Combinational logic vs Sequential logic (Petrinets) in PLC

[Peter Nachtwey]

Sorry... not my definition , but the generally accepted !
A Moore implementation is very dangerous in case a sensor fails !
You know why ... ?
It's nearly human-impossible to manually convert a Mealy to a Moore and prevent such dangerous behavior !

(therefore Mealy bites the dog )

You can, it takes more states. I was using Moore state machines back in the early 1980s. One thing we had to add is a time out so if a input didn't get activated within a certain amount of time the machine would stop but then it was easy to indicate what sensor caused the machine to stop.

 

[Sioan]

You can, it takes more states. I was using Moore state machines back in the early 1980s. One thing we had to add is a time out so if a input didn't get activated within a certain amount of time the machine would stop but then it was easy to indicate what sensor caused the machine to stop.

The problem bei a Moore FSM ist that if a input is missing (failed sensor) you (can) have another (unwanted/unexpected) output , a time limit do not really help here ! A Moore FSM do not blocks itself in such a case.
On the contrary a Mealy FSM just blocks itself in such a case ... It is not so easy to manually convert this Mealy to a sure Moore
( ...you must think every sensor has an accident-limited life , even if you do not know precisely which one ... "dies" first )


But do not be worried : all this "theory" is true only for hardware/real INPUTS ( ...not for software/abstract inputs ... they maybe cannot fail )

 

[Sioan]

ExAMPLE :

If you have (a Moore FSM):
inputs -> output
a,b,c---> x
a,b,Nc -> y

(Nc-means-NOTc)...what happens if your input c fails ? ...how can help a time limit here ?

 

[Sioan]

)))))))))))))))))

I think (to myself) you can solve it ...?... using integers as additionally inputs/outputs ...
Is it TRUE ?

 

[Peter Nachtwey]

Some how the world keeps turning and doesn't stop just because of a hardware failure

If it is critical then use a redundancy.

 

[Sioan]

If it is critical then use a redundancy.

I think it's a (weak] solution , but otherwise someone could lose his hands in the machine ...

 

[DamianInRochester]

Why?

I think it's a (weak] solution , but otherwise someone could lose his hands in the machine ...

Huh? Because all of a sudden it is OK for people to put their hands in an unguarded machine just because it is programmed really well?

Any sequencer of any fashion should be checking to ensure that the sensors are properly responding to outputs. If they are not the program should be handling those failures in an appropriate fashion. This is a function of programming practice and has nothing to do with state machines themselves.

If failure of a sensor puts the machine in an unsafe condition, then this is a matter of poor mechanical and controls design. Poor programming should not be allowed to cause a dangerous situation.

None of this is particularly difficult to overcome. So I am failing to see where you are going with all this.

Whether it is "difficult" to manually convert between and Mealy and a Moore irrelevant. Why would this even be a requirement? Unless I am programming an FPGA, why would my boss NOT fire me for spending my time doing this conversion rather than writing the program correctly the first time and getting the job done?

 

[Sioan]

Not necessary ...hands ... not necessary machine ! But necessary Mealy automaton !
It can also be a rocket ( for example Ariane 5, do you remember )...or Chernobyl's factory ...etc.

The whole story here is :

You can not easy programing a sequencer in Moore's way to do that ...because it's no sequence there

 

[Vartile]

I don't understand (not that I never had courses of state machines ). Why not just make working program.

If you have PID running is that somesort of state in the (paper)statemachine?

 

[Peter Nachtwey]

I don't understand (not that I never had courses of state machines ). Why not just make working program.

Pragmatic but it is best if you know how to organize a state machine. It saves time when you need to do one. Since I get involved with a lot of machine control I consider state machines to be very basic.

The truth is that too many of you can't write a decent state machine. I know because I make motion controllers. In the past the state machine had to be done in the PLC and too often we had to help the PLC programmers get it right.

If you have PID running is that somesort of state in the (paper)statemachine?

PIDs are a continuous process. Controlling temperature is a continuous process. In machine control there is a lot of "do this" "wait for this to be done" then "do that" then "wait for that to be done" then "do the next thing" then "wait for the next thing to be done" etc then repeat for the next part.

Vartile, can you see the difference between a PID and machine control?

Sioan is just trying to screw with your mind.

SFCs or Sequential Function Charts are basically Moore state machines. Once you are in a state you stay there until a transition has occurred to go to the next state. Our motion controller has a build in state machine that functions in a similar way to SFCs. It looks different but pretty much works the same way.

Now Sioan is trying to tell us the IECC and we are all wrong. Meally and Moore came up with the formal idea of state machine in the mid 1950s. Sioan would have us believe the whole world has been doing it wrong since then or that the whole world never did get it right. I think Sioan should study more and get more experience before making his comments. At least he should back up his claims with links to credible articles.

 

[Sioan]

I don't understand (not that I never had courses of state machines ). Why not just make working program.

If you have PID running is that somesort of state in the (paper)statemachine?

Sure that a PID is a finite state machine (in a loop) ! ... isn't it DIGITAL ?

 

[DamianInRochester]

Sure that a PID is a finite state machine (in a loop) ! ... isn't it DIGITAL ?

It may very well be at a lower level. Clearly not at the development in the context of what anyone here is going to be talking about. That is obvious right? We are talking about PLC application programs, not the OS those applications run on.

 

[Sioan]

Sioan is just trying to screw with your mind.

SFCs or Sequential Function Charts are basically Moore state machines. Once you are in a state you stay there until a transition has occurred to go to the next state. Our motion controller has a build in state machine that functions

A Moore state machine has a next state but never knows it !
A Mealy state machine has a next state and knows it !

☯

...big difference isn't it ?

 

[Sioan]

It may very well be at a lower level. Clearly not at the development in the context of what anyone here is going to be talking about. That is obvious right? We are talking about PLC application programs, not the OS those applications run on.

I beg your pardon !
A (digital) PID is very well at a normal level a finite state machine : ( is a just a finite number of ... CPU instructions/states the ... PID)

How many lines of code need you for ...

someone have to really see the FSM states between the lines of his code ... to know about ... cybernetics

 

[DamianInRochester]

I beg your pardon !
A (digital) PID is very well at a normal level a finite state machine : ( is a just a finite number of ... CPU instructions/states the ... PID)

How many lines of code need you for ...

someone have to really see the FSM states between the lines of his code ... to know about ... cybernetics

You clearly did not understand what I said. I would suggest you read it again.