1. Rescheduling Manufacturing Systems: a framework of strategies, policies, and methods By Guilherme E. Vieria & Jeffrey W. Herrmann, Edward Lin Represented by Nuriye Kaptanlar

2. Introduction... Production Schedules  state when certain controllable activities should take place  help in coordinating activities to * increase productivity * reduce operating costs !!!BUT not ENOUGH. Rescheduling  updating a production schedule when it becomes INFEASIBLE.

3. Introduction... Aim of the paper To standardize the terms To classify * the strategies * the policiesframework * the methods in rescheduling literature. To show how rescheduling affects the performance criteria. To bring practice and theory closer.

4. Introduction... Objectives in scheduling Generate high quality schedules React quickly to disruptions Revise schedules in a cost effective way Notation in static scheduling  /  /   : scheduling environment  : characteristics of the jobs to be scheduled  : objective function Classifications DO NOT consider rescheduling context.

5. Introduction... Types of studies in rescheduling literature * for repairing a schedule that has been disrupted * for creating a schedule that is robust with respect to disruptions (robustness will be considered later) * how rescheduling policies affect the performance criteria.

6. Introduction... Framework * Rescheduling environment: identifies the set of jobs that need to be scheduled * Rescheduling strategy: describes whether or not production schedules are generated * Rescheduling policy: specifies when rescheduling should occur

7. Introduction... Paper Organization * Rescheduling in general * Scheduling terms, rescheduling framework * Performance measures * Rescheduling strategies & policies * Robust schedules & updating them * Effect of rescheduling policies on system Performance * Gap between theory and practice

8. Rescheduling in general … A production schedule CAN Identify resource conflicts Control the release of jobs to the shop Ensure that required raw material are ordered in time Determine whether delivery promises can be met Identify time periods available for prev. maint.

9. Rescheduling in general … Disruptions Arrival of new jobs M/C failures M/C repairs Types of manufacturing systems considered Single M/C Parallel M/C Flow shop Job Shop Flexible Manufacturing cells and systems

10. Rescheduling in general … Factors & triggered actions M/C failure Urgent job arrival Job cancellation Due date change Delay in the arrival Shortage of materials Change in job priority Rework or quality problems Over or underestimation of process time Operator absenteeism Overtime In-process subcontracting Process change or re-routing M/C substitution Limited manpower Setup times Equipment release Cause

11. Terminology & framework… Definitions Manufacturing system: organizes equipment, people and information to fabricate and assemble finished goods. Order release: controls a manufacturing system’s input by determining which orders should be moved into production. Shop floor control: determines which operation each person and piece of equipment should do and when they should do it.

12. Terminology & framework… Definitions Production schedule: specifies for each resource required for production, the planned start and end time of each job assigned to that resource. Scheduling: process of creating a production schedule for a given set of jobs and resources. Rescheduling: process of updating an existing production schedule in response to disruptions or other changes.

13. Terminology & framework… Definitions Rescheduling environment: identifies the set of jobs that the schedule should include. Rescheduling strategy: describes whether or not production schedules are generated. Rescheduling policy: specifies when and how rescheduling is done. Rescheduling methods: generate and update production schedules.

14. Terminology & framework… The focus of studies in Rescheduling: on one or more well-defined scheduling problems Establishes the problem’s computational complexity Identifies properties of optimal schedules Proves the optimality of an exact solution approach Compares the performance of heuristic solution approaches experimentally.

15. Terminology & framework…

16. Rescheduling Environment Static rescheduling environment: have a finite set of jobs Dynamic rescheduling: have an infinite set of jobs

17. Terminology & framework… Deterministic static: finite set of jobs and no uncertainty. Stochastic static: finite set of jobs but some variables are uncertain e.g. processing times are random variable * modify the schedule during execution to react additional information. * construct a solution that only partially specifies the schedule. * uncertainty is not modeled as a probability distribution.

