pr Ml Alsan AKhtar Hsin Professor, IPE Dept, BUET ad gsample I: Consider the 6/3/ F/F ax problem below. Solve it ps | Actual process Constructed processing times | Jobs Tm | Me Istmachine_[ 2nd macl T 4 1 z | 6 2 3 3 1 4 5 3 10 5 & 2 8 6 4 1 5 | First, let’s check whether conditions in. Eq. (1) hold or not. Here, we have, 6 6 min (Pa) max{Pia}=3; — min (Pa) ist isi i=l * n So, Equation (I): min {Py} =323= nan Pa) holds. Thus, Johnson’s rule is 1 is applicable. We get the following Gantt chart using “constructed Processing times”. This is indeed optimal, bi ut & Ta TaTa] Mi © 19332639 bow bh Hy My 68 4 19 2324 97 31 hs 7 L Ja [aT] 8 7 24 30 33-36 O Ignatt Schrage Algorithm: Branch and Bound il and Schrage Proposed an algorithm or 3 machine flow shop problem. Later 02 *permutation schedule, for branch and bound enumeration, which is applicable , Lomnicki also contributed ideas on ths topic. Tis is 1 basic branching tree, applied earlier in single-machine problem, is similar to this case (in “"all-Schrage algorithm), except that o represents a Partial permutation occurring at the ing of the sequence instead of at the end. In other words, the job sequence is constructed in "ard direction, (instead of a backward direction), For each node of the tree, a lower bound ‘Ne makespan associated with any completion of the corresponding partial sequence a “4 by considering the work remaining on each machine. To ilustate the proved { * @ ‘sein1 IPE Depts a Dr. Ma. Ahsan Akhtar Hasin oem wD Er 1 ii in umber of machines m = 3, let o” denote the set of jobs that are not contained in the Pani) Permutation @. For a given partial sequence 0, let — it jobs in @ (hence the earliest 41 = latest completion time on machine 1 among jobs in & ( Lin which some job j € a’ could begin processing. 42 = latest completion time on machine 2 among jobs in 0. 43 = latest completion time on machine 3 among jobs in 0. ire amount of processing yet required of machine 1 is — fa fea’ | One lower bound on the makespan is — b= a + Dty + min {ly + ba} jee! jeo! | Similarly, a second lower bound based on the processing yet required of machine 2 yi by = det Ste + min (3) j€o" sea’ A third lower bound based on the processing yet required of machine 3 is ~ =H + Sty jeo" Final lower bound, proposed by Ignall and Schrage, is ~ B = max{b,, by, bj} Z The Branch and Bound procedure,“Proposed by Tgnall and Schrage follows jumptrackin, along with Dominance Property, as used in case of single machine sequencing, Example 2: A 4,Job, 3-machine flow shop problem, with processing times as follows. Prepare permutation schedule, using Branch and Bound (B&B) procedure: Jobs [M; | Mz ] My ty | 2 | ti 1 [3 4 [0 2 [arf fs] vee 3 [7] 9 [43 4 [il 2[2 The first node generated by the BAB algo nigh well eomespond to the sub-problem for which job 1 is assigned the first position in sequence, and o!= {2, 3, 4}. 4. @ 80 ijF al-Apsan ARhtar Hasin Professor, IPE Dept, BUET 7 ae i ial sequence I, i.e, Job | first; BS nl / u) a=3 bj =3+284+6=37 “a @=7 42242531 a] q=17 by=174 20237 M2 i So, lower bound LB = max (37, 31,37) =37 M: 2 : 7 7 * Partial sequence 2, i, Job 2 first: 4 wi quell b= 11+20+14=45 ! w= 12 Le =17 M, 7 Me So, LB = max (45, 39, 42} = 45 My ha 12 17 Similarly, for Partial sequence 3, ie, Job 3 first, LB = max {37, 35, 46} = 46 and 4 for Partial sequence 4, ie, Job 4 first, LB = max (37, 41, $2) = 52 Based on the above four lower bound values, we construct a B&B diagram (nodes show which job to come first in the sequence): 31 45 SSG ? Since node 1 (i.e. job 1 first) has the minimum LB, we sub-branch that node, in order to find [hich ob to come in the 2nd position. It wil be either Job 2, 9 Ine 3, oF Job 4. We calculate the lower bound values again, Partial sequence 1-2, i.e. Job 2 the 2nd: M, qu dy qa 14 by =14417414=45 3 14 m= 15 b= 15+214+2=38 h i = 22 by=22415=37 Saag So, lower bound LB = max (45, 38, 37} =45 M; yf 7 17 22 °milarly for Partial Sequence 1-3, i.e. Job 3 the 2nd, LB = max (37, 34,39} =39, and Foe, Pars, ‘ Partial sequence 1-4, i.e, Job 4 the 2nd, LB = max (37, 40, 45} = 45 GyProfessor, IPE Dep, py ty 4 Dr. Md. Ahsan Akhtar Hasin liagram to another j | tend the B&B diagram level, | Based on the above three lower bound values, We | Out of 3 LB values, partial sequence 1-3 (ie. node 13) has the minimum value, So, we Sb-brang, that node again, where Job 2 ot Job 4 will come in the 3rd position. Partial sequence 1-3-2, ic. Job 2 the 3rd: qh ds jy 10 21 Me h 5 | i 37 by=37+2=39 ° 710 19 2 M by = 21 +104 14=45 So, LB = max (45, 36,39} = 45 Ms hy hh 7 1719 32037 } Similarly, for Partial sequence 1-3-4, i.e, Job 4 the 3rd, LB = max {37, 38, 39} =39 Out of 2 LB values, partial sequence 1-3-4 (ie. node 134) has the minimum value. So, we sub. branch that node again, where Job 2 will obviously come the last (because no other jobs lef), @ So, for overall permutation schedule: 1-3-4-2 Makespan M = 39 oF, Fina = 39 OF, Cinax = 39.pr Md. Ahsan Akhtar Hasin Professor, IPE Dept, BUET pumple 3: A 4job, 3-machine flow shop problem, with processing times as follows. Prepare a pemutation schedule, using Branch and Bound (B&B) procedure [Jobs] Mi | Mz | My ta | te | ty | i lees 2b le 3 10 3 7 el dels © Heuristic approaches opiinaum Result Fa» Time wR Ne | which has high level of complexity, thereby leading to exponentially high time requirements, He avoid this major drawback. However, a heuristic approach has also a major limitation: it can not Buarantee optimality. Some constructive heuristics are being discussed here, which can give quick 3h Leasonably good solution for makespan problem, but without any guarantee of optimality. These are: Johnson’s algorithm (discussed earlier) Palmer's method . sin, Gapta’s sitinad ruhuchiye huaristes Campbell, Dudek, Smith (CDS) method Dannenbring method * (a) Nawaz, Enscore and Ham (NEH) method, et @ Palmer’s method gests t Jobs having the strongest tendency to limes in the sequence of operations. For this purpose, Palmer proposed the calculation of slope index Sj for each job, as follows: For job j and machine k:j = J, 2, 3....m: 2 Dek - m= 1h, ft The above equation ean be expressed in other (alternative) way also ~ (m= Dim + (I= BD tj ayaa + (= Staats = (m= 5)tis —(m = 3)t2 = (m= 1th Then, a permutation schedule is constructed in non-increasing (i.e. decreasing) order of slope indices 5; values — S2S>., 2S, Fer 2machine problem (i.e. m = 2), Palmer's heuristies sequences the jobs in non-inereasing ‘ver of (rm — 1) ty — (me — 1) tj oF, (tha — ty), since m= 2. An example follows: aa G