Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror

Comment Re:Still no reason for putting idiots on the job (Score 1) 96

Sensors that spit out text? Who in their right mind would want that?

SCADA grabs sensor readings from the underlying control system, most likely running on some PLCs, where you have to do calculations on the data in order to feed back control values to the process being controlled.

Now, a PLC is, admittedly, sort of like the general purpose CPU's dumb brother, and the instructions it accepts are rather limited. But, for a number of reasons, they're immensely suited for their task. The single most important one being the ability to safely and easily change a program that is in production. This feature is important because control systems often have to be tuned when they're commisioned, they don't just work out of the box. You have to fiddle with constants in order to get it working, perhaps even change the structure of the control algorithm.

Because we control engineers have to fiddle with the program while it is running, we really don't want to do string to int/float/whatever conversion and the reverse when working on the PLCs. That would just be yet another place where we could scew up horribly, causing a country wide electrical blackout in the proces. It's hard enough as it is, so keep it simple, stupid!

Comment Re:Replication of results? (Score 1) 73

There were two teams at the LHC that independantly came to the conclusion that there was a particle.

And, though you might have been ironic, you caught the exact reason there are two independant sets of sensors, data analysis, etc. (Everything besides the accelerator ring.) Now, a third, completely independant reproduction of the result would be golden, but until we get it we'll have to make do with just a single reproduction of the result.

Comment Re:Quick couple of questions (Score 3, Interesting) 73

Consensus has nothing to do with science. Following the scientific method is science.

And the particle the people at CERN have discovered is consistent with what we believe to be a Higgs boson, and might therefore actually be one. We haven't had time to do enough experiments to tell. The only things we know about the particle is that it's there and that it has a mass of ~126 GeV. We're assuming that it's the Higgs, because the Higgs is the only particle that is missing with a mass in that neighbourhood. (Gravitons ought to be heavier, as do dark matter, etc.)

Comment Re:Doesn't matter in the end (Score 1) 472

A feedback controller is a good example. Sure you can test it against a model, but you just won't know until you've had it running against the hardware. Also, models are expensive, so clients often don't want to pay for their development, but will allow for a bit more debugging and tuning against the hardware instead.

Comment Re:You should never stop learning (Score 1) 260

I would recommend the same approach, i.e., write papers and get them published, but for another reason.

If, after a while, you realise that grad school just isn't you, perhaps because of a stingy professor or something along those lines, then, when you quit, you'll still have your publications even if you didn't get your degree. And, since the papers have been through the peer review process, you'll have the word of your peers that you are a capable researcher.

Comment Re:Color me unimpressed... (Score 1) 237

It is a "solved" problem. But that doesn't make it easy. Far from it.

When controlling an object like this you'd usually use a state-space controller, most likely an optimal controller.

First you need a mathematical model of the system. Since this is almost entirely a kinematic system they've most likely constructed a non-linear model using first order principles and then derived a first order approximation of it. The model won't be completely accurate, so they'll have to test it up against the actual system to see if it is "good enough". (Where the ability to judge if a model is "good enough" comes with practice.)

Then, using the linear model, you can construct an observer so that you'll be able to estimate the states from the actual measurements. In this case I'd probably go for a Kalman filter.

Finally you can synthesize you optimal controller, that generates the control signal from the estimated state.

As far as I can tell, this is the first untethered flight of this system. Therefore they haven't had the chance to test if their mathematical model is correct, since there's a good chance the tether will change the dynamics of the systems, and errors in the model will result in a bad estimator and a bad controller, that, because the estimator is bad, will be fed false information. Getting reality to conform to the used math is no simple feat.

...and that was just the mathematical core of the control system. All the calculations has to run on a RT-OS, and all the electrical components, e.g. computers, sensors and actuators, have to work perfectly. Add to that all the various start up procedures that also have to be controlled, the trajectory calculations, etc. etc.

Comment Re:The last sentence (Score 2) 98

You've got it backwards. Coloumb is defined as c = a*s, while Ampere is defined as the constant current that will produce an attractive force of 2 × 10^–7 newton per metre of length between two straight, parallel conductors of infinite length and negligible circular cross section placed one metre apart in a vacuum.

Coloumb, as a unit, is derived fra Ampere. Furthermore, Coloumb is a measure of charge, not electrons, in the same way that Ampere is a measure of current, not electrions/s. If you know that your current results from a stream of electrons, instead of say, ions, protons, or positrons, then you can calculate the corresponding electrons/s.

Still, you're right that GP is wrong.

Comment Re:Um (Score 2) 84

It seems they were, but catching fabricated results like these isn't exactly easy and it won't happen in the review process.

In order to catch fabricated results you'd either have to repeat the experiment, which nobody wanted to do since the research was low impact, or catch discrepancies in the data, which was how he was caught out.

Slashdot Top Deals

The value of a program is proportional to the weight of its output.

Working...