Balancing Operations

[EH003268]

I have a worksheet that collects data for various operations within a process. These process times are variable depending on the process. See attached. I have tried several methods, but it still does not work correctly.
The idea is that i can find the manning level that gives the best balanced process. Work time for each operator is as close as possible.
I can use 1 to 12 operators.
There may not always be 12 cycle times.
I need to be able to change manning level.
Any 1 operator can do 1 or more operations.
Balance tolerance ideally would be 15%.

Hope someone can help.

[Leith Ross]

Hello EH003268,

I would have posted this sooner if I hadn't forgotten were the article was. This is a interesting method that you may find useful in your situation. It uses the Solver program (part of the Analysis ToolPak add-in for Excel) to help schedule employees. Hopefully you will find the article useful.

http://office.microsoft.com/en-us/ex...245981033.aspx

Sincerely,
Leith Ross

[shg]

I sorted your list of tasks in descending order by time, then assigned workers in column E as (1 to n, n to 1, 1 to n ...) e.g.,

for 5 workers: 1,2,3,4,5,5,4,3,2,1,1,2
for 6 workers: 1,2,3,4,5,6,6,5,4,3,2,1
for 7 workers, 1,2,3,4,5,6,7,7,6,5,4,3

Then I wrote a formula that computed the deviation of the work times (=STDEV(OFFSET(G4,0,1,1,nWorkers)), and wrote a routine that swapped every task between every pair of workers to try to reduce the deviation, but it never made a successful swap -- so the simple method above is not bad.

I can't tell what your formula in column D is supposed to do.

[EH003268]

Thanks for your replys.
I am looking into using the solver.
One of the problems is that the work must be conducted in the same order as listed. It is a one piece flow production line, and so it is generally not feasable to change the order of the operations.

[shg]

One of the problems is that the work must be conducted in the same order as listed.

OK; please take a look at the attachment. I think it does a decent job.

[EH003268]

Thanks for the input, but still not quite there.
Operations cannot change order.
Also Operators cannot change order. Operator 1's work must be followed by operator 2's work etc.
The operators are sitting in a line and cannot move around the assembly process.
There are not always 12 operation times.

[shg]

Operations cannot change order.

Got that.

Also Operators cannot change order. Operator 1's work must be followed by operator 2's work etc.

That's not true if what you said in your first post remains valid:

Any 1 operator can do 1 or more operations.

If what you mean is that operation 2 must follow operation 1, then you already said it.

So: N operators are sitting at stations 1 thru N of a M-station (N<=M) assembly line. As each completes his task for all the day's widgets, he goes to the next vacant station, where either work has piled up, or he waits for work to arrive, or perhaps arrives just in time to start work. This continues until the last widget is out the door for the day.

So what are the opportunities for load balancing?

[EH003268]

We are running a one piece flow, so all required operators sit together at a line of benches. The first operator begins the process i.e. cut the cable to length, and when the designated operation/s are complete, the semi complete assembly is passed to the next operator. Second operator completes the assigned tasks i.e strip the wire, and fit a sleeve, then passes the semi complete assembly to the next operator in the line. This continues through the operators until the assembly is complete. There is only 1 assembly at each work station at a time.
If operator 2's work (1 assembly) is longer than operator 3's then operator 3 will be waiting. On the other hand if operator 2's work is shorter, we get a build up of semi complete assemblies at that part of the line, and this is not allowed.
To make the process efficient, with no waste, the operations must be balanced between the operators. Each operator should have approx the same amount of work.

Hope this helps explain the issue.