18. Terminology & framework… Dynamic rescheduling: 1.No uncertainty or variability in the arrival process  cyclic scheduling 2.Uncertainty in job arrivals, same route for all jobs, arrival rate steady  significant set ups 3.Variability in process flow and job arrival Presence of additional capacity (subcontracting & overtime  relax in capacity versus overtime costs)

19. Terminology & framework… Definitions (continued) Scheduling point: point in time when a scheduling decision is made Rescheduling period: time between two consecutive scheduling points Rescheduling frequency: measures how often rescheduling is performed Scheduling stability: measures the number of revisions or changes that a schedule undergoes during execution.(Cowling & Johansson)

20. Cowling: Utility versus stability… Utility will measure the benefit which may be gained by using a particular rescheduling strategy. There is a continuum of different strategic approaches to dealing with real time information between “do nothing” and “always reschedule to maximize utility without considering stability” where we may trade-off Cbar and stability.

21. Cowling: Utility versus stability…

23. Terminology & framework… Definitions (continued) Scheduling nervousness: significant changes in the schedule (opposite of schedule stability) Schedule robustness: measures how much disruptions would degrade the performance of the system as it executes the schedule !!! NOTE: Stability & nervousness measure the changes to a schedule BUT robustness measures the changes to system-level performance.

24. Performance Measures… Measures of schedule * efficiency * stability * cost. Schedule efficiency: used when generating a schedule. e.g. makespan, mean tardiness, mean flow time, average resource utilization, maximum lateness.

25. Performance Measures… Schedule stability: * static, deterministic; not considered * Impact of schedule change defined as:  the starting time deviations btw the new schedule and the original one  a measure of the sequence difference btw the two schedules [Wu, Storer & Chang] * M/C failure; major concern (stable & robust)

26. Performance Measures… Schedule Robustness: * as the level of uncertainty, frequent scheduling becomes more effective in improving the robustness. Economic Performance of the system: * due to lack of efficient time-based performance measure * Min COST (starting jobs too early, WIP, Tardiness)

27. Performance Measures… Economic Performance of the system: cont’ed * Computational costs * Setup costs * Transportation costs Computational costs: * computational burden on the computer running the scheduling system * non-recurring costs of investment * recurring costs of administration, maintenance and upgrades * time sent for generating and updating the schedule

28. Performance Measures… Setup Costs: * occur when tooling & fixtures are created or allocated Transportation Costs: (material handling) * delivering materials earlier than required * additional material handling work to transport jobs form one scheduled M/C to other points

29. Performance Measures… The relative values of the rescheduling period and the total mean processing time will affect the performance measures used in predictive- reactive rescheduling. Large rescheduling period versus small ones In job shop, scheduling objectives are complex.

30. Rescheduling strategies… Comparison of two strategies in dynamic rescheduling environment that have uncertain job arrivals; * Dynamic versus * Predictive-Reactive  periodic  event-driven  hybrid

31. Rescheduling strategies… Dynamic Scheduling: (reactive, online) Does not create schedules Decentralized production control methods dispatch jobs * when necessary * use information at the time of dispatching Dispatching rules & pull mechanisms: * when a M/C becomes available it chooses from among the jobs in its queue by using a dispatching rule that sorts the jobs by some criteria

32. Rescheduling strategies… Dynamic Scheduling: (reactive, online) Dispatching rules are classified as: * simple dispatching rules * combinations of simple rules * weighted priority indexes * heuristic scheduling rules [Panwalkar & Iskander] Coalition rule highly valued, individual rule rejected, traditional and theoretical not highly valued. [Green & Appel]

33. Rescheduling strategies… Dynamic Scheduling: (reactive, online) Control Theoretic Models: are used to develop rules for deciding which action to take an when to take it in response to random disruptions. M/Cs without setup: least slack policy (due date-E(amount of time until the job is completed)) M/Cs with setups: focus on completing all waiting jobs of one type before performing a setup and processing of another type. [Kumar]

34. Rescheduling strategies… Predictive- Reactive Scheduling: * generate a schedule * updates the schedule in response to a disruption or other event to minimize its impact on system performance Iterative process: [Wu &Li] * Evaluation step evaluates the impact of a disruption * Solution step determines the rescheduling solutions that can enhance the performance of the existing schedule * Revision step updates the existing production schedule or generates a new one

35. Rescheduling strategies… Predictive- Reactive Scheduling: Rescheduling approach: [Yamamoto & Nof] * Planning Phase constructs an initial schedule just prior to the start of a new period * Control Phase compares the actual progress of operations during a given period. * Rescheduling Phase constructs a revised schedule considering the operational changes that have triggered rescheduling.

