Stochastic Models in Planning Complex EngineerToOrder Products

1
Stochastic Models in Planning Complex
Engineer-To-Order Products

• By Dong-Ping Song
• Supervisors Dr. Chris Hicks,
• Prof. Chris F. Earl

Department of Mechanical, Materials and
Manufacturing Engineering, University of
Newcastle upon Tyne, UK, Oct. 2001
2
Objectives of Thesis

• Developing a series of effective methods for the
  planning of ETO manufacturing systems producing
  products with complex product structure and
  various uncertainties.

3
Engineer-To-Order System

• ETO products are engineered and produced based
  on the specifications of the customer. Order
  tendering and engineering design are included.
• Characteristics highly customised low volume
  complex product structure uncertainties two
  major stages.

4
Work-flow and Planning Levels of an ETO System
5
Special Issues in ETO Planning

• Product due date planning (-- level 1)
• Stage due date and activity start times planning
  (-- level 2)
• Production scheduling (-- level 3)
• Dynamic production scheduling (-- level 3)

6
Structure of Thesis
Ch 1 Ch 2
high
Ch 3
Product due date planning
Ch 4 Ch 5
Stage due date planning Activity start times
planning
Planning level
Ch 6
Ch 7
Production scheduling Dynamic scheduling
Ch 8
Ch 9
low
Ch 10
7
Key Assumptions

• Probability distribution of operation processing
  times are available
• Products have specified structures of
  manufacturing and assembly operations

8
Product Due Date Planning

• Objective find a better product due date using
  the information on processing time distributions
  of operations in a multistage product structure.
• Method moment-based approximation.
• Possible application quote reliable delivery
  date at order tendering stage.

9
Stage Due Date and Activity Start Time Planning

• Objective planning due dates at each stage to
  meet specified service targets.
• Method recursive procedure with moment-based
  approximation.

• Objective planning activity start times with
  given product due date to minimise expected
  earliness and tardiness cost.
• Method Perturbation Analysis Stochastic
  Approximation (PASA).

• Possible application set good start and due
  dates for each stage in multistage assembly
  systems by taking into account the effects of
  uncertainty.

10
Production Scheduling

• In stochastic systems, a schedule must be capable
  of dealing with the situation,
• when a resource finishes one operation, the next
  scheduled operation has not arrived but several
  other operations are queuing at this resource
  which operation should the resource do next?

Choice I Keeping original schedule ? scheduling
problem I
Choice II Applying a priority rule ? scheduling
problem II
11
Two Type Scheduling Problems

• Type-I scheduling problem find optimal
  sequences and timings by minimising expected
  total earliness and tardiness cost, where the
  current operation sequence is followed when
  determining the cost of a particular candidate
  schedule during optimisation.
• Type- II scheduling problem find the optimal
  operation timings directly by minimising expected
  total earliness and tardiness cost, where the
  operation sequences are unknown in advance due to
  uncertainty of operation durations (a priority
  rule is used).

12
Two-phase optimisation method to solve type-I
scheduling problem
Ch 6
Ch 7
13
One-phase optimisation method to solve type-II
scheduling problem
Ch 8
Sn a schedule. EPST earliest planned start
time
14
Production Scheduling applicable situations

• Deterministic scheduling problems using
  Heuristic, Simulated Annealing, Evolution
  Strategy methods in Ch 6.
• Stochastic timing scheduling with fixed sequences
  using PASA, Simulated Annealing, Evolution
  Strategy methods in Ch 7.
• Type-I scheduling in stochastic situation using
  two-phase optimisation method in Ch 8 based on Ch
  6 and 7.
• Type-II scheduling in stochastic situation
  using one-phase optimisation method in Ch 8.

15
Dynamic Production Scheduling

• Incremental planning generate an incremental
  plan for the new order without affecting the
  production schedule for the existing orders.
• Regenerative planning regenerate a plan for the
  new order and those unfinished existing orders.
• Assumption deterministic operation times.

16
Dynamic Scheduling methods
Incremental Planning
Regenerative Planning

• Forward incremental planning
• Backward incremental planning
• Evolution Strategy incremental planning

• Forward regenerative planning
• Evolution Strategy regenerative planning

17
Conclusions

• Developed a series of effective methods for the
  planning of stochastic ETO products.
• Investigated the effects of complex product
  structure and uncertainty in processing times.
• Examined the effects of dynamic arriving orders.
• Addressed specific planning problems such as
• product due date planning,
• stage due date and activity start time planning,
• production scheduling,
• dynamic incremental planning and rescheduling.

18
Limitations

• Distribution of processing times are required.
• Random variables in Ch 35 are assumed to be
  independent.
• Process planning is not considered and product
  structure is pre-specified.
• Set-up and transfer times are not explicitly
  considered.