Calculation software hours for a project

I have always a problem when I have to do calculation for the number of hours for software.
Is there some kind of rule to calculate number of hours for PLC programming and HMI design?
In addition to that how do you calculate the hours for installation and testing of the software?

There's no substitute for experience.
There are several ways of doing it. We have an estimator here that does it by using the number of I/O and number of HMI screens and assigns a time accordingly.
If I can track him down I'll ask him what figures/formula etc he uses.
As for debug, really the same applies, it will depend on the I/O and the complexity of the program. It also helps if you're pretty good at getting your program as close as possible on the first attempt.

My method has always been as follows:

1) Define the scope and/or complexity of the project. In other words, I need to program ten conveyors, five solenoids, one HMI, and twenty pushbuttons/pilot lights.

2) Make some kind of educated guess as to what each will take. For example, I can write the logic for ten conveyors in twenty hours.

3) The most important step: Track your time accurately. Be fair to yourself. If you get stuck on a strange problem that will not happen again, for example you are unfamiliar with how to use a certain instruction, don't count that against you for this step.

4) Compare your estimate to actual.

5) Make adjustments as required.

Over a longer period of time, you will get darn good at estimating your programming time.

However, when I was a consultant, I found my total time to break down as follows:

30% Electrical design - generate CAD drawings of schematics and layouts

30% PLC and HMI programming

40% Start-up and troubleshooting - This was usually due to a mechanical machine delay issue. For example, if a double acting solenoid is wired with the A and B coils swapped, I can move a wire on a terminal strip quick. If the pipefitter has to swap hard-piped hydraulic tubes, we're talking two hours.

I always had enough time to make my software changes while the mechanical people made their fixes... In fact, I used to leave myself a little something to do to kill time while the machine problems were resolved, such as fault logic implementation.

For IO count based estimation -
16 pt IO card = 8 hours.
Analog = 1-5 hours/point depending on funciton of the point.
Then I add in some extra based on the complexity of the task.

As MartB said, there is no substitue for experience - but even that often fails. I usually still take it in the shorts. I've gotten fairly decent at estimating programming time (mainly because my programming time has decreased with expereince, not because I've become a better estimator) But fifteen years of experience and I'm still underestimating the startup and debug time.

Thanks MartB,



It would be nice if you give me the figures/formula’s from your estimator. Can you also ask him if there are other things which are calculated with formulas for example motion, regulators etc.

And then there is PID

Hi there Henry

I use the I/O count and a fixed price per HMI screen. For the Network comms I usaully work on the I/O point price x 16 per network point. For Analouge the I/O point price x 12 per analouge input or output.

Then a bit of a thumbs suck for the commisioning onsite as the plant always has a couple of bugs.

Hope this will help
Cheers
Andre

my rule of thumb is $100.00 per i/o point, including hmi tags

It usually boils down to "is it 2 weeks or 3?"
Most of the time I give estimated costs, not fixed prices and can usually justify this to the client because the software processes are rarely written with all control variables considered and commissioning is always slowed down by electrical/mechanical changes required.
I also never come to the client and say "your scope says 32 drives and its now 33, give me more money".
My estimate is really based on the specification a little on the IO count and 8 hours per SCADA page.

i do not calculate hours, i just do work right in time for fixed price per project, with $100 per i/o+HMI tags i would get more money then the cost of whole project always is

At my old job, I would guess. Then my boss would double it. Turned out to be pretty accurate.


Here, it is a little more complicated. We have some software with numbers generated by experience.