36. Rescheduling strategies… Predictive- Reactive Scheduling: Policies studied: * periodic * hybrid rolling time horizon * event-driven Rolling time horizon: the overall scheduling problem is decomposed into smaller and static scheduling problems.

37. Rescheduling strategies… Predictive- Reactive Scheduling: Periodic Policy: reschedules the facility periodically. * used in environments where there is online data acquisition from the shop floor * yields more schedule stability than constant rescheduling * significant changes compromise performance * determination of the optimal rescheduling period is not easy

38. Rescheduling strategies… Predictive- Reactive Scheduling: Event Driven Policy: rescheduling can happen repeatedly in dynamic environments or a single event can revise a schedule in a static system. * popular in static environment and used when M/C failure occurs. * in dynamic environments used when total # of job arrivals reaches a threshold. [Viera] * require fast & reliable electronic data collection.

39. Rescheduling strategies… Predictive- Reactive Scheduling: Hybrid Policy: reschedules the system periodically and also when special (or major) events take place. e.g. major events; M/C failures, arrival of urgent jobs, job cancellation or job priority changes

40. Rescheduling methods… Methods used as a part of reactive-predictive scheduling is discussed. Concentrate on methods that generate robust schedules and methods that update schedules. Approaches for static and stochastic environments are considered.

41. Rescheduling methods… Generating Robust Schedules Attempt to maintain good system performance with simple schedule adjustments. Some studies and results: * As the amount of processing time variability increases, dispatching rules led to better performance. [Wu] (branch & bound) * Schedules that are robust to stochastic disturbances could be generated without much degradation. [Mehta & Uzsoy] (inserted idle time, shifting bottleneck algorithm)

42. Rescheduling methods… Generating Robust Schedules * When probability distributions are not available, worst case scenario is key objective. [Daniels & Kouvelis] * O’Donovan uses first a dispatching rule and then a simple policy to insert idle time btw jobs based on E(downtime) in his method aiming at minimizing E(deviation in completion times). * As the level of uncertainty increases, frequent scheduling becomes more effective in improving the robustness. [Shafaei & Brunn]

43. Rescheduling methods… Repairing Schedules Even if the managers and supervisors do not explicitly update the schedule, schedule repair occurs as the operators react to disruptions, delaying tasks or performing tasks out of order. Methods used to repair a schedule: * Right shift scheduling * Regeneration * Partial scheduling

44. Rescheduling methods… Repairing Schedules Right shift scheduling: postpones each remaining operation by the amount of time needed to make the schedule feasible

45. Rescheduling methods… Repairing Schedules Partial rescheduling: reschedules only the operations affected directly or indirectly by the disruptions. * tends to maintain schedule stability by preserving the initial schedule Studies: * Comparison of performance under the proposed affected operations method to the total rescheduling and right shift scheduling methods. [Abunaizar & Svestka]

46. Rescheduling methods… Repairing Schedules Studies: Continued * Matchup scheduling procedure uses heuristic ordering rules to resequence all jobs scheduled before a matchup point. Optimal, when the disruptions are infrequent. [Bean] * Matchup scheduling that partially reschedules a modified flow shop when a M/C failure occurs. [Akturk & Gorgulu] * Constraint based schedule repair [Smith] * Repair tactics; adjusting start times, swapping operations and switching to alternative resources. [Miyashita & Sycara]

47. Rescheduling methods… Repairing Schedules Regeneration: reschedules the entire set of operations not processed before the rescheduling point, including those not affected by the disruption. Disadvantage, excessive computational effort and unsatisfactory response time. Genetic algorithm that reuses the previous solution to solve a job shop scheduling problem every time a new job arrives. [Bierwirth & Mattfeld]

48. The impact of rescheduling policies… The impact of * Type of events that trigger rescheduling * Rescheduling frequency on performance is studied.

