job shop scheduling

[snibril]

i'm trying to develop an algorithm/heuristic for scheduling in a job shop. it's a dynamic problem,the jobs have quite random routing around the shop and the times used for tasks are subject to probability. can anyone suggest any good texts to refer to?

 

[TonyE]

In a job shop environment where job times, resources, etc. are almost random, the only suggestion that I could make is to view the backlog and estimate the resources and times to complete the job.  Those estimates can then be used in scheduling.  Depending on the types of jobs that the shop will take on will determine if statistical quantification can even be used.  As far as a "good text" to recommend, I've not found any that really address this issue.

 

[johnkarlo]

When you have many resources and their use is aleatory, the best option is focusing on the least vital, it means, focus on that things which restricts the throughput of your business.&nbsp;&nbsp;That theory has been extensively explaned by Eliyahu Goldratt in his books, which are&nbsp;&nbsp;&quot;The Goal&quot;, &quot;The Race&quot;, and &quot;El sindrome del pajar&quot; (I don´t know the name in english but it probably is &quot;the barn syndrome&quot.<br><br>I know there is some web pages which treats this issue, try in some search engine with &quot;goldratt&quot;, or &quot;constrains&quot;.<br><br>I hope this helps

 

[betojau]

Why didn't you use Theory if Constrains, linking your demand with your schedule.
Or use the Drummer software.

 

[Drazen]

It seems very interesting to establish some theoretic criteria for shop task scheduling, but I am not sure if it can prove itself on shop floor.
Shop is a real life and statistics can show some conformity to real life over a long period, but scheduling is on day-to-day basis and you could end up with total chaos many days while total result over, say, two years is not so bad.
This doesn't seem accepatble for practice. I strongly beleive in systems derived from long successful practice. One such system is throughly explained in Doc Palmer's &quot;Maintenance Planning and Scheduling Handbook&quot;.
You are guided through all steps of implementation, which are persuasively supported with practical examples and common sense arguing.
Maybe that can give you most of answers.

 

[IainMcF]

Job-shop scheduling is almost always done in a batch modality. Jobs, routings, material and support resources for rouings, machines, operations, along with operation networks, delays, lags, etc are all fed into a scheduling engine and the schedule is generated. This is often referred to as the &quot;paper tiger&quot; approach and 'can' be useful in a job-shop. At least one can generate dispatch reports and try to follow them. I have used this type of system many times and have had good results. Especially useful for determining resource usage patterns and material requirments for known demand. Scenarios can be put through a &quot;bake-off&quot; and an course of action determined with relative confidence. The &quot;batch&quot; approach does, however, result in a production schedules that may be hard to actually achieve. As was mentioned, the real-world intercedes with &quot;optimization&quot; techniques.

Real-time scheduling brings together the algorithms of optimization with the infrastructure of information technology. This &quot;always on&quot; scheduling approach is ideal for job-shop environments. Coupled with &quot;visibility&quot; tools and monitoring technologies (OPC, historians, etc), these types of systems can offer valuable trending information for longer-term analysis while ensuring that production bottlenecks are addressed (TOC-style) and throughput maintained.

Of special note for the job-shop environment is the efficacy of &quot;visibility&quot; tools. Knowing that some job/process is going off-side ahead of time is invaluable in maintaining schedule compliance. The main problem here is that it requires a high degree of shop-floor discipline to ensure that feedback on production status is timely.

There are a variety of systems available today that can address your needs. Developing an algorithm is only part of the problem; creating a system that handles the real-world is a bigger problem.
Iain

 

[gato]

I suggest the book &quot;Factory Physics&quot;.