Axxom: What happened so far

1
Axxom What happened so far
2
Basic case study

• lacquer production scheduling
• 3 recipes for lacquers,
• specifying processing steps,
• resources used (shared resources)
• timing dependencies between processing steps
• 29 orders with
• starting time, due date
• recipe, amount

first question is there a feasible schedule?
Solved with heuristics (non-laziness,
non-overtaking...)with IF and UPPAAL
3
First extension

• performance factors describe break down of
  resources
• availability factors describe working hour
  constraints

Axxom approach extend processing times by
performance And availability factors.
second question is there a feasible schedule for
the extended processing times?
Solved with heuristics (non-laziness,
non-overtaking...)(with IF and ) UPPAAL
4
Costs

• storage costs for products that are finished too
  early
• delay costs for products that are finished too
  late
• set up costs (colour change on resources)

question what is the cost-optimal schedule?
Martijn and Gerd will report on modelling and the
solution..
5
Working hours

• working hours are Monday till Friday 8-20hrs.
• there are no processes running outside working
  hours

question how to model? What are cost-optimal
solutions?
Martijn and Gerd will report on this..
6
Scaling up

• a 73 job version
• a 219 job version
• with extended processing times
• no costs, no working hours

question is there a feasible schedule?
The 73-job version could be solved with more or
less the29-job approach.For the 219 job version
we came into problems with clock numbers (one
clock for each job) Idea treat also clocks as
shared resources, only active jobs use a clock.
-gt Gerd
7
Probabilistic evaluation

• Approach
• gt take schedules generated by UPPAAL,
• i.e. take job starting times from schedules
  generated.
• gt Make a probabilistic model taking the machine
  breakdowns into account (MODEST)
• gt Simulate the processes (Moebius)

Result the schedules derived with extended
processing timeshad a higher probability for
delay
Reasonif we reserve time for possible break
down, this timeis wasted when there is no
break-down.
-gtQEST
8
Generating schedules taking probabilism into
account

• Basic scheduling questions
• Long term scheduling how many orders can be
  treated?
• Short term scheduling what to do now?

Questions Should both questions be treated with
the same models? What parameters go into which
model? (colour changing costs,performance
factors, ....) What do performance factors mean
in the context of shortterm scheduling?
Holger will discuss this in more detail
9
Similarities between long-term scheduling and
performance analysis
-gt ideas from Henrik