49. The impact of rescheduling policies… Church & Uzsoy * Hybrid, event-driven, for single and parallel M/Cs with dynamic job arrivals * EDD rule is used * Analytical models to bound the max(completion time) * Periodic rescheduling lead to near opt. (min max lateness) when order release is periodic. * Rescheduling at the arrival of a rush job is useful but more frequent rescheduling does not improve the performance.

50. The impact of rescheduling policies… Vieira * Single M/C, periodic & event-driven based on queue size * Analytical models can accurately predict the performance * Extend to Parallel M/Cs * Rescheduling frequency can significantly affect the average flow time.

51. The impact of rescheduling policies… Vieira (continued) * Lower freq. lowers the # of setups, increases cycle time and WIP. * Higher freq. Allows the system to react more quickly to disruptions but may increase the # of setups. * Event-driven and periodic strategies give similar performance. * Rescheduling when a M/C fails after a repair decreases cycle time but increases the freq.

52. The impact of rescheduling policies… Farn & Muhlemann * Via simulation * Single M/C with sequence dep. setup times * FIFO * Often rescheduling leads to lower setup costs. Muhlemann * Dynamic job shop * Compare heuristics across scenarios scheduling period length, # of jobs in the backlog, the amount of certainty in processing times and M/C failures

53. The impact of rescheduling policies… Bean * Matchup algorithm leads to better {less total tardiness} Wu * Robust, partial schedule leads to better {less weighted tardiness} * Processing time variability increases, dispatching rules lead better performance Mehta & Uzsoy * Predictive schedules (idle time) increase predictability but do not significantly degrade {max lateness}

54. The impact of rescheduling policies… Kim & Kim and Sabuncuoglu & Karabuk * Advantage to check the performance periodically * Too-long monitoring periods and too-frequent monitoring negatively affect performance Shafaei & Brunn * Under loose due date conditions, the performance is not sensitive to changes in rescheduling interval. * Under tight due date conditions, the rescheduling interval had a much more significant effect on performance.

55. The impact of rescheduling policies… Through better coordination a longer rescheduling period can improve performance Herrmann & Delaio Material inexpensive  decrease freq. Reduce costs Material expensive  changing freq. Does not affect costs much.

56. Scheduling theory & practice… Understanding the rescheduling can address the gap btw theory and practice. Theory has had limited impact on practice since: * results do not consider important characteristics of the environment * the dynamic aspects of the manufacturing system has not been considered fully !!! Solving scheduling problems is IMPORTANT for controlling dynamic, stochastic systems.

57. Scheduling theory & practice… Importance of planning period: * Job cycle times>planning period  careful scheduling * Job cycle times<planning period  scheduling is seldom important & satisfying the production target should set the constraints and objectives [Portougal & Robb]

58. Scheduling theory & practice… Principles that explain practical scheduling processes: * generates partial solutions for partial problems * anticipates, reacts to and adjusts for disturbances * Is sensitive to and adjusts to the meaning of time in the production situation

59. Summary & conclusion… Papers formulate scheduling as a combinatorial optimization problem. Describe: * algorithms for generating or updating schedules * new rescheduling policies that specify when schedules are generated and updated * studies on dispatching rules, optimal control policies, rescheduling strategies

60. Summary & conclusion… This paper standardizes the rescheduling concepts and shows the gap btw theory and practice. Key results: * Rescheduling policy needs to be considered in system design. However existing models provide little support for rescheduling.

61. Summary & conclusion… * More research is needed to  compare the performance of systems under predictive-reactive policies to their performance under dynamic scheduling.  understand how interactions btw rescheduling policies and other production planning functions affect performance.