Tracking hourly production

[Jeebs]

Do you mind explaining this a little further? Variable length? I didn't know that you could make the length variable, thought it was a constant.

This just means that your FIFO won't have ~700 items in it at all times. One minute it'll have 558, the next it might have 567.

With this you can show nice graphs every hour, on the hour. And you can make nice pop-ups, flashy things once they reach a new record, or when they go over their production target.

Depending on the impact on scan time, you can either add the 500+ items or you can subtract the few, that fall outside of the hour, from the previous count.

 

[bernie_carlton]

If you notice in my first post, even though there are 60 items they don't have to be added each minute. The addition is happening continuously. The COPy command which shifts the table (much like a FFU though it takes less time) is the only command which uses any significant time.

 

[monkeyhead]

My question now is...what is the best way to add 60 integers together in a SLC 5/03? As Bernie said, there is no SUM function.

I used a similar scheme on a program I did and my solution was to add the new value that was about to be pushed to an integer that served as the overall count, and then subtract the value that was about to be popped just before loading the FIFO.

Newest_Sample + Running_Count - Oldest_sample

 

[brucechase]

Or, with a lot a work, you could just throw all the data into a database then use VB or something similar to do all your calculations. Every time the switch activates, have the database store the time in a register. All kinds of graphs, calculations, downtime, rates, etc can be then done using simple web pages. At least that is alot like what we do here - and man oh man do we have some graphs to look at.

One thing to be careful of, as soon as the production people figure out which switch it is and the reward is enticing enough (money, pizza, etc), someone will go over and activate the switch dozens of times to get their number up. Soon, your plant will be putting out 800% of the daily totals.

 

[milldrone]

One thing to be careful of, as soon as the production people figure out which switch it is and the reward is enticing enough (money, pizza, etc), someone will go over and activate the switch dozens of times to get their number up. Soon, your plant will be putting out 800% of the daily totals.

To get around the people problem. I have used other parts of the process. Like another switch that activates in a one to one ratio with the production monitoring switch. If none are available, monitoring inbound feed conveyor(s) for material flowing through the machine interlocked with the production counting switch will work.

 

[shoelesscraig]

One thing to be careful of, as soon as the production people figure out which switch it is and the reward is enticing enough (money, pizza, etc), someone will go over and activate the switch dozens of times to get their number up. Soon, your plant will be putting out 800% of the daily totals.

Already thought of this...I know how sneaky of a guy the department manager can be. So, I've got a few tricks up my sleeve for that one.

And Johnster, my managers have plenty of different ways to use this information. I work for a MASSIVE company, and I am creating most of this information for corporate use, historical reports, and permanent records. The "contest" that one of our managers come up with was just a little something extra to build a little competition between shifts. I am also tracking about 20 or 30 other pieces of information for them as well, so trust me, they know how to manage the data.

 

[demmons]

How about a low tech method. Convert production rate to an analog signal, hook the analog output to a chart recorder, then let the production gurus sort the data...

 

[shoelesscraig]

How about a low tech method. Convert production rate to an analog signal, hook the analog output to a chart recorder, then let the production gurus sort the data...

Yeah, but I'm a high-tech kinda guy...

Na, don't feel like running wires across the building and having the charts to keep up with. Everything else in the plant is 100% completely automated...I wouldn't want to go back old school when I can do it easier this way.

 

[shoelesscraig]

I hated to dig up this old thread of mine, but I thought I would post to say what I ended up doing...thought it might help someone who searches for something similar.

I ended up using the "rolling average" method that some people suggested. It worked perfectly in my particular application. Basically, I count how many packages pass the limit switch every minute. I take this value (each minute) and load it into a FIFO stack with a length of 60. When the length of the FIFO has reached 60, I start unloading the oldest record, in other words I get rid of the data that was recorded 61 minutes ago.

When production starts, I start accumulating an integer as my total. During the first hour, my total is just this integer alone. After an hour has passed, each minute I just subtract the oldest minute and add the newest minute, thus getting my last hour's production. From here I can divide by 60 to get the average minute's production, etc.

And just as I thought, and others of you did as well, it wasn't long before people started trying to figure out where the numbers were counted from so they could try and cheat and make their numbers higher. So, I set up some logic to determine if the limit switch was being actuated by hand. In the event that it is, for that minute it adds 0 to the FIFO stack (regardless) of what the actual production number should have been. So, you actually get penalized for messing with the limit switch. That'll teach em'!

My thanks to those who responded with ideas and suggestions. Your help made it happen.

 

[Honga]

Personally I prefer the LiLo approach... just me though.