Operations Sequencing/ Scheduling (Week 6)

Outline
Work Center Defined Objectives of Sequencing & Scheduling Priority rules for Sequencing for 1-Machine/N-jobs Johnsons Algorithm for 2-Machine/N-Jobs Modified Johnsons Algorithm for 3-Machine/NJobs Algorithm for N-Machine/N-Jobs Shop Floor Control Functions Input/ Output Control Sequence dependent set-up time Heuristic solution
2

Work Center Defined

A work center is an area in a business in which productive resources are organized and work is completed Can be a single machine, a group of machines, or an area where a particular type of work is done

Objectives of Sequencing and Scheduling

Meet due dates Minimize lead time Minimize setup time or cost Minimize work-in-process inventory Maximize machine utilization
4

Priority Rules for Job Sequencing (1-Machine/N-Jobs)

1. First-come, first-served (FCFS) 2. Shortest processing time (SPT) 3. Earliest due date (EDD) 4. Critical ratio (CR) 5. Last come, first served (LCFS)
5

Example of Job Sequencing: FCFS

Suppose you have the four jobs to the right arrive for processing on one machine
Jobs (in order of arrival) A B C D Processing Due Date Time (days) (days hence) 4 5 7 10 3 6 1 4

What is the FCFS schedule?

Do all the jobs get done on time? No, Jobs B, C, and D are going to be late

Answer: FCFS Schedule

4 + 11 + 14 + 15 Mean flow time= = 11days 4

Average lateness=
0 + 1 + 8 + 11 = 5days 4

Example of Job Sequencing: SPT

Suppose you have the four jobs to the right arrive for processing on one machine
Jobs (in order of arrival) A B C D Processing Due Date Time (days) (days hence) 4 5 7 10 3 6 1 4

What is the SPT schedule?

Do all the jobs get done on time?

Answer: Shortest Operating Time Schedule

Jobs (in order of arrival) D C A B Processing Due Date Flow Time (days) Time (days) (days hence) 1 4 1 3 6 4 4 5 8 7 10 15

No, Jobs A and B are going to be late

Example of Job Sequencing: EDD

Suppose you have the four jobs to the right arrive for processing on one machine
Jobs (in order of arrival) A B C D Processing Due Date Time (days) (days hence) 4 5 7 10 3 6 1 4

What is the earliest due date first schedule?

Do all the jobs get done on time?

Answer: Earliest Due Date First

Jobs (in order of arrival) D A C B Processing Due Date Flow Time Time (days) (days hence) (days) 1 4 1 4 5 5 3 6 8 7 10 15

No, Jobs C and B are going to be late

Example of Job Sequencing: CRM

Suppose you have the four jobs to the right arrive for processing on one machine
What is the CR schedule?
Jobs (in order of arrival) A B C D Processing Due Date Time (days) (days hence) 4 5 7 10 3 6 1 4

Do all the jobs get done on time?

In order to do this schedule, the CRs have to be calculated for each job. If we let today be Day 1 and allow a total of 15 days to do the work. The resulting CRs and order schedule are: CR(A)=(5-4)/15=0.06 (Do this job last) CR(B)=(10-7)/15=0.20 (Do this job first, tied with C and D) CR(C)=(6-3)/15=0.20 (Do this job first, tied with B and D) CR(D)=(4-1)/15=0.20 (Do this job first, tied with B and C)

No, but since there is threeway tie, only the first job or two will be on time

10

Example of Job Sequencing: LCFS

Suppose you have the four jobs to the right arrive for processing on one machine
Jobs (in order of arrival) A B C D Processing Due Date Time (days) (days hence) 4 5 7 10 3 6 1 4

What is the LCFS schedule?

Do all the jobs get done on time?

Answer: Last-Come First-Served Schedule

Jobs (in order of arrival) D C B A Processing Time (days) 1 3 7 4 Due Date Flow Time (days hence) (days) 4 1 6 4 10 11 5 15

No, Jobs B and A are going to be late

11

Comparison of Priority Rules

Rule Total flow time (days) 15 15 15 15 Mean flow time (days) 11 7 7.25 7.75 Average lateness (days) 5 2 1.75 2.75

FCFS SOT DDATE LCFS

SOT is better than rest of the rules,. It is called most important concept in entire subject of sequencing.

12

Job Sequencing: Johnsons Rule (2-Machine/N-Jobs)

Suppose you have the following five jobs with time requirements in two stages of production. What is the job sequence using Johnsons Rule?

Jobs A B C D

Time in Hours Stage 1 Stage 2 1.50 1.25 2.00 3.00 2.50 2.00 1.00 2.00
13

Job Sequencing: Johnsons Rule

First, select the job with the smallest time in either stage. That is Job D with the smallest time in the first stage. Place that job as early as possible in the unfilled job sequence below. Jobs A B C D Time in Hours Stage 1 Stage 2 1.50 1.25 2.00 3.00 2.50 2.00 1.00 2.00

Drop D out, select the next smallest time (Job A), and place it 4th in the job sequence.

