a timer will probably be needed to improve accuracy, (to any desired level, at the cost of longer simulation time — but the timer only needs to be set at low mass over the impact duration).

Also, avoid models with long flat surfaces bearing against each other, e.g. a long flat-sided piston in a cylinder. Keep bearing surfaces short.

Hollow objects: Use negative masses for cutouts. See how silux do this in their clockwork model. Otherwise rotational inertia will be wrong.

Micromovements: If the model is not starting from rest, assign correct velocity and rotational speed values for all objects, no matter how light, at the start of a simulation. Otherwise graphs may be "noisy". Occasionally, e.g. if accurate forces in links must be graphed, it may be necessary to calculate initial forces on some objects (e.g. centrifugal forces), and pre-load the model with these. So: calculate and assign them, (making sure that objects can only move in allowable directions for the pre-loading); run the model, periodically stopping all motion until it has settled down; then remove the forces. When restarted and running as intended, graphs should then be very smooth and accurate.

Yes, it may be tedious to do all this assigning of velocities, pre-loading etc, but on the other hand it can give a lot more confidence that the model really is set up and working correctly, and will give correct results.

Significant figures: Silux often displays only three significant figures e.g. for spring maximum force Fn. But for accurate work you can enter values to many more significant figures and silux will act on these, and will record and display to six significant figures in data associated with graphs.

Compression springs: These can display incorrect forces in graphs, but will still act correctly in models.

Dampers: Always graph damper force vs time to be sure a damper is working correctly. Either it will be, or if not, it will display zero force, and it will have to be re-made.

Gears: If gears are created using the silux method "Create Gear" and "Create Link Gearing" etc, this is the one case where the simulation results will probably be incorrect. Energy does not seem to transfer correctly across the gear link. If that's important, it would be best to model the gears as normal objects, complete with their teeth (only use simple arcs and/or straight lines — silux can't handle involutes!) The simulation will take longer, but it will be correct.

Saving: Never save a model with a graph minimised. You won't be able to recover the graph, and you may get an unhandled exception error which will crash the program.

Black screen: If the model window goes mostly black, e.g. after running a macro that has STOP SIMULATION, just click on that window.

Closing: To close (exit) the program easily, don't close the model window first.

Torque vs angle analysis

As a final tip, when analysing a proposed perpetual motion wheel, especially one that has multiple mechanisms of the same kind, I often find it useful to:—

1.  Model the wheel with just a single mechanism installed. Also add a single counterbalancing weight (if continuous balance of the mechanism's weight is important).

2.  Activate a macro to force the wheel to rotate at a constant rotational speed.
3.  Record wheel torque vs wheel angle over a full revolution (in silux, set up a graph of these quantities, and use the Record button).
4.  Transfer the recorded data into a drawing program, draw the graph of torque vs angle, and integrate it (i.e. find the area under the graph) to see if there is any net energy output.

This method can give a detailed and accurate result in less time than it would take to build a complex model with multiple mechanisms. Yes, such models can be built in silux using Script macros, but I still prefer the method described. One reason is that it's easy to see the magnitude of the torque, and exactly where it goes positive, and where it goes negative, etc, allowing the model's behaviour to be better understood.