MEC-E5003 - Fluid Power Basics, 07.01.2019-14.03.2019

Comments on Simscape Fluids exercise

Teacher’s overall view is that the students could manage the use of Simulink/Simscape surprisingly well.

The exercise is still new and it will developed according to teacher’s experiences and students’ feedback, which was very welcome.

The ergonomics of Simscape is rather good. Even beginners learn to use it quickly. The challenges are in selection of system parameters. You should know what to component parameters and system inputs to give, otherwise you can get whatever results. The case is similar to usage of other simulators as well. You should know the technology you study and the system characteristics so precisely that you are able to criticize the numerical solutions and you can interpret the results correctly.  

Simscape’s challenge/problem is that the sub-models are quite “black boxes”. You no not really know what is happening inside the modules. Mathworks publishes the system equations used in the models but it is hard to know the way the numerical solution is made.

If you make the system model by yourself for example in Simulink environment you should know accurately how the system behaves and what phenomena your system covers … and what phenomena it does not describe.

Therefore also the calculations we have done during the Basics course are important. You should have the intuition and also view based on these basic calculations what kind of operation and numerical results should be expected from the system and its simulation model. You should compare your own, handmade results with the simulation output and be critical.

The position servo we studied works quite fine with only P control. The experiences could be different if we would had a velocity servo under investigation. The integrator which is already in the position servo system takes care that final error is small. However, there is small error left and we could not get rid of it completely. With higher P gain the error is reduced but that will eventually lead to system instability.

The stability problems appear (in this case) as a bounded vibration. In position signal this can be quite difficult to observe. However, if you study the velocity or acceleration outputs this oscillation would be observable more easily. In our model with only inertia (mass) load the cylinder pressure signals (A and B) together determine the acceleration, so the oscillation is also visible in cylinder pressures.

The velocity and displacement signals are integrated or double integrated acceleration signals so those signals are heavily filtered (by integration).

To get rid of the error probably small integrator (I term with small gain) in controller would be needed. High gain would be a stability risk again.

Making the system work “faster” (small rise time) is rather impossible (in this case) with just manipulating controller parameters since during the (fast) movement the proportional control valve is fully open and to get more speed we would need more pressure difference over the valve or a cylinder with smaller piston areas. Higher pressure difference could be realized by raising pump pressure (PRV or pressure compensated pump setting). Also the acceleration phase could be shorter/faster by having more pressure and thus also force.  

The overshoot in our case is small or can be absent. The definition in this case is “the ratio of difference between output y’s first maximum and its new steady-state value to its new steady-state value”. To be very precise we should compare the “first maximum” and the “new steady state value” which in our case is not exactly the command value because of steady state error.  However, the steady state error is only about 1 mm, so the numerical “error in overshoot value” is not very meaningful.

The 95% rise time used (time it takes for the response to rise from zero to 95% of the steady-state response) is probably not as practical as 5-95% or 10-90% would be. In our simulated case the numerical noise is very small but in practical systems it can be significant so it could be more functional to use a criteria which does not start from zero. In systems with noise it can be difficult to identify when the variable is “in zero”.

Settling time (the time that after a stepwise change in system’s setting value w is required for the process output y to reach and remain inside a band whose width is equal to ±5% of the total change in y, sometimes also other bandwidths are used, e.g., ±1 %, ±2 %) in our studied case was (about) the same as the rise time because the signal never got out of the ±5% “pipe” after it had entered it.

The teacher is happy with the students’ results. Encouraged by this year’s experiences this exercise will be further developed and used in a new and hopefully enhanced form the next year.