Drop A out, select the next smallest time. There is a tie in two stages for two different jobs. In this case, place the job with the smallest time in the first stage as early as possible in the unfilled job sequence. Then place the job with the smallest time in the second stage as late as possible in the unfilled sequence. Job Sequence 1 Job Assigned D 2 B 3 C 4 A

14

Shop-Floor Control:Major Functions

1. Assigning priority of each shop order 2. Maintaining work-in-process quantity information 3. Conveying shop-order status information to the office
15

4. Providing actual output data for capacity control purposes 5. Providing quantity by location by shop order for WIP inventory and accounting purposes 6. Providing measurement of efficiency, utilization, and productivity of manpower and machines
16

Input/Output Control
Input Work Center Output

Planned input should never exceed planned output

Focuses attention on bottleneck work centers

17

Principles of Work Center Scheduling

1. There is a direct equivalence between work flow and cash flow 2. The effectiveness of any job shop should be measured by speed of flow through the shop 3. Schedule jobs as a string, with process steps back-to-back 4. A job once started should not be interrupted
18

5. Speed of flow is most efficiently achieved by focusing on bottleneck work centers and jobs

6. Reschedule every day

7. Obtain feedback each day on jobs that are not completed at each work center

8. Match work center input information to what the worker can actually do
19

9. When seeking improvement in output, look for incompatibility between engineering design and process execution

10. Certainty of standards, routings, and so forth is not possible in a job shop, but always work towards achieving it
20

Personnel Scheduling in Services

Scheduling consecutive days off Scheduling daily work times Scheduling hourly work times

21

Sequence Dependent Set-up time

Consider a single production facility through which numerous jobs must be processed. Typically, the facility may have to shut down after processing one job in order to set up for the next. Such downtime is termed setup (or changeover) time. The length of this setup time may depend on the next job to be processed or the job just completed. A sequence of similar jobs would be interrupted by less 22 setup time than a sequence of dissimilar jobs.

A typical managerial problem would be to sequence the jobs in such a way as to minimize total setup time. From a combinatorial point of view, this can be a very large model. For example, if there are only 3 jobs to be processed, say A, B, and C, then any of the three could be taken first, with either of the remaining two second and third determined. The 6 possible sequences are displayed as a tree with each branch representing one sequence.
A B C C B A C B C A A B C B A 23

In general, with n jobs, there are n! = n(n-1)(n-2) 1 possible combinations or sequences. For example, 10 jobs produces 10! = 10(9)(8)(7)(6)(5)(4)(3)(2)(1) = 3,268,800 different sequences. Complete enumeration can be used to solve this minimization problem. However, even though this would provide a true optimum, it is not practical even for modest values of n.
24

HEURISTIC SOLUTIONS
For example, consider a machine operator who has three rather long batch jobs to be run on Monday afternoon. The machine is currently idle. For each of these jobs, there From Job is a setup time which includes cleaning 0 the machine from the last job, A setting up the individual components and other B auxiliary equipment for the new job, etc. C
To Job

A 27

B 21 35

C 32 22 46

49 46 12

Since there are only 3! = 3(2)(1) = 6 possible sequences, they can all be enumerated.
SEQUENCE 0 0 0 0 0 0 A A B B C C B C C A A B C B A C B A SETUP TIME 27 + 35 + 46 27 + 22 + 12 21 + 46 + 46 21 + 49 + 22 32 + 46 + 35 32 + 12 + 49 TOTAL (MIN) 108 61 113 92 113 93

As you can see above, the optimal (minimum total setup time) sequence is 0 A C B.

26

A Greedy Heuristic The heuristic rule to be applied to this model is the next best rule, sometimes called a greedy algorithm. 1. At step 1, perform the task with least initial startup time. 2. At each subsequent step, select the task with least setup time, based on the current state. From the previous table, the task with the least initial setup time is B. Hence, the first step is 0 B. According to the greedy algorithm, given that we have just completed B, the task to be selected is C, since the setup for B C is less than for B A.
27

Thus, since we have 0 B with A. Thus, we obtain

C, then we can only finish

Greedy heuristic: 0 B C A Total setup time: 21 + 46 + 46 = 113 Although easy to apply, for sequential decision models, the greedy algorithm does not lead to an optimal solution. However, statistically, for the above type sequencing model, the rule is not bad and will often produce better results than could be obtained by a purely random selection of tasks.
28

A Better Heuristic The following modified heuristic gives even better results: 1. Transform the original setup time data by subtracting the minimum setup time in each column from all other entries in that column.
A 0 A B C 49 - 27 46 - 27 12 - 12 C 19 0 27 - 27 B 21 - 12 35 - 12 C A 32 - 22 0 22 - 22 A 46 - 22 B 22 24 23 0 0 9 10 B C

29

2. Apply the greedy algorithm to this set of transformed data. Doing this, we obtain Best first step: 0 A Best second step: A C Third step: C B And thus the modified heuristic produces the sequence 0 A C B, which was already shown to be optimal for this model. Although this modified heuristic will not always give the optimal solution, it is easy to implement, and in practice, for large models, it often produces good